RFC 6693






Internet Research Task Force (IRTF)                          A. Lindgren
Request for Comments: 6693                                          SICS
Category: Experimental                                          A. Doria
ISSN: 2070-1721                                           Technicalities
                                                               E. Davies
                                                        Folly Consulting
                                                               S. Grasic
                                          Lulea University of Technology
                                                             August 2012


  Probabilistic Routing Protocol for Intermittently Connected Networks

Abstract



   This document is a product of the Delay Tolerant Networking Research
   Group and has been reviewed by that group.  No objections to its
   publication as an RFC were raised.

   This document defines PRoPHET, a Probabilistic Routing Protocol using
   History of Encounters and Transitivity.  PRoPHET is a variant of the
   epidemic routing protocol for intermittently connected networks that
   operates by pruning the epidemic distribution tree to minimize
   resource usage while still attempting to achieve the best-case
   routing capabilities of epidemic routing.  It is intended for use in
   sparse mesh networks where there is no guarantee that a fully
   connected path between the source and destination exists at any time,
   rendering traditional routing protocols unable to deliver messages
   between hosts.  These networks are examples of networks where there
   is a disparity between the latency requirements of applications and
   the capabilities of the underlying network (networks often referred
   to as delay and disruption tolerant).  The document presents an
   architectural overview followed by the protocol specification.

Status of This Memo



   This document is not an Internet Standards Track specification; it is
   published for examination, experimental implementation, and
   evaluation.

   This document defines an Experimental Protocol for the Internet
   community.  This document is a product of the Internet Research Task
   Force (IRTF).  The IRTF publishes the results of Internet-related
   research and development activities.  These results might not be
   suitable for deployment.  This RFC represents the consensus of the
   Delay Tolerant Networking Research Group of the Internet Research





Lindgren, et al.              Experimental                      [Page 1]

RFC 6693                         PRoPHET                     August 2012


   Task Force (IRTF).  Documents approved for publication by the IRSG
   are not a candidate for any level of Internet Standard; see Section 2
   of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc6693.

Copyright Notice



   Copyright (c) 2012 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.
































Lindgren, et al.              Experimental                      [Page 2]

RFC 6693                         PRoPHET                     August 2012


Table of Contents



   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   4
     1.1.  Relation to the Delay-Tolerant Networking Architecture  .   7
     1.2.  Applicability of the Protocol . . . . . . . . . . . . . .   8
     1.3.  PRoPHET as Compared to Regular Routing Protocols  . . . .  10
     1.4.  Requirements Notation . . . . . . . . . . . . . . . . . .  11
   2.  Architecture  . . . . . . . . . . . . . . . . . . . . . . . .  11
     2.1.  PRoPHET . . . . . . . . . . . . . . . . . . . . . . . . .  11
       2.1.1.  Characteristic Time Interval  . . . . . . . . . . . .  12
       2.1.2.  Delivery Predictability Calculation . . . . . . . . .  12
       2.1.3.  Optional Delivery Predictability Optimizations  . . .  17
       2.1.4.  Forwarding Strategies and Queueing Policies . . . . .  18
     2.2.  Bundle Protocol Agent to Routing Agent Interface  . . . .  19
     2.3.  PRoPHET Zone Gateways . . . . . . . . . . . . . . . . . .  20
     2.4.  Lower-Layer Requirements and Interface  . . . . . . . . .  21
   3.  Protocol Overview . . . . . . . . . . . . . . . . . . . . . .  22
     3.1.  Neighbor Awareness  . . . . . . . . . . . . . . . . . . .  22
     3.2.  Information Exchange Phase  . . . . . . . . . . . . . . .  23
       3.2.1.  Routing Information Base Dictionary . . . . . . . . .  25
       3.2.2.  Handling Multiple Simultaneous Contacts . . . . . . .  26
     3.3.  Routing Algorithm . . . . . . . . . . . . . . . . . . . .  28
     3.4.  Bundle Passing  . . . . . . . . . . . . . . . . . . . . .  32
       3.4.1.  Custody . . . . . . . . . . . . . . . . . . . . . . .  33
     3.5.  When a Bundle Reaches Its Destination . . . . . . . . . .  33
     3.6.  Forwarding Strategies . . . . . . . . . . . . . . . . . .  34
     3.7.  Queueing Policies . . . . . . . . . . . . . . . . . . . .  36
   4.  Message Formats . . . . . . . . . . . . . . . . . . . . . . .  38
     4.1.  Header  . . . . . . . . . . . . . . . . . . . . . . . . .  39
     4.2.  TLV Structure . . . . . . . . . . . . . . . . . . . . . .  44
     4.3.  TLVs  . . . . . . . . . . . . . . . . . . . . . . . . . .  45
       4.3.1.  Hello TLV . . . . . . . . . . . . . . . . . . . . . .  45
       4.3.2.  Error TLV . . . . . . . . . . . . . . . . . . . . . .  47
       4.3.3.  Routing Information Base Dictionary TLV . . . . . . .  48
       4.3.4.  Routing Information Base TLV  . . . . . . . . . . . .  50
       4.3.5.  Bundle Offer and Response TLVs (Version 2)  . . . . .  51
   5.  Detailed Operation  . . . . . . . . . . . . . . . . . . . . .  55
     5.1.  High-Level State Tables . . . . . . . . . . . . . . . . .  56
     5.2.  Hello Procedure . . . . . . . . . . . . . . . . . . . . .  59
       5.2.1.  Hello Procedure State Tables  . . . . . . . . . . . .  61
     5.3.  Information Exchange Phase  . . . . . . . . . . . . . . .  62
       5.3.1.  State Definitions for the Initiator Role  . . . . . .  66
       5.3.2.  State Definitions for the Listener Role . . . . . . .  71
       5.3.3.  Recommendations for Information Exchange Timer
               Periods . . . . . . . . . . . . . . . . . . . . . . .  77
       5.3.4.  State Tables for Information Exchange . . . . . . . .  78
     5.4.  Interaction with Nodes Using Version 1 of PRoPHET . . . .  92




Lindgren, et al.              Experimental                      [Page 3]

RFC 6693                         PRoPHET                     August 2012


   6.  Security Considerations . . . . . . . . . . . . . . . . . . .  93
     6.1.  Attacks on the Operation of the Protocol  . . . . . . . .  94
       6.1.1.  Black-Hole Attack . . . . . . . . . . . . . . . . . .  94
       6.1.2.  Limited Black-Hole Attack / Identity Spoofing . . . .  95
       6.1.3.  Fake PRoPHET ACKs . . . . . . . . . . . . . . . . . .  95
       6.1.4.  Bundle Store Overflow . . . . . . . . . . . . . . . .  96
       6.1.5.  Bundle Store Overflow with Delivery Predictability
               Manipulation  . . . . . . . . . . . . . . . . . . . .  96
     6.2.  Interactions with External Routing Domains  . . . . . . .  97
   7.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .  97
     7.1.  DTN Routing Protocol Number . . . . . . . . . . . . . . .  98
     7.2.  PRoPHET Protocol Version  . . . . . . . . . . . . . . . .  98
     7.3.  PRoPHET Header Flags  . . . . . . . . . . . . . . . . . .  99
     7.4.  PRoPHET Result Field  . . . . . . . . . . . . . . . . . .  99
     7.5.  PRoPHET Codes for Success and Codes for Failure . . . . .  99
     7.6.  PRoPHET TLV Type  . . . . . . . . . . . . . . . . . . . . 100
     7.7.  Hello TLV Flags . . . . . . . . . . . . . . . . . . . . . 101
     7.8.  Error TLV Flags . . . . . . . . . . . . . . . . . . . . . 101
     7.9.  RIB Dictionary TLV Flags  . . . . . . . . . . . . . . . . 102
     7.10. RIB TLV Flags . . . . . . . . . . . . . . . . . . . . . . 102
     7.11. RIB Flags . . . . . . . . . . . . . . . . . . . . . . . . 103
     7.12. Bundle Offer and Response TLV Flags . . . . . . . . . . . 103
     7.13. Bundle Offer and Response B Flags . . . . . . . . . . . . 104
   8.  Implementation Experience . . . . . . . . . . . . . . . . . . 104
   9.  Deployment Experience . . . . . . . . . . . . . . . . . . . . 105
   10. Acknowledgements  . . . . . . . . . . . . . . . . . . . . . . 105
   11. References  . . . . . . . . . . . . . . . . . . . . . . . . . 105
     11.1. Normative References  . . . . . . . . . . . . . . . . . . 105
     11.2. Informative References  . . . . . . . . . . . . . . . . . 106
   Appendix A.  PRoPHET Example  . . . . . . . . . . . . . . . . . . 108
   Appendix B.  Neighbor Discovery Example . . . . . . . . . . . . . 110
   Appendix C.  PRoPHET Parameter Calculation Example  . . . . . . . 110

1.  Introduction



   The Probabilistic Routing Protocol using History of Encounters and
   Transitivity (PRoPHET) algorithm enables communication between
   participating nodes wishing to communicate in an intermittently
   connected network where at least some of the nodes are mobile.

   One of the most basic requirements for "traditional" (IP) networking
   is that there must exist a fully connected path between communication
   endpoints for the duration of a communication session in order for
   communication to be possible.  There are, however, a number of
   scenarios where connectivity is intermittent so that this is not the
   case (thus rendering the end-to-end use of traditional networking
   protocols impossible), but where it still is desirable to allow
   communication between nodes.



Lindgren, et al.              Experimental                      [Page 4]

RFC 6693                         PRoPHET                     August 2012


   Consider a network of mobile nodes using wireless communication with
   a limited range that is less than the typical excursion distances
   over which the nodes travel.  Communication between a pair of nodes
   at a particular instant is only possible when the distance between
   the nodes is less than the range of the wireless communication.  This
   means that, even if messages are forwarded through other nodes acting
   as intermediate routes, there is no guarantee of finding a viable
   continuous path when it is needed to transmit a message.

   One way to enable communication in such scenarios is by allowing
   messages to be buffered at intermediate nodes for a longer time than
   normally occurs in the queues of conventional routers (cf. Delay-
   Tolerant Networking [RFC4838]).  It would then be possible to exploit
   the mobility of a subset of the nodes to bring messages closer to
   their destination by transferring them to other nodes as they meet.
   Figure 1 shows how the mobility of nodes in such a scenario can be
   used to eventually deliver a message to its destination.  In this
   figure, the four sub-figures (a) - (d) represent the physical
   positions of four nodes (A, B, C, and D) at four time instants,
   increasing from (a) to (d).  The outline around each letter
   represents the range of the radio communication used for
   communication by the nodes: communication is only possible when the
   ranges overlap.  At the start time, node A has a message -- indicated
   by an asterisk (*) next to that node -- to be delivered to node D,
   but there does not exist a path between nodes A and D because of the
   limited range of available wireless connections.  As shown in sub-
   figures (a) - (d), the mobility of the nodes allows the message to
   first be transferred to node B, then to node C, and when finally node
   C moves within range of node D, it can deliver the message to its
   final destination.  This technique is known as "transitive
   networking".

   Mobility and contact patterns in real application scenarios are
   likely to be non-random, but rather be predictable, based on the
   underlying activities of the higher-level application (this could,
   for example, stem from human mobility having regular traffic patterns
   based on repeating behavioral patterns (e.g., going to work or the
   market and returning home) and social interactions, or from any
   number of other node mobility situations where a proportion of nodes
   are mobile and move in ways that are not completely random over time
   but have a degree of predictability over time).  This means that if a
   node has visited a location or been in contact with a certain node
   several times before, it is likely that it will visit that location
   or meet that node again.







Lindgren, et al.              Experimental                      [Page 5]

RFC 6693                         PRoPHET                     August 2012


   PRoPHET can also be used in some networks where such mobility as
   described above does not take place.  Predictable patterns in node
   contacts can also occur among static nodes where varying radio
   conditions or power-saving sleeping schedules cause connection
   between nodes to be intermittent.

   In previously discussed mechanisms to enable communication in
   intermittently connected networks, such as Epidemic Routing
   [vahdat_00], very general approaches have been taken to the problem
   at hand.  In an environment where buffer space and bandwidth are
   infinite, epidemic routing will give an optimal solution to the
   problem of routing in an intermittently connected network with regard
   to message delivery ratio and latency.  However, in most cases,
   neither bandwidth nor buffer space is infinite, but instead they are
   rather scarce resources, especially in the case of sensor networks.

   PRoPHET is fundamentally an epidemic protocol with strict pruning.
   An epidemic protocol works by transferring its data to each and every
   node it meets.  As data is passed from node to node, it is eventually
   passed to all nodes, including the target node.  One of the
   advantages of an epidemic protocol is that by trying every path, it
   is guaranteed to try the best path.  One of the disadvantages of an
   epidemic protocol is the extensive use of resources with every node
   needing to carry every packet and the associated transmission costs.
   PRoPHET's goal is to gain the advantages of an epidemic protocol
   without paying the price in storage and communication resources
   incurred by the basic epidemic protocol.  That is, PRoPHET offers an
   alternative to basic epidemic routing, with lower demands on buffer
   space and bandwidth, with equal or better performance in cases where
   those resources are limited, and without loss of generality in
   scenarios where it is suitable to use PRoPHET.

   In a situation where PRoPHET is applicable, the patterns are expected
   to have a characteristic time (such as the expected time between
   encounters between mobile stations) that is in turn related to the
   expected time that traffic will take to reach its destination in the
   part of the network that is using PRoPHET.  This characteristic time
   provides guidance for configuration of the PRoPHET protocol in a
   network.  When appropriately configured, the PRoPHET protocol
   effectively builds a local model of the expected patterns in the
   network that can be used to optimize the usage of resources by
   reducing the amount of traffic sent to nodes that are unlikely to
   lead to eventual delivery of the traffic to its destination.








Lindgren, et al.              Experimental                      [Page 6]

RFC 6693                         PRoPHET                     August 2012


     +----------------------------+   +----------------------------+
     |                      ___   |   |                      ___   |
     |      ___            /   \  |   |                     /   \  |
     |     /   \          (  D  ) |   |                    (  D  ) |
     |    (  B  )          \___/  |   |     ___             \___/  |
     |     \___/    ___           |   |    /___\    ___            |
     |___          /   \          |   |   (/ B*\)  /   \           |
     |   \        (  C  )         |   |   (\_A_/) (  C  )          |
     | A* )        \___/          |   |    \___/   \___/           |
     |___/                        |   |                            |
     +----------------------------+   +----------------------------+
              (a) Time t                     (b) Time (t + dt)
     +----------------------------+   +----------------------------+
     |        _____         ___   |   |        ___           ___   |
     |       / / \ \       /   \  |   |       /   \         /___\  |
     |      ( (B C* )     (  D  ) |   |      (  B  )       (/ D*\) |
     |       \_\_/_/       \___/  |   |       \___/        (\_C_/) |
     |     ___                    |   |     ___             \___/  |
     |    /   \                   |   |    /   \                   |
     |   (  A  )                  |   |   (  A  )                  |
     |    \___/                   |   |    \___/                   |
     |                            |   |                            |
     +----------------------------+   +----------------------------+
          (c) Time (t + 2*dt)               (d) Time (t + 3*dt)

               Figure 1: Example of transitive communication

   This document presents a framework for probabilistic routing in
   intermittently connected networks, using an assumption of non-random
   mobility of nodes to improve the delivery rate of messages while
   keeping buffer usage and communication overhead at a low level.
   First, a probabilistic metric called delivery predictability is
   defined.  The document then goes on to define a probabilistic routing
   protocol using this metric.

1.1.  Relation to the Delay-Tolerant Networking Architecture



   The Delay-Tolerant Networking (DTN) architecture [RFC4838] defines an
   architecture for communication in environments where traditional
   communication protocols cannot be used due to excessive delays, link
   outages, and other extreme conditions.  The intermittently connected
   networks considered here are a subset of those covered by the DTN
   architecture.  The DTN architecture defines routes to be computed
   based on a collection of "contacts" indicating the start time,
   duration, endpoints, forwarding capacity, and latency of a link in
   the topology graph.  These contacts may be deterministic or may be





Lindgren, et al.              Experimental                      [Page 7]

RFC 6693                         PRoPHET                     August 2012


   derived from estimates.  The architecture defines some different
   types of intermittent contacts.  The ones called "opportunistic" and
   "predicted" are the ones addressed by this protocol.

   Opportunistic contacts are those that are not scheduled, but rather
   present themselves unexpectedly and frequently arise due to node
   mobility.  Predicted contacts are like opportunistic contacts, but,
   based on some information, it might be possible to draw some
   statistical conclusion as to whether or not a contact will be present
   soon.

   The DTN architecture also introduces the bundle protocol [RFC5050],
   which provides a way for applications to "bundle" an entire session,
   including both data and metadata, into a single message, or bundle,
   that can be sent as a unit.  The bundle protocol also provides end-
   to-end addressing and acknowledgments.  PRoPHET is specifically
   intended to provide routing services in a network environment that
   uses bundles as its data transfer mechanism but could be also be used
   in other intermittent environments.

1.2.  Applicability of the Protocol



   The PRoPHET routing protocol is mainly targeted at situations where
   at least some of the nodes are mobile in a way that creates
   connectivity patterns that are not completely random over time but
   have a degree of predictability.  Such connectivity patterns can also
   occur in networks where nodes switch off radios to preserve power.
   Human mobility patterns (often containing daily or weekly periodic
   activities) provide one such example where PRoPHET is expected to be
   applicable, but the applicability is not limited to scenarios
   including humans.

   In order for PRoPHET to benefit from such predictability in the
   contact patterns between nodes, it is expected that the network exist
   under similar circumstances over a longer timescale (in terms of node
   encounters) so that the predictability can be accurately estimated.

   The PRoPHET protocol expects nodes to be able to establish a local
   TCP link in order to exchange the information needed by the PRoPHET
   protocol.  Protocol signaling is done out-of-band over this TCP link,
   without involving the bundle protocol agent [RFC5050].  However, the
   PRoPHET protocol is expected to interact with the bundle protocol
   agent to retrieve information about available bundles as well as to
   request that a bundle be sent to another node (it is expected that
   the associated bundle protocol agents are then able to establish a
   link (probably over the TCP convergence layer [CLAYER]) to perform
   this bundle transfer).




Lindgren, et al.              Experimental                      [Page 8]

RFC 6693                         PRoPHET                     August 2012


   TCP provides a reliable bidirectional channel between two peers and
   guarantees in-order delivery of transmitted data.  When using TCP,
   the guarantee of reliable, in-order delivery allows information
   exchanges of each category of information to be distributed across
   several messages without requiring the PRoPHET protocol layer to be
   concerned that all messages have been received before starting the
   exchange of the next category of information.  At most, the last
   message of the category needs to be marked as such.  This allows the
   receiver to process earlier messages while waiting for additional
   information and allows implementations to limit the size of messages
   so that IP fragmentation will be avoided and memory usage can be
   optimized if necessary.  However, implementations MAY choose to build
   a single message for each category of information that is as large as
   necessary and rely on TCP to segment the message.

   While PRoPHET is currently defined to run over TCP, in future
   versions the information exchange may take place over other transport
   protocols, and these may not provide message segmentation or
   reliable, in-order delivery.  The simple message division used with
   TCP MUST NOT be used when the underlying transport does not offer
   reliable, in-order delivery, as it would be impossible to verify that
   all the messages had arrived.  Hence, the capability is provided to
   segment protocol messages into submessages directly in the PRoPHET
   layer.  Submessages are provided with sequence numbers, and this,
   together with a capability for positive acknowledgements, would allow
   PRoPHET to operate over an unreliable protocol such as UDP or
   potentially directly over IP.

   Since TCP offers reliable delivery, it is RECOMMENDED that the
   positive acknowledgment capability is not used when PRoPHET is run
   over a TCP transport or similar protocol.  When running over TCP,
   implementations MAY safely ignore positive acknowledgments.

   Whatever transport protocol is used, PRoPHET expects to use a
   bidirectional link for the information exchange; this allows for the
   information exchange to take place in both directions over the same
   link avoiding the need to establish a second link for information
   exchange in the reverse direction.

   In a large Delay- and Disruption-Tolerant Network (DTN), network
   conditions may vary widely, and in different parts of the network,
   different routing protocols may be appropriate.  In this
   specification, we consider routing within a single "PRoPHET zone",
   which is a set of nodes among which messages are routed using
   PRoPHET.  In many cases, a PRoPHET zone will not span the entire DTN,
   but there will be other parts of the network with other
   characteristics that run other routing protocols.  To handle this,
   there may be nodes within the zone that act as gateways to other



Lindgren, et al.              Experimental                      [Page 9]

RFC 6693                         PRoPHET                     August 2012


   nodes that are the destinations for bundles generated within the zone
   or that insert bundles into the zone.  Thus, PRoPHET is not
   necessarily used end-to-end, but only within regions of the network
   where its use is appropriate.

1.3.  PRoPHET as Compared to Regular Routing Protocols



   While PRoPHET uses a mechanism for pruning the epidemic forwarding
   tree that is similar to the mechanism used in metric-based vector
   routing protocols (where the metric might be distance or cost), it
   should not be confused with a metric vector protocol.

   In a traditional metric-based vector routing protocol, the
   information passed from node to node is used to create a single non-
   looping path from source to destination that is optimal given the
   metric used.  The path consists of a set of directed edges selected
   from the complete graph of communications links between the network
   nodes.

   In PRoPHET, that information is used to prune the epidemic tree of
   paths by removing paths that look less likely to provide an effective
   route for delivery of data to its intended destination.  One of the
   effects of this difference is that the regular notions of split
   horizon, as described in [RFC1058], do not apply to PRoPHET.  The
   purpose of split horizon is to prevent a distance vector protocol
   from ever passing a packet back to the node that sent it the packet
   because it is well known that the source does not lie in that
   direction as determined when the directed path was computed.

   In an epidemic protocol, where that previous system already has the
   data, the notion of passing the data back to the node is redundant:
   the protocol can readily determine that such a transfer is not
   required.  Further, given the mobility and constant churn of
   encounters possible in a DTN that is dominated by opportunistic
   encounters, it is quite possible that, on a future encounter, the
   node might have become a better option for reaching the destination.
   Such a later encounter may require a re-transfer of the data if
   resource constraints have resulted in the data being deleted from the
   original carrier between the encounters.

   The logic of metric routing protocols does not map directly onto the
   family of epidemic protocols.  In particular, it is inappropriate to
   try to assess such protocols against the criteria used to assess
   conventional routing protocols such as the metric vector protocols;
   this is not to say that the family of epidemic protocols do not have
   weaknesses but they have to be considered independently of
   traditional protocols.




Lindgren, et al.              Experimental                     [Page 10]

RFC 6693                         PRoPHET                     August 2012


1.4.  Requirements Notation



   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in RFC 2119 [RFC2119].

2.  Architecture



2.1.  PRoPHET



   This section presents an overview of the main architecture of
   PRoPHET, a Probabilistic Routing Protocol using History of Encounters
   and Transitivity.  The protocol leverages the observations made on
   the non-randomness of mobility patterns present in many application
   scenarios to improve routing performance.  Instead of doing blind
   epidemic replication of bundles through the network as previous
   protocols have done, it applies "probabilistic routing".

   To accomplish this, a metric called "delivery predictability",
   0 <= P_(A,B) <= 1, is established at every node A for each known
   destination B.  This metric is calculated so that a node with a
   higher value for a certain destination is estimated to be a better
   candidate for delivering a bundle to that destination (i.e., if
   P_(A,B)>P_(C,B), bundles for destination B are preferable to forward
   to A rather than C).  It is later used when making forwarding
   decisions.  As routes in a DTN are likely to be asymmetric, the
   calculation of the delivery predictability reflects this, and P_(A,B)
   may be different from P_(B,A).

   The delivery predictability values in each node evolve over time both
   as a result of decay of the metrics between encounters between nodes
   and due to changes resulting from encounters when metric information
   for the encountered node is updated to reflect the encounter and
   metric information about other nodes is exchanged.

   When two PRoPHET nodes have a communication opportunity, they
   initially enter a two-part Information Exchange Phase (IEP).  In the
   first part of the exchange, the delivery predictabilities for all
   destinations known by each node are shared with the encountered node.
   The exchanged information is used by each node to update the internal
   delivery predictability vector as described below.  After that, the
   nodes exchange information (including destination and size) about the
   bundles each node carries, and the information is used in conjunction
   with the updated delivery predictabilities to decide which bundles to
   request to be forwarded from the other node based on the forwarding
   strategy used (as discussed in Section 2.1.4).  The forwarding of
   bundles is carried out in the latter part of the Information Exchange
   Phase.



Lindgren, et al.              Experimental                     [Page 11]

RFC 6693                         PRoPHET                     August 2012


2.1.1.  Characteristic Time Interval



   When an application scenario makes PRoPHET applicable, the mobility
   pattern will exhibit a characteristic time interval that reflects the
   distribution of time intervals between encounters between nodes.  The
   evolution of the delivery predictabilities, which reflects this
   mobility pattern, should reflect this same characteristic time
   interval.  Accordingly, the parameters used in the equations that
   specify the evolution of delivery predictability (see Section 2.1.2)
   need to be configured appropriately so that the evolution reflects a
   model of the mobility pattern.

2.1.2.  Delivery Predictability Calculation



   As stated above, PRoPHET relies on calculating a metric based on the
   probability of encountering a certain node, and using that to support
   the decision of whether or not to forward a bundle to a certain node.
   This section describes the operations performed on the metrics stored
   in a node when it encounters another node and a communications
   opportunity arises.  In the operations described by the equations
   that follow, the updates are being performed by node A, P_(A,B) is
   the delivery predictability value that node A will have stored for
   the destination B after the encounter, and P_(A,B)_old is the
   corresponding value that was stored before the encounter.  If no
   delivery predictability value is stored for a particular destination
   B, P_(A,B) is considered to be zero.

   As a special case, the metric value for a node itself is always
   defined to be 1 (i.e., P_(A,A)=1).

   The equations use a number of parameters that can be selected to
   match the characteristics of the mobility pattern in the PRoPHET zone
   where the node is located (see Section 2.1.1).  Recommended settings
   for the various parameters are given in Section 3.3.  The impact on
   the evolution of delivery predictabilities if encountering nodes have
   different parameter setting is discussed in Section 2.1.2.1.

   The calculation of the updates to the delivery predictabilities
   during an encounter has three parts.

   When two nodes meet, the first thing they do is to update the
   delivery predictability for each other, so that nodes that are often
   encountered have a high delivery predictability.  If node B has not
   met node A for a long time or has never met node B, such that
   P_(A,B) < P_first_threshold, then P_(A,B) should be set to
   P_encounter_first.  Because PRoPHET generally has no prior knowledge
   about whether this is an encounter that will be repeated relatively
   frequently or one that will be a rare event, P_encounter_first SHOULD



Lindgren, et al.              Experimental                     [Page 12]

RFC 6693                         PRoPHET                     August 2012


   be set to 0.5 unless the node has extra information obtained other
   than through the PRoPHET protocol about the likelihood of future
   encounters.  Otherwise, P_(A,B) should be calculated as shown in
   Equation 1, where 0 <= P_encounter <= 1 is a scaling factor setting
   the rate at which the predictability increases on encounters after
   the first, and delta is a small positive number that effectively sets
   an upper bound for P_(A,B).  The limit is set so that
   predictabilities between different nodes stay strictly less than 1.
   The value of delta should normally be very small (e.g., 0.01) so as
   not to significantly restrict the range of available
   predictabilities, but it can be chosen to make calculations efficient
   where this is important.

   P_(A,B) =
   P_(A,B)_old + ( 1 - delta - P_(A,B)_old ) * P_encounter  (Eq. 1)

   There are practical circumstances where an encounter that is
   logically a single encounter in terms of the proximity of the node
   hardware and/or from the point of view of the human users of the
   nodes results in several communication opportunities closely spaced
   in time.  For example, mobile nodes communicating with each other
   using Wi-Fi ad hoc mode may produce apparent multiple encounters with
   a short interval between them but these are frequently due to
   artifacts of the underlying physical network when using wireless
   connections, where transmission problems or small changes in location
   may result in repeated reconnections.  In this case, it would be
   inappropriate to increase the delivery predictability by the same
   amount for each opportunity as it would be increased when encounters
   occur at longer intervals in the normal mobility pattern.

   In order to reduce the distortion of the delivery predictability in
   these circumstances, P_encounter is a function of the interval since
   the last encounter resulted in an update of the delivery
   predictabilities.  The form of the function is as shown in Figure 2.

















Lindgren, et al.              Experimental                     [Page 13]

RFC 6693                         PRoPHET                     August 2012


              P_encounter
                   ^
                   |
   P_encounter_max +  -  - .-------------------------------------
                   |      /
                   |     / .
                   |    /
                   |   /   .
                   |  /
                   | /     .
                   |/
                   +-------+-------------------------------------> I
                          I_typ

          Figure 2: P_encounter as function of time interval, I,
                              between updates

   The form of the function is chosen so that both the increase of
   P_(A,B) resulting from Equation 1 and the decrease that results from
   Equation 2 are related to the interval between updates for short
   intervals.  For intervals longer than the "typical" time (I_typ)
   between encounters, P_encounter is set to a fixed value
   P_encounter_max.  The break point reflects the transition between the
   "normal" communication opportunity regime (where opportunities result
   from the overall mobility pattern) and the closely spaced
   opportunities that result from what are effectively local artifacts
   of the wireless technology used to deliver those opportunities.

   P_encounter_max is chosen so that the increment in P_(A,B) provided
   by Equation 1 significantly exceeds the decay of the delivery
   predictability over the typical interval between encounters resulting
   from Equation 2.

   Making P_encounter dependent on the interval time also avoids
   inappropriate extra increments of P_(A,B) in situations where node A
   is in communication with several other nodes simultaneously.  In this
   case, updates from each of the communicating nodes have to be
   distributed to the other nodes, possibly leading to several updates
   being carried out in a short period.  This situation is discussed in
   more detail in Section 3.2.2.

   If a pair of nodes do not encounter each other during an interval,
   they are less likely to be good forwarders of bundles to each other,
   thus the delivery predictability values must age, being reduced in
   the process.  The second part of the updates of the metric values is
   application of the aging equation shown in Equation 2, where
   0 <= gamma <= 1 is the aging constant, and K is the number of time
   units that have elapsed since the last time the metric was aged.  The



Lindgren, et al.              Experimental                     [Page 14]

RFC 6693                         PRoPHET                     August 2012


   time unit used can differ and should be defined based on the
   application and the expected delays in the targeted network.

   P_(A,B) = P_(A,B)_old * gamma^K  (Eq. 2)

   The delivery predictabilities are aged according to Equation 2 before
   being passed to an encountered node so that they reflect the time
   that has passed since the node had its last encounter with any other
   node.  The results of the aging process are sent to the encountered
   peer for use in the next stage of the process.  The aged results
   received from node B in node A are referenced as P_(B,x)_recv.

   The delivery predictability also has a transitive property that is
   based on the observation that if node A frequently encounters node B,
   and node B frequently encounters node C, then node C probably is a
   good node to which to forward bundles destined for node A.
    Equation 3 shows how this transitivity affects the delivery
   predictability, where 0 <= beta <= 1 is a scaling constant that
   controls how large an impact the transitivity should have on the
   delivery predictability.

   P_(A,C) = MAX( P_(A,C)_old, P_(A,B) * P_(B,C)_recv * beta )  (Eq. 3)

   Node A uses Equation 3 and the metric values received from the
   encountered node B (e.g., P_(B,C)_recv) in the third part of updating
   the metric values stored in node A.

2.1.2.1.  Impact of Encounters between Nodes with Different Parameter
          Settings



   The various parameters used in the three equations described in
   Section 2.1.2 are set independently in each node, and it is therefore
   possible that encounters may take place between nodes that have been
   configured with different values of the parameters.  This section
   considers whether this could be problematic for the operation of
   PRoPHET in that zone.

   It is desirable that all the nodes operating in a PRoPHET zone should
   use closely matched values of the parameters and that the parameters
   should be set to values that are appropriate for the operating zone.
   More details of how to select appropriate values are given in
   Section 3.3.  Using closely matched values means that delivery
   predictabilities will evolve in the same way in each node, leading to
   consistent decision making about the bundles that should be exchanged
   during encounters.






Lindgren, et al.              Experimental                     [Page 15]

RFC 6693                         PRoPHET                     August 2012


   Before going on to consider the impact of reasonable but different
   settings, it should be noted that malicious nodes can use
   inappropriate settings of the parameters to disrupt delivery of
   bundles in a PRoPHET zone as described in Section 6.

   Firstly and importantly, use of different, but legitimate, settings
   in encountering nodes will not cause problems in the protocol itself.
   Apart from P_encounter_first, the other parameters control the rate
   of change of the metric values or limit the range of valid values
   that will be stored in a node.  None of the calculations in a node
   will be invalidated or result in illegal values if the metric values
   received from another node were calculated using different
   parameters.  Furthermore, the protocol is designed so that it is not
   possible to carry delivery predictabilities outside the permissible
   range of 0 to 1.

   A node MAY consider setting received values greater than (1 - delta)
   to (1 - delta) if this would simplify operations.  However, there are
   some special situations where it may be appropriate for the delivery
   predictability for another node to be 1.  For example, if a DTN using
   PRoPHET has multiple gateways to the continuously connected Internet,
   the delivery predictability seen from PRoPHET in one gateway for the
   other gateway nodes can be taken as 1 since they are permanently
   connected through the Internet.  This would allow traffic to be
   forwarded into the DTN through the most advantageous gateway even if
   it initially arrives at another gateway.

   Simulation work indicates that the update calculations are quite
   stable in the face of changes to the rate parameters, so that minor
   discrepancies will not have a major impact on the performance of the
   protocol.  The protocol is explicitly designed to deal with
   situations where there are random factors in the opportunistic nature
   of node encounters, and this randomness dominates over the
   discrepancies in the parameters.

   More major discrepancies may lead to suboptimal behavior of the
   protocol, as certain paths might be more preferred or more deprecated
   inappropriately.  However, since the protocol overall is epidemic in
   nature, this would not generally lead to non-delivery of bundles, as
   they would also be passed to other nodes and would still be
   delivered, though possibly not on the optimal path.










Lindgren, et al.              Experimental                     [Page 16]

RFC 6693                         PRoPHET                     August 2012


2.1.3.  Optional Delivery Predictability Optimizations



2.1.3.1.  Smoothing



   To give the delivery predictability a smoother rate of change, a node
   MAY apply one of the following methods:

   1.  Keep a list of NUM_P values for each destination instead of only
       a single value.  (The recommended value is 4, which has been
       shown in simulations to give a good trade-off between smoothness
       and rate of response to changes.)  The list is held in order of
       acquisition.  When a delivery predictability is updated, the
       value at the "newest" position in the list is used as input to
       the equations in Section 2.1.2.  The oldest value in the list is
       then discarded and the new value is written in the "newest"
       position of the list.  When a delivery predictability value is
       needed (either for sending to a peering PRoPHET node, or for
       making a forwarding decision), the average of the values in the
       list is calculated, and that value is then used.  If less than
       NUM_P values have been entered into the list, only the positions
       that have been filled should be used for the averaging.

   2.  In addition to keeping the delivery predictability as described
       in Section 2.1.2, a node MAY also keep an exponential weighted
       moving average (EWMA) of the delivery predictability.  The EWMA
       is then used to make forwarding decisions and to report to
       peering nodes, but the value calculated according to
       Section 2.1.2 is still used as input to the calculations of new
       delivery predictabilities.  The EWMA is calculated according to
       Equation 4, where 0 <= alpha <= 1 is the weight of the most
       current value.

   P_ewma = P_ewma_old * (1 - alpha) + P * alpha  (Eq. 4)

   The appropriate choice of alpha may vary depending on application
   scenario circumstances.  Unless prior knowledge of the scenario is
   available, it is suggested that alpha is set to 0.5.

2.1.3.2.  Removal of Low Delivery Predictabilities



   To reduce the data to be transferred between two nodes, a node MAY
   treat delivery predictabilities smaller than P_first_threshold, where
   P_first_threshold is a small number, as if they were zero, and thus
   they do not need to be stored or included in the list sent during the
   Information Exchange Phase.  If this optimization is used, care must
   be taken to select P_first_threshold to be smaller than delivery
   predictability values normally present in the network for
   destinations for which this node is a forwarder.  It is possible that



Lindgren, et al.              Experimental                     [Page 17]

RFC 6693                         PRoPHET                     August 2012


   P_first_threshold could be calculated based on delivery
   predictability ranges and the amount they change historically, but
   this has not been investigated yet.

2.1.4.  Forwarding Strategies and Queueing Policies



   In traditional routing protocols, choosing where to forward a message
   is usually a simple task; the message is sent to the neighbor that
   has the path to the destination with the lowest cost (often the
   shortest path).  Normally, the message is also sent to only a single
   node since the reliability of paths is relatively high.  However, in
   the settings we envision here, things are radically different.  The
   first possibility that must be considered when a bundle arrives at a
   node is that there might not be a path to the destination available,
   so the node has to buffer the bundle, and upon each encounter with
   another node, the decision must be made whether or not to transfer a
   particular bundle.  Furthermore, having duplicates of messages (on
   different nodes, as the bundle offer/request mechanism described in
   Section 4.3.5 ensures that a node does not receive a bundle it
   already carries) may also be sensible, as forwarding a bundle to
   multiple nodes can increase the delivery probability of that bundle.

   Unfortunately, these decisions are not trivial to make.  In some
   cases, it might be sensible to select a fixed threshold and only give
   a bundle to nodes that have a delivery predictability over that
   threshold for the destination of the bundle.  On the other hand, when
   encountering a node with a low delivery predictability, it is not
   certain that a node with a higher metric will be encountered within a
   reasonable time.  Thus, there can also be situations where we might
   want to be less strict in deciding who to give bundles to.
   Furthermore, there is the problem of deciding how many nodes to give
   a certain bundle to.  Distributing a bundle to a large number of
   nodes will of course increase the probability of delivering that
   particular bundle to its destination, but this comes at the cost of
   consuming more system resources for bundle storage and possibly
   reducing the probability of other bundles being delivered.  On the
   other hand, giving a bundle to only a few nodes (maybe even just a
   single node) will use less system resources, but the probability of
   delivering a bundle is lower, and the delay incurred is high.

   When resources are constrained, nodes may suffer from storage
   shortage, and may have to drop bundles before they have been
   delivered to their destinations.  They may also wish to consider the
   length of bundles being offered by an encountered node before
   accepting transfer of the bundle in order to avoid the need to drop
   the new bundle immediately or to ensure that there is adequate space
   to hold the bundle offered, which might require other bundles to be
   dropped.  As with the decision as to whether or not to forward a



Lindgren, et al.              Experimental                     [Page 18]

RFC 6693                         PRoPHET                     August 2012


   bundle, deciding which bundles to accept and/or drop to still
   maintain good performance might require different policies in
   different scenarios.

   Nodes MAY define their own forwarding strategies and queueing
   policies that take into account the special conditions applicable to
   the nodes, and local resource constraints.  Some default strategies
   and policies that should be suitable for most normal operations are
   defined in Section 3.6 and Section 3.7.

2.2.  Bundle Protocol Agent to Routing Agent Interface



   The bundle protocol [RFC5050] introduces the concept of a "bundle
   protocol agent" that manages the interface between applications and
   the "convergence layers" that provide the transport of bundles
   between nodes during communication opportunities.  This specification
   extends the bundle protocol agent with a routing agent that controls
   the actions of the bundle protocol agent during an (opportunistic)
   communications opportunity.

   This specification defines the details of the PRoPHET routing agent,
   but the interface defines a more general interface that is also
   applicable to alternative routing protocols.

   To enable the PRoPHET routing agent to operate properly, it must be
   aware of the bundles stored at the node, and it must also be able to
   tell the bundle protocol agent of that node to send a bundle to a
   peering node.  Therefore, the bundle protocol agent needs to provide
   the following interface/functionality to the routing agent:

   Get Bundle List
        Returns a list of the stored bundles and their attributes to the
        routing agent.

   Send Bundle
        Makes the bundle protocol agent send a specified bundle.

   Accept Bundle
        Gives the bundle protocol agent a new bundle to store.

   Bundle Delivered
        Tells the bundle protocol agent that a bundle was delivered to
        its destination.

   Drop Bundle Advice
        Advises the bundle protocol agent that a specified bundle should
        not be offered for forwarding in future and may be dropped by
        the bundle protocol agent if appropriate.



Lindgren, et al.              Experimental                     [Page 19]

RFC 6693                         PRoPHET                     August 2012


   Route Import
        Can be used by a gateway node in a PRoPHET zone to import
        reachability information about endpoint IDs (EIDs) that are
        external to the PRoPHET zone.  Translation functions dependent
        on the external routing protocol will be used to set the
        appropriate delivery predictabilities for imported destinations
        as described in Section 2.3.

   Route Export
        Can be used by a gateway node in a PRoPHET zone to export
        reachability information (destination EIDs and corresponding
        delivery predictabilities) for use by routing protocols in other
        parts of the DTN.

      Implementation Note: Depending on the distribution of functions in
      a complete bundle protocol agent supporting PRoPHET, reception and
      delivery of bundles may not be carried out directly by the PRoPHET
      module.  In this case, PRoPHET can inform the bundle protocol
      agent about bundles that have been requested from communicating
      nodes.  Then, the Accept Bundle and Bundle Delivered functions can
      be implemented as notifications of the PRoPHET module when the
      relevant bundles arrive at the node or are delivered to local
      applications.

2.3.  PRoPHET Zone Gateways



   PRoPHET is designed to handle routing primarily within a "PRoPHET
   zone", i.e., a set of nodes that all implement the PRoPHET routing
   scheme.  However, since we recognize that a PRoPHET routing zone is
   unlikely to encompass an entire DTN, there may be nodes within the
   zone that act as gateways to other nodes that are the destinations
   for bundles generated within the zone or that insert bundles into the
   zone.

   PRoPHET MAY elect to export and import routes across a bundle
   protocol agent interface.  The delivery predictability to use for
   routes that are imported depends on the routing protocol used to
   manage those routes.  If a translation function between the external
   routing protocol and PRoPHET exists, it SHOULD be used to set the
   delivery predictability.  If no such translation function exists, the
   delivery predictability SHOULD be set to 1.  For those routes that
   are exported, the current delivery predictability will be exported
   with the route.








Lindgren, et al.              Experimental                     [Page 20]

RFC 6693                         PRoPHET                     August 2012


2.4.  Lower-Layer Requirements and Interface



   PRoPHET can be run on a large number of underlying networking
   technologies.  To accommodate its operation on all kinds of lower
   layers, it requires the lower layers to provide the following
   functionality and interfaces.

   Neighbor discovery and maintenance
        A PRoPHET node needs to know the identity of its neighbors and
        when new neighbors appear and old neighbors disappear.  Some
        wireless networking technologies might already contain
        mechanisms for detecting neighbors and maintaining this state.
        To avoid redundancies and inefficiencies, neighbor discovery is
        thus not included as a part of PRoPHET, but PRoPHET relies on
        such a mechanism in lower layers.  The lower layers MUST provide
        the two functions listed below.  If the underlying networking
        technology does not support such services, a simple neighbor
        discovery scheme using local broadcasts of beacon messages could
        be run in between PRoPHET and the underlying layer.  An example
        of a simple neighbor discovery mechanism that could be used is
        in Appendix B.

        New Neighbor
             Signals to the PRoPHET agent that a new node has become a
             neighbor.  A neighbor is defined here as another node that
             is currently within communication range of the wireless
             networking technology in use.  The PRoPHET agent should now
             start the Hello procedure as described in Section 5.2.

        Neighbor Gone
             Signals to the PRoPHET agent that one of its neighbors has
             left.

   Local Address
        An address used by the underlying communication layer (e.g., an
        IP or Media Access Control (MAC) address) that identifies the
        sender address of the current message.  This address must be
        unique among the nodes that can currently communicate and is
        only used in conjunction with an Instance Number to identify a
        communicating pair of nodes as described in Section 4.1.  This
        address and its format is dependent on the communication layer
        that is being used by the PRoPHET layer.









Lindgren, et al.              Experimental                     [Page 21]

RFC 6693                         PRoPHET                     August 2012


3.  Protocol Overview



   The PRoPHET protocol involves two principal phases:

   o  becoming aware of new neighbors that implement the protocol and
      establishing a point-to-point connection between each pair of
      encountering nodes, and

   o  using the connection for information exchange needed to establish
      PRoPHET routing and to exchange bundles.

3.1.  Neighbor Awareness



   Since the operation of the protocol is dependent on the encounters of
   nodes running PRoPHET, the nodes must be able to detect when a new
   neighbor is present.  The protocol may be run on several different
   networking technologies, and as some of them might already have
   methods available for detecting neighbors, PRoPHET does not include a
   mechanism for neighbor discovery.  Instead, it requires the
   underlying layer to provide a mechanism to notify the protocol of
   when neighbors appear and disappear as described in Section 2.4.

   When a new neighbor has been detected, the protocol starts to set up
   a link with that node through the Hello message exchange as described
   in Section 5.2.  The Hello message exchange allows for negotiation of
   capabilities between neighbors.  At present, the only capability is a
   request that the offering node should or should not include bundle
   payload lengths with all offered bundles rather than just for
   fragments.  Once the link has been set up, the protocol may continue
   to the Information Exchange Phase (see Section 3.2).  Once this has
   been completed, the nodes will normally recalculate the delivery
   predictabilities using the equations and mechanisms described in
   Sections 2.1.2 and 2.1.3.

   As described in Section 2.1.2, there are some circumstances in which
   a single logical encounter may result in several actual communication
   opportunities.  To avoid the delivery predictability of the
   encountered node being increased excessively under these
   circumstances, the value of P_encounter is made dependent on the
   interval time between delivery predictability updates when the
   interval is less than the typical interval between encounters, but it
   is a constant for longer intervals.

   In order to make use of this time dependence, PRoPHET maintains a
   list of recently encountered nodes identified by the Endpoint
   Identifier (EID) that the node uses to identify the communication
   session and containing the start time of the last communication
   session with that node.  The size of this list is controlled because



Lindgren, et al.              Experimental                     [Page 22]

RFC 6693                         PRoPHET                     August 2012


   nodes that are not in contact and that started their last connection
   more than a time I_typ before the present can be dropped from the
   list.  It also maintains a record of the time at which the decay
   function (Equation 2) was last applied to the delivery
   predictabilities in the node.

3.2.  Information Exchange Phase



   The Information Exchange Phase involves two parts:

   o  establishing the Router Information Base (RIB Exchange Sub-Phase),
      and

   o  exchanging bundles using this information (Bundle Passing Sub-
      Phase).

   Four types of information are exchanged during this process:

   o  Routing Information Base Dictionary (RIB Dictionary or RIBD),

   o  Routing Information Base (RIB),

   o  Bundle Offers, and

   o  Bundle Responses.

   During a communication opportunity, several sets of each type of
   information may be transferred in each direction as explained in the
   rest of this section.  Each set can be transferred in one or more
   messages.  When (and only when) using a connection-oriented reliable
   transport protocol such as TCP as envisaged in this document, a set
   can be partitioned across messages by the software layer above the
   PRoPHET protocol engine.

   In this case, the last message in a set is flagged in the protocol.
   This allows the higher-level software to minimize the buffer memory
   requirements by avoiding the need to build very large messages in one
   go and allows the message size to be controlled outside of PRoPHET.
   However, this scheme is only usable if the transport protocol
   provides reliable, in-order delivery of messages, as the messages are
   not explicitly sequence numbered and the overall size of the set is
   not passed explicitly.

   The specification of PRoPHET also provides a submessage mechanism and
   retransmission that allows large messages specified by the higher
   level to be transmitted in smaller chunks.  This mechanism was
   originally provided to allow PRoPHET to operate over unreliable
   transport protocols such as UDP, but can also be used with reliable



Lindgren, et al.              Experimental                     [Page 23]

RFC 6693                         PRoPHET                     August 2012


   transports if the higher-level software does not want to handle
   message fragmentation.  However, the sequencing and length adds
   overhead that is redundant if the transport protocol already provides
   reliable, in-order delivery.

   The first step in the Information Exchange Phase is for the protocol
   to send one or more messages containing a RIB Dictionary TLV (Type-
   Length-Value message component) to the node with which it is peering.
   This set of messages contain a dictionary of the Endpoint Identifiers
   (EIDs) of the nodes that will be listed in the Routing Information
   Base (RIB); see Section 3.2.1 for more information about this
   dictionary.  After this, one or more messages containing a Routing
   Information Base TLV are sent.  This TLV contains a list of the EIDs
   that the node has knowledge of, and the corresponding delivery
   predictabilities for those nodes, together with flags describing the
   capabilities of the sending node.  Upon reception of a complete set
   of these messages, the peer node updates its delivery predictability
   table according to the equations in Section 2.1.2.  The peer node
   then applies its forwarding strategy (see Section 2.1.4) to determine
   which of its stored bundles it wishes to offer the node that sent the
   RIB; that node will then be the receiver for any bundles to be
   transferred.

   After making this decision, one or more Bundle Offer TLVs are
   prepared, listing the bundle identifiers and their destinations for
   all bundles the peer node wishes to offer to the receiver node that
   sent the RIB.  As described in [RFC5050], a bundle identifier
   consists of up to five component parts.  For a complete bundle, the
   identifier consists of

   o  source EID,

   o  creation timestamp - time of creation, and

   o  creation timestamp - sequence number.

   Additionally, for a bundle fragment, the identifier also contains

   o  offset within the payload at which the fragment payload data
      starts, and

   o  length of the fragment payload data.

   If any of the Bundle Offer TLVs lists a bundle for which the source
   or destination EID was not included in the previous set of RIBD
   information sent, one or more new RIBD TLVs are sent next with an
   incremental update of the dictionary.  When the receiver node has a
   dictionary with all necessary EIDs, the Bundle Offer TLVs are sent to



Lindgren, et al.              Experimental                     [Page 24]

RFC 6693                         PRoPHET                     August 2012


   it.  The Bundle Offer TLVs also contain a list of PRoPHET ACKs (see
   Section 3.5).  If requested by the receiver node during the Hello
   phase, the Bundle Offer TLV will also specify the payload length for
   all bundles rather than for just fragments.  This information can be
   used by the receiving node to assist with the selection of bundles to
   be accepted from the offered list, especially if the available bundle
   storage capacity is limited.

   The receiving node then examines the list of offered bundles and
   selects bundles that it will accept according to its own policies,
   considering the bundles already present in the node and the current
   availability of resources in the node.  The list is sorted according
   to the priority that the policies apply to the selected bundles, with
   the highest priority bundle first in the list.  The offering node
   will forward the selected bundles in this order.  The prioritized
   list is sent to the offering node in one or more Bundle Response TLVs
   using the same EID dictionary as was used for the Bundle Offer TLV.

   When a new bundle arrives at a node, the node MAY inspect its list of
   available neighbors, and if one of them is a candidate to forward the
   bundle, a new Bundle Offer TLV MAY be sent to that node.  If two
   nodes remain connected over a longer period of time, the Information
   Exchange Phase will be periodically re-initiated to allow new
   delivery predictability information to be spread through the network
   and new bundle exchanges to take place.

   The Information Exchange Phase of the protocol is described in more
   detail in Section 5.3.

3.2.1.  Routing Information Base Dictionary



   To reduce the overhead of the protocol, the Routing Information Base
   and Bundle Offer/Response TLVs utilize an EID dictionary.  This
   dictionary maps variable-length EIDs (as defined in [RFC4838]), which
   may potentially be quite long, to shorter numerical identifiers,
   coded as Self-Delimiting Numeric Values (SDNVs -- see Section 4.1. of
   RFC 5050 [RFC5050]), which are used in place of the EIDs in
   subsequent TLVs.

   This dictionary is a shared resource between the two peering nodes.
   Each can add to the dictionary by sending a RIB Dictionary TLV to its
   peer.  To allow either node to add to the dictionary at any time, the
   identifiers used by each node are taken from disjoint sets:
   identifiers originated by the node that started the Hello procedure
   have the least significant bit set to 0 (i.e., are even numbers)
   whereas those originated by the other peer have the least significant
   bit set to 1 (i.e., are odd numbers).  This means that the dictionary




Lindgren, et al.              Experimental                     [Page 25]

RFC 6693                         PRoPHET                     August 2012


   can be expanded by either node at any point in the Information
   Exchange Phase and the new identifiers can then be used in subsequent
   TLVs until the dictionary is re-initialized.

   The dictionary that is established only persists through a single
   encounter with a node (i.e., while the same link set up by the Hello
   procedure, with the same instance numbers, remains open).

   Having more then one identifier for the same EID does not cause any
   problems.  This means that it is possible for the peers to create
   their dictionary entries independently if required by an
   implementation, but this may be inefficient as a dictionary entry for
   an EID might be sent in both directions between the peers.
   Implementers can choose to inspect entries sent by the node that
   started the Hello procedure and thereby eliminate any duplicates
   before sending the dictionary entries from the other peer.  Whether
   postponing sending the other peer's entries is more efficient depends
   on the nature of the physical link technology and the transport
   protocol used.  With a genuinely full-duplex link, it may be faster
   to accept possible duplication and send dictionary entries
   concurrently in both directions.  If the link is effectively half-
   duplex (e.g., Wi-Fi), then it will generally be more efficient to
   wait and eliminate duplicates.

   If a node receives a RIB Dictionary TLV containing an identifier that
   is already in use, the node MUST confirm that the EID referred to is
   identical to the EID in the existing entry.  Otherwise, the node must
   send an error response to the message with the TLV containing the
   error and ignore the TLV containing the error.  If a node receives a
   RIB, Bundle Offer, or Bundle Response TLV that uses an identifier
   that is not in its dictionary, the node MUST send an error response
   and ignore the TLV containing the error.

3.2.2.  Handling Multiple Simultaneous Contacts



   From time to time, a mobile node may, for example, be in wireless
   range of more than one other mobile node.  The PRoPHET neighbor
   awareness protocol will establish multiple simultaneous contacts with
   these nodes and commence information exchanges with each of them.

   When updating the delivery predictabilities as described in
   Section 2.1.2 using the values passed from each of the contacts in
   turn, some special considerations apply when multiple contacts are in
   progress:







Lindgren, et al.              Experimental                     [Page 26]

RFC 6693                         PRoPHET                     August 2012


   SC1  When aging the delivery predictabilities according to
        Equation 2, the value of K to be used in each set of
        calculations is always the amount of time since the last aging
        was done.  For example, if node Z makes contact with node A and
        then with node B, the value of K used when the delivery
        predictabilities are aged in node Z for the contact with node B
        will be the time since the delivery predictabilities were aged
        for the contact with node A.

   SC2  When a new contact starts, the value of P_encounter used when
        applying Equation 1 for the newly contacted node is always
        selected according to the time since the last encounter with
        that node.  Thus, the application of Equation 1 to update
        P_(Z,A) when the contact of nodes Z and A starts (in the aging
        example just given) and the updating of P_(Z,B) when the contact
        of nodes Z and B starts will use the appropriate value of
        P_encounter according to how long it is since node Z previously
        encountered node A and node B, respectively.

   SC3  If, as with the contact between nodes Z and B, there is another
        active contact in progress, such as with node A when the contact
        with node B starts, Equation 1 should *also* be applied to
        P_(z,x) for all the nodes "x" that have ongoing contacts with
        node Z (i.e., node A in the example given).  However, the value
        of P_encounter used will be selected according to the time since
        the previous update of the delivery predictabilities as a result
        of information received from any other node.  In the example
        given here, P_(Z,A) would also have Equation 1 applied when the
        delivery predictabilities are received from node B, but the
        value of P_encounter used would be selected according to the
        time since the updates done when the encounter between nodes Z
        and A started rather than the time since the previous encounter
        between nodes A and Z.

   If these simultaneous contacts persist for some time, then, as
   described in Section 3.2, the Information Exchange Phase will be
   periodically rerun for each contact according to the configured timer
   interval.  When the delivery predictability values are recalculated
   during each rerun, Equation 1 will be applied as in special
   consideration SC3 above, but it will be applied to the delivery
   predictability for each active contact using the P_encounter value
   selected according to the time since the last set of updates were
   performed on the delivery predictabilities, irrespective of which
   nodes triggered either the previous or current updates.  This means
   that, in the example discussed here, P_(Z,A) and P_(Z,B) will be
   updated using the same value of P_encounter whether node A or node B
   initiated the update while the three nodes remain connected.




Lindgren, et al.              Experimental                     [Page 27]

RFC 6693                         PRoPHET                     August 2012


   The interval between reruns of the information exchange will
   generally be set to a small fraction of the expected time between
   independent encounters of pairs of nodes.  This ensures that, for
   example, the delivery predictability information obtained by node Z
   from node A will be passed on to node B whether or not nodes A and B
   can communicate directly during this encounter.  This avoids problems
   that may arise from peculiarities of radio propagation during this
   sort of encounter, but the scaling of the P_encounter factor
   according to the time between updates of the delivery
   predictabilities means that the predictabilities for the nodes that
   are in contact are not increased excessively as would be the case if
   each information exchange were treated as a separate encounter with
   the value of P_encounter_max used each time.  When several nodes are
   in mutual contact, the delivery predictabilities in each node
   stabilize after a few exchanges due to the scaling of P_encounter as
   well as the form of Equation 3 where a "max" function is used.  This
   has been demonstrated by simulation.

   The effect of the updates of the delivery predictabilities when there
   are multiple simultaneous contacts is that the information about good
   routes on which to forward bundles is correctly passed between sets
   of nodes that are simultaneously in contact through the transitive
   update of Equation 3 during each information exchange, but the
   delivery predictabilities for the direct contacts are not
   exaggerated.

3.3.  Routing Algorithm



   The basic routing algorithm of the protocol is described in
   Section 2.1.  The algorithm uses some parameter values in the
   calculation of the delivery predictability metric.  These parameters
   are configurable depending on the usage scenario, but Figure 3
   provides some recommended default values.  A brief explanation of the
   parameters and some advice on setting appropriate values is given
   below.

   I_typ
        I_typ provides a fundamental timescale for the mobility pattern
        in the PRoPHET scenario where the protocol is being applied.  It
        represents the typical or mean time interval between encounters
        between a given pair of nodes in the normal course of mobility.
        The interval should reflect the "logical" time between
        encounters and should not give significant weight to multiple
        connection events as explained in Section 2.1.2.  This time
        interval informs the settings of many of the other parameters
        but is not necessarily directly used as a parameter.
        Consideration needs to be given to the higher statistical
        moments (e.g., standard deviation) as well as the mean (first



Lindgren, et al.              Experimental                     [Page 28]

RFC 6693                         PRoPHET                     August 2012


        moment) of the distribution of intervals between encounters and
        the nature of that distribution (e.g., how close to a normal
        distribution it is).  There is further discussion of this point
        later in this section and in Appendix C.

   P_encounter_max
        P_encounter_max is used as the upper limit of a scaling factor
        that increases the delivery predictability for a destination
        when the destination node is encountered.  A larger value of
        P_encounter_max will increase the delivery predictability
        faster, and fewer encounters will be required for the delivery
        predictability to reach a certain level.  Given that relative
        rather than absolute delivery predictability values are what is
        interesting for the forwarding mechanisms defined, the protocol
        is very robust to different values of P_encounter as long as the
        same value is chosen for all nodes.  The value should be chosen
        so that the increase in the delivery predictability resulting
        from using P_encounter_max in Equation 1 more than compensates
        for the decay of the delivery predictability resulting from
        Equation 3 with a time interval of I_typ.

   P_encounter(intvl)
        As explained in Section 2.1.2, the parameter P_encounter used in
        Equation 1 is a function of the time interval "intvl".  The
        function should be an approximation to

             P_encounter(intvl) =
             P_encounter_max * (intvl / I_typ) for 0<= intvl <= I_typ
             P_encounter_max for intvl > I_typ

        The function can be quantized and adapted to suit the mobility
        pattern and to make implementation easier.  The overall effect
        should be that be that if Equation 1 is applied a number of
        times during a long-lived communication opportunity lasting
        I_typ, the overall increase in the delivery predictability
        should be approximately the same as if there had been two
        distinct encounters spaced I_typ apart.  This second case would
        result in one application of Equation 1 using P_encounter_max.

   P_first_threshold
        As described in Section 2.1.2, the delivery predictability for a
        destination is gradually reduced over time unless increased as a
        result of direct encounters or through the transitive property.
        If the delivery predictability falls below the value
        P_first_threshold, then the node MAY discard the delivery
        predictability information for the destination and treat
        subsequent encounters as if they had never encountered the node
        previously.  This allows the node to reduce the storage needed



Lindgren, et al.              Experimental                     [Page 29]

RFC 6693                         PRoPHET                     August 2012


        for delivery predictabilities and decreases the amount of
        information that has to be exchanged between nodes; otherwise,
        the reduction algorithm would result in very small but non-zero
        predictabilities being maintained for nodes that were last
        encountered a long time ago.

   P_encounter_first
        As described in Section 2.1.2, PRoPHET does not, by default,
        make any assumptions about the likelihood that an encountered
        node will be encountered repeatedly in the future or,
        alternatively, that this is a one-off chance encounter that is
        unlikely to be repeated.  During an encounter where the
        encountering node has no delivery predictability information for
        the encountered destination node, either because this is really
        the first encounter between the nodes or because the previous
        encounter was so long ago that the predictability had fallen
        below P_first_threshold and therefore had been discarded, the
        encountering node sets the delivery predictability for the
        destination node to P_encounter_first.  The suggested value for
        P_encounter_first is 0.5: this value is RECOMMENDED as
        appropriate in the usual case where PRoPHET has no extra (e.g.,
        out-of-band) information about whether future encounters with
        this node will be regular or otherwise.

   alpha
        The alpha parameter is used in the optional smoothing of the
        delivery predictabilities described in Section 2.1.3.1.  It is
        used to determine the weight of the most current P-value in the
        calculation of an EWMA.

   beta
        The beta parameter adjusts the weight of the transitive property
        of PRoPHET, that is, how much consideration should be given to
        information about destinations that is received from encountered
        nodes.  If beta is set to zero, the transitive property of
        PRoPHET will not be active, and only direct encounters will be
        used in the calculation of the delivery predictability.  The
        higher the value of beta, the more rapidly encounters will
        increase predictabilities through the transitive rule.

   gamma
        The gamma parameter determines how quickly delivery
        predictabilities age.  A lower value of gamma will cause the
        delivery predictability to age faster.  The value of gamma
        should be chosen according to the scenario and environment in
        which the protocol will be used.  If encounters are expected to
        be very frequent, a lower value should be chosen for gamma than
        if encounters are expected to be rare.



Lindgren, et al.              Experimental                     [Page 30]

RFC 6693                         PRoPHET                     August 2012


   delta
        The delta parameter sets the maximum value of the delivery
        predictability for a destination other than for the node itself
        (i.e., P_(A,B) for all cases except P_(A,A)) as (1 - delta).
        Delta should be set to a small value to allow the maximum
        possible range for predictabilities but can be configured to
        make the calculation efficient if needed.

   To set an appropriate gamma value, one should consider the "average
   expected delivery" time I_aed in the PRoPHET zone where the protocol
   is to be used, and the time unit used (the resolution with which the
   delivery predictability is being updated).  The I_aed time interval
   can be estimated according to the average number of hops that bundles
   have to pass and the average interval between encounters I_typ.
   Clearly, if bundles have a Time To Live (TTL), i.e., the time left
   until the expiry time stored in the bundle occurs, that is less than
   I_aed, they are unlikely to survive in the network to be delivered to
   a node in this PRoPHET zone.  However, the TTL for bundles created in
   nodes in this zone should not be chosen solely on this basis because
   they may pass through other networks.

   After estimating I_aed and selecting how much we want the delivery
   predictability to age in one I_aed time period (call this A), we can
   calculate K, the number of time units in one I_aed, using
   K = (I_aed / time unit).  This can then be used to calculate gamma as
   gamma = K'th-root( A ).

   I_typ, I_aed, K, and gamma can then be used to inform the settings of
   P_encounter_first, P_encounter_max, P_first_threshold, delta, and the
   detailed form of the function P_encounter(intvl).

   First, considering the evolution of the delivery predictability
   P_(A,B) after a single encounter between nodes A and B, P_(A,B) is
   initially set to P_encounter_first and will then steadily decay until
   it reaches P_first_threshold.  The ratio between P_encounter_first
   and P_first_threshold should be set so that P_first_threshold is
   reached after a small multiple (e.g., 3 to 5) of I_aed has elapsed,
   making it likely that any subsequent encounter between the nodes
   would have occurred before P_(A,B) decays below P_first_threshold.
   If the statistics of the distribution of times between encounters is
   known, then a small multiple of the standard deviation of the
   distribution would be a possible period instead of using a multiple
   of I_aed.

   Second, if a second encounter between A and B occurs, the setting of
   P_encounter_max should be sufficiently high to reverse the decay that
   would have occurred during I_typ and to increase P_(A,B) above the
   value of P_encounter_first.  After several further encounters,



Lindgren, et al.              Experimental                     [Page 31]

RFC 6693                         PRoPHET                     August 2012


   P_(A,B) will reach (1 - delta), its upper limit.  As with setting up
   P_first_threshold, P_encounter_max should be set so that the upper
   limit is reached after a small number of encounters spaced apart by
   I_typ have occurred, but this should generally be more than 2 or 3.

   Finally, beta can be chosen to give some smoothing of the influence
   of transitivity.

   These instructions on how to set the parameters are only given as a
   possible method for selecting appropriate values, but network
   operators are free to set parameters as they choose.  Appendix C goes
   into some more detail on linking the parameters defined here and the
   more conventional ways of expressing the mobility model in terms of
   distributions of times between events of various types.

   Recommended starting parameter values when specific network
   measurements have not been done are below.  Note: There are no "one
   size fits all" default values, and the ideal values vary based on
   network characteristics.  It is not inherently necessary for the
   parameter values to be identical at all nodes, but it is recommended
   that similar values are used at all nodes within a PRoPHET zone as
   discussed in Section 2.1.2.1.

     +========================================+
     |      Parameter     | Recommended value |
     +========================================+
     |   P_encounter_max  |       0.7         |
     +----------------------------------------+
     |  P_encounter_first |       0.5         |
     +----------------------------------------+
     |  P_first_threshold |       0.1         |
     +----------------------------------------+
     |        alpha       |       0.5         |
     +----------------------------------------+
     |        beta        |       0.9         |
     +----------------------------------------+
     |        gamma       |       0.999       |
     +----------------------------------------+
     |        delta       |       0.01        |
     +========================================+

                   Figure 3: Default parameter settings

3.4.  Bundle Passing



   Upon reception of the Bundle Offer TLV, the node inspects the list of
   bundles and decides which bundles it is willing to store for future
   forwarding or that it is able to deliver to their destinations.  This



Lindgren, et al.              Experimental                     [Page 32]

RFC 6693                         PRoPHET                     August 2012


   decision has to be made using local policies and considering
   parameters such as available buffer space and, if the node requested
   bundle lengths, the lengths of the offered bundles.  For each such
   acceptable bundle, the node sends a Bundle Response TLV to its
   peering node, which responds by sending the requested bundle.  If a
   node has some bundles it would prefer to receive ahead of others
   offered (e.g., bundles that it can deliver to their final
   destination), it MAY request the bundles in that priority order.
   This is often desirable as there is no guarantee that the nodes will
   remain in contact with each other for long enough to transfer all the
   acceptable bundles.  Otherwise, the node SHOULD assume that the
   bundles are listed in a priority order determined by the peering
   node's forwarding strategy and request bundles in that order.

3.4.1.  Custody



   To free up local resources, a node may give custody of a bundle to
   another node that offers custody.  This is done to move the
   retransmission requirement further toward the destination.  The
   concept of custody transfer, and more details on the motivation for
   its use can be found in [RFC4838].  PRoPHET takes no responsibilities
   for making custody decisions.  Such decisions should be made by a
   higher layer.

3.5.  When a Bundle Reaches Its Destination



   A PRoPHET ACK is only a confirmation that a bundle has been delivered
   to its destination in the PRoPHET zone (within the part of the
   network where PRoPHET is used for routing, bundles might traverse
   several different types of networks using different routing
   protocols; thus, this might not be the final destination of the
   bundle).  When nodes exchange Bundle Offer TLVs, bundles that have
   been ACKed are also listed, having the "PRoPHET ACK" flag set.  The
   node that receives this list updates its own list of ACKed bundles to
   be the union of its previous list and the received list.  To prevent
   the list of ACKed bundles growing indefinitely, each PRoPHET ACK
   should have a timeout that MUST NOT be longer than the timeout of the
   bundle to which the ACK corresponds.

   When a node receives a PRoPHET ACK for a bundle it is carrying, it
   MAY delete that bundle from its storage, unless the node holds
   custody of that bundle.  The PRoPHET ACK only indicates that a bundle
   has been delivered to its destination within the PRoPHET zone, so the
   reception of a PRoPHET ACK is not a guarantee that the bundle has
   been delivered to its final destination.






Lindgren, et al.              Experimental                     [Page 33]

RFC 6693                         PRoPHET                     August 2012


   Nodes MAY track to which nodes they have sent PRoPHET ACKs for
   certain bundles, and MAY in that case refrain from sending multiple
   PRoPHET ACKs for the same bundle to the same node.

   If necessary in order to preserve system resources, nodes MAY drop
   PRoPHET ACKs prematurely but SHOULD refrain from doing so if
   possible.

   It is important to keep in mind that PRoPHET ACKs and bundle ACKs
   [RFC5050] are different things.  PRoPHET ACKs are only valid within
   the PRoPHET part of the network, while bundle ACKs are end-to-end
   acknowledgments that may go outside of the PRoPHET zone.

3.6.  Forwarding Strategies



   During the Information Exchange Phase, nodes need to decide on which
   bundles they wish to exchange with the peering node.  Because of the
   large number of scenarios and environments that PRoPHET can be used
   in, and because of the wide range of devices that may be used, it is
   not certain that this decision will be based on the same strategy in
   every case.  Therefore, each node MUST operate a _forwarding
   strategy_ to make this decision.  Nodes may define their own
   strategies, but this section defines a few basic forwarding
   strategies that nodes can use.  Note: If the node being encountered
   is the destination of any of the bundles being carried, those bundles
   SHOULD be offered to the destination, even if that would violate the
   forwarding strategy.  Some of the forwarding strategies listed here
   have been evaluated (together with a number of queueing policies)
   through simulations, and more information about that and
   recommendations on which strategies to use in different situations
   can be found in [lindgren_06].  If not chosen differently due to the
   characteristics of the deployment scenario, nodes SHOULD choose GRTR
   as the default forwarding strategy.

   The short names applied to the forwarding strategies should be read
   as mnemonic handles rather than as specific acronyms for any set of
   words in the specification.

   We use the following notation in our descriptions below.  A and B are
   the nodes that encounter each other, and the strategies are described
   as they would be applied by node A.  The destination node is D.
   P_(X,Y) denotes the delivery predictability stored at node X for
   destination Y, and NF is the number of times node A has given the
   bundle to some other node.







Lindgren, et al.              Experimental                     [Page 34]

RFC 6693                         PRoPHET                     August 2012


   GRTR
        Forward the bundle only if P_(B,D) > P_(A,D).

        When two nodes meet, a bundle is sent to the other node if the
        delivery predictability of the destination of the bundle is
        higher at the other node.  The first node does not delete the
        bundle after sending it as long as there is sufficient buffer
        space available (since it might encounter a better node, or even
        the final destination of the bundle in the future).

   GTMX
        Forward the bundle only if P_(B,D) > P_(A,D) && NF < NF_max.

        This strategy is like the previous one, but each bundle is given
        to at most NF_max other nodes in addition to the destination.

   GTHR
        Forward the bundle only if
        P_(B,D) > P_(A,D) OR P_(B,D) > FORW_thres,
        where FORW_thres is a threshold value above which a bundle
        should always be given to the node unless it is already present
        at the other node.

        This strategy is similar to GRTR, but among nodes with very high
        delivery predictability, bundles for that particular destination
        are spread epidemically.

   GRTR+
        Forward the bundle only if Equation 5 holds, where P_max is the
        largest delivery predictability reported by a node to which the
        bundle has been sent so far.

             P_(B,D) > P_(A,D) && P_(B,D) > P_max  (Eq. 5)

        This strategy is like GRTR, but each node forwarding a bundle
        keeps track of the largest delivery predictability of any node
        it has forwarded this bundle to, and only forwards the bundle
        again if the currently encountered node has a greater delivery
        predictability than the maximum previously encountered.

   GTMX+
        Forward the bundle only if Equation 6 holds.

            P_(B,D) > P_(A,D) && P_(B,D) > P_max && NF < NF_max  (Eq. 6)

        This strategy is like GTMX, but nodes keep track of P_max as in
        GRTR+.




Lindgren, et al.              Experimental                     [Page 35]

RFC 6693                         PRoPHET                     August 2012


   GRTRSort
        Select bundles in descending order of the value of
        P_(B,D) - P_(A,D).
        Forward the bundle only if P_(B,D) > P_(A,D).

        This strategy is like GRTR, but instead of just going through
        the bundle queue linearly, this strategy looks at the difference
        in delivery predictabilities for each bundle between the two
        nodes and forwards the bundles with the largest difference
        first.  As bandwidth limitations or disrupted connections may
        result in not all bundles that would be desirable being
        exchanged, it could be desirable to first send bundles that get
        a large improvement in delivery predictability.

   GRTRMax
        Select bundles in descending order of P_(B,D).
        Forward the bundle only if P_(B,D) > P_(A,D).

        This strategy begins by considering the bundles for which the
        encountered node has the highest delivery predictability.  The
        motivation for doing this is the same as in GRTRSort, but based
        on the idea that it is better to give bundles to nodes with high
        absolute delivery predictabilities, instead of trying to
        maximize the improvement.

3.7.  Queueing Policies



   Because of limited buffer resources, nodes may need to drop some
   bundles.  As is the case with the forwarding strategies, which bundle
   to drop is also dependent on the scenario.  Therefore, each node MUST
   also operate a queueing policy that determines how its bundle queue
   is handled.  This section defines a few basic queueing policies, but
   nodes MAY use other policies if desired.  Some of the queueing
   policies listed here have been evaluated (together with a number of
   forwarding strategies) through simulations.  More information about
   that and recommendations on which policies to use in different
   situations can be found in [lindgren_06].  If not chosen differently
   due to the characteristics of the deployment scenario, nodes SHOULD
   choose FIFO as the default queueing policy.

   The short names applied to the queueing policies should be read as
   mnemonic handles rather than as specific acronyms for any set of
   words in the specification.

   FIFO - First In First Out.
        The bundle that was first entered into the queue is the first
        bundle to be dropped.




Lindgren, et al.              Experimental                     [Page 36]

RFC 6693                         PRoPHET                     August 2012


   MOFO - Evict most forwarded first.
        In an attempt to maximize the delivery rate of bundles, this
        policy requires that the routing agent keep track of the number
        of times each bundle has been forwarded to some other node.  The
        bundle that has been forwarded the largest number of times is
        the first to be dropped.

   MOPR - Evict most favorably forwarded first.
        Keep a variable FAV for each bundle in the queue, initialized to
        zero.  Each time the bundle is forwarded, update FAV according
        to Equation 7, where P is the predictability metric that the
        node the bundle is forwarded to has for its destination.

             FAV_new = FAV_old + ( 1 - FAV_old ) * P  (Eq. 7)

        The bundle with the highest FAV value is the first to be
        dropped.

   Linear MOPR - Evict most favorably forwarded first; linear increase.
        Keep a variable FAV for each bundle in the queue, initialized to
        zero.  Each time the bundle is forwarded, update FAV according
        to Equation 8, where P is the predictability metric that the
        node the bundle is forwarded to has for its destination.

             FAV_new = FAV_old + P  (Eq. 8)

        The bundle with the highest FAV value is the first to be
        dropped.

   SHLI - Evict shortest life time first.
        As described in [RFC5050], each bundle has a timeout value
        specifying when it no longer is meaningful to its application
        and should be deleted.  Since bundles with short remaining Time
        To Live will soon be dropped anyway, this policy decides to drop
        the bundle with the shortest remaining lifetime first.  To
        successfully use a policy like this, there needs to be some form
        of time synchronization between nodes so that it is possible to
        know the exact lifetimes of bundles.  However, this is not
        specific to this routing protocol, but a more general DTN
        problem.

   LEPR - Evict least probable first.
        Since the node is least likely to deliver a bundle for which it
        has a low delivery predictability, drop the bundle for which the
        node has the lowest delivery predictability, and that has been
        forwarded at least MF times, where MF is a minimum number of
        forwards that a bundle must have been forwarded before being
        dropped (if such a bundle exists).



Lindgren, et al.              Experimental                     [Page 37]

RFC 6693                         PRoPHET                     August 2012


   More than one queueing policy MAY be combined in an ordered set,
   where the first policy is used primarily, the second only being used
   if there is a need to tie-break between bundles given the same
   eviction priority by the primary policy, and so on.  As an example,
   one could select the queueing policy to be {MOFO; SHLI; FIFO}, which
   would start by dropping the bundle that has been forwarded the
   largest number of times.  If more than one bundle has been forwarded
   the same number of times, the one with the shortest remaining
   lifetime will be dropped, and if that also is the same, the FIFO
   policy will be used to drop the bundle first received.

   It is worth noting that a node MUST NOT drop bundles for which it has
   custody unless the bundle's lifetime expires.

4.  Message Formats



   This section defines the message formats of the PRoPHET routing
   protocol.  In order to allow for variable-length fields, many numeric
   fields are encoded as Self-Delimiting Numeric Values (SDNVs).  The
   format of SDNVs is defined in [RFC5050].  Since many of the fields
   are coded as SDNVs, the size and alignment of fields indicated in
   many of the specification diagrams below are indicative rather than
   prescriptive.  Where SDNVs and/or text strings are used, the octets
   of the fields will be packed as closely as possible with no
   intervening padding between fields.

   Explicit-length fields are specified for all variable-length string
   fields.  Accordingly, strings are not null terminated and just
   contain the exact set of octets in the string.

   The basic message format shown in Figure 4 consists of a header (see
   Section 4.1) followed by a sequence of one or more Type-Length-Value
   components (TLVs) taken from the specifications in Section 4.2.


















Lindgren, et al.              Experimental                     [Page 38]

RFC 6693                         PRoPHET                     August 2012


       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~                            Header                             ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~                             TLV 1                             ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                .                              |
      ~                                .                              ~
      |                                .                              |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~                             TLV n                             ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                  Figure 4: Basic PRoPHET Message Format

4.1.  Header



       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |Protocol Number|Version| Flags |     Result    |     Code      |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |      Receiver Instance        |      Sender Instance          |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                    Transaction Identifier                     |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |S|      SubMessage Number      |         Length (SDNV)         |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~                          Message Body                         ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                     Figure 5: PRoPHET Message Header

   Protocol Number
        The DTN Routing Protocol Number encoded as 8-bit unsigned
        integer in network bit order.  The value of this field is 0.
        The PRoPHET header is organized in this way so that in principle
        PRoPHET messages could be sent as the Protocol Data Unit of an
        IP packet if an IP protocol number was allocated for PRoPHET.



Lindgren, et al.              Experimental                     [Page 39]

RFC 6693                         PRoPHET                     August 2012


        At present, PRoPHET is only specified to use a TCP transport for
        carriage of PRoPHET packets, so that the protocol number serves
        only to identify the PRoPHET protocol within DTN.  Transmitting
        PRoPHET packets directly as an IP protocol on a public IP
        network such as the Internet would generally not work well
        because middleboxes (such as firewalls and NAT boxes) would be
        unlikely to allow the protocol to pass through, and the protocol
        does not provide any congestion control.  However, it could be
        so used on private networks for experimentation or in situations
        where all communications are between isolated pairs of nodes.
        Also, in the future, other protocols that require transmission
        of metadata between DTN nodes could potentially use the same
        format and protocol state machinery but with a different
        Protocol Number.

   Version
        The version of the PRoPHET Protocol.  Encoded as a 4-bit
        unsigned integer in network bit order.  This document defines
        version 2.

   Flags
        Reserved field of 4 bits.

   Result
        Field that is used to indicate whether a response is required to
        the request message if the outcome is successful.  A value of
        "NoSuccessAck" indicates that the request message does not
        expect a response if the outcome is successful, and a value of
        "AckAll" indicates that a response is expected if the outcome is
        successful.  In both cases, a failure response MUST be generated
        if the request fails.  If running over a TCP transport or
        similar protocol that offers reliable in order delivery,
        deployments MAY choose not to send "Success" responses when an
        outcome is successful.  To achieve this, the Result field is set
        to the "NoSuccessAck" value in all request messages.

        In a response message, the result field can have two values:
        "Success" and "Failure".  The "Success" result indicates a
        success response.  All messages that belong to the same success
        response will have the same Transaction Identifier.  The
        "Success" result indicates a success response that may be
        contained in a single message or the final message of a success
        response spanning multiple messages.








Lindgren, et al.              Experimental                     [Page 40]

RFC 6693                         PRoPHET                     August 2012


        ReturnReceipt is a value of the result field used to indicate
        that an acknowledgement is required for the message.  The
        default for messages is that the controller will not acknowledge
        responses.  In the case where an acknowledgement is required, it
        will set the Result Field to ReturnReceipt in the header of the
        Message.

        The result field is encoded as an 8-bit unsigned integer in
        network bit order.  The following values are currently defined:

           NoSuccessAck:       Result = 1
           AckAll:             Result = 2
           Success:            Result = 3
           Failure:            Result = 4
           ReturnReceipt       Result = 5

   Code
        This field gives further information concerning the result in a
        response message.  It is mostly used to pass an error code in a
        failure response but can also be used to give further
        information in a success response message or an event message.
        In a request message, the code field is not used and is set to
        zero.

        If the Code field indicates that the Error TLV is included in
        the message, further information on the error will be found in
        the Error TLV, which MUST be the first TLV after the header.

        The Code field is encoded as an 8-bit unsigned integer in
        network bit order.  Separate number code spaces are used for
        success and failure response messages.  In each case, a range of
        values is reserved for use in specifications and another range
        for private and experimental use.  For success messages, the
        following values are defined:

                  Generic Success                  0x00
                  Submessage Received              0x01
                  Unassigned                   0x02 - 0x7F
                  Private/Experimental Use     0x80 - 0xFF

        The Submessage Received code is used to acknowledge reception of
        a message segment.  The Generic Success code is used to
        acknowledge receipt of a complete message and successful
        processing of the contents.







Lindgren, et al.              Experimental                     [Page 41]

RFC 6693                         PRoPHET                     August 2012


        For failure messages, the following values are defined:

                  Reserved                     0x00 - 0x01
                  Unspecified Failure              0x02
                  Unassigned                   0x03 - 0x7F
                  Private/Experimental Use     0x80 - 0xFE
                  Error TLV in message             0xFF

        The Unspecified Failure code can be used to report a failure for
        which there is no more specific code or Error TLV value defined.

   Sender Instance
        For messages during the Hello phase with the Hello SYN, Hello
        SYNACK, and Hello ACK functions (which are explained in
        Section 5.2), it is the sender's instance number for the link.
        It is used to detect when the link comes back up after going
        down or when the identity of the entity at the other end of the
        link changes.  The instance number is a 16-bit number that is
        guaranteed to be unique within the recent past and to change
        when the link or node comes back up after going down.  Zero is
        not a valid instance number.  For the RSTACK function (also
        explained in detail in Section 5.2), the Sender Instance field
        is set to the value of the Receiver Instance field from the
        incoming message that caused the RSTACK function to be
        generated.  Messages sent after the Hello phase is completed
        should use the sender's instance number for the link.  The
        Sender Instance is encoded as a 16-bit unsigned integer in
        network bit order.

   Receiver Instance
        For messages during the Hello phase with the Hello SYN, Hello
        SYNACK, and Hello ACK functions, it is what the sender believes
        is the current instance number for the link, allocated by the
        entity at the far end of the link.  If the sender of the message
        does not know the current instance number at the far end of the
        link, this field MUST be set to zero.  For the RSTACK message,
        the Receiver Instance field is set to the value of the Sender
        Instance field from the incoming message that caused the RSTACK
        message to be generated.  Messages sent after the Hello phase is
        completed should use what the sender believes is the current
        instance number for the link, allocated by the entity at the far
        end of the link.  The Sender Instance is encoded as a 16-bit
        unsigned integer in network bit order.








Lindgren, et al.              Experimental                     [Page 42]

RFC 6693                         PRoPHET                     August 2012


   Transaction Identifier
        Used to associate a message with its response message.  This
        should be set in request messages to a value that is unique for
        the sending host within the recent past.  Reply messages contain
        the Transaction Identifier of the request to which they are
        responding.  The Transaction Identifier is a bit pattern of 32
        bits.

   S-flag
        If S is set (value 1), then the SubMessage Number field
        indicates the total number of SubMessage segments that compose
        the entire message.  If it is not set (value 0), then the
        SubMessage Number field indicates the sequence number of this
        SubMessage segment within the whole message.  The S field will
        only be set in the first submessage of a sequence.

   SubMessage Number
        When a message is segmented because it exceeds the MTU of the
        link layer or otherwise, each segment will include a SubMessage
        Number to indicate its position.  Alternatively, if it is the
        first submessage in a sequence of submessages, the S-flag will
        be set, and this field will contain the total count of
        SubMessage segments.  The SubMessage Number is encoded as a
        15-bit unsigned integer in network bit order.  The SubMessage
        number is zero-based, i.e., for a message divided into n
        submessages, they are numbered from 0 to (n - 1).  For a message
        that is not divided into submessages, the single message has the
        S-flag cleared (value 0), and the SubMessage Number is set to 0
        (zero).

   Length
        Length in octets of this message including headers and message
        body.  If the message is fragmented, this field contains the
        length of this SubMessage.  The Length is encoded as an SDNV.

   Message Body
        As specified in Section 4, the Message Body consists of a
        sequence of one or more of the TLVs specified in Section 4.2.

   The protocol also requires extra information about the link that the
   underlying communication layer MUST provide.  This information is
   used in the Hello procedure described in more detail in Section 5.2.
   Since this information is available from the underlying layer, there
   is no need to carry it in PRoPHET messages.  The following values are
   defined to be provided by the underlying layer:






Lindgren, et al.              Experimental                     [Page 43]

RFC 6693                         PRoPHET                     August 2012


   Sender Local Address
        An address that is used by the underlying communication layer as
        described in Section 2.4 and identifies the sender address of
        the current message.  This address must be unique among the
        nodes that can currently communicate, and it is only used in
        conjunction with the Receiver Local Address, Receiver Instance,
        and Sender Instance to identify a communicating pair of nodes.

   Receiver Local Address
        An address that is used by the underlying communication layer as
        described in Section 2.4 and identifies the receiver address of
        the current message.  This address must be unique among the
        nodes that can currently communicate, and is only used in
        conjunction with the Sender Local Address, Receiver Instance,
        and Sender Instance to identify a communicating pair of nodes.

   When PRoPHET is run over TCP, the IP addresses of the communicating
   nodes are used as Sender and Receiver Local Addresses.

4.2.  TLV Structure



   All TLVs have the following format, and can be nested.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |    TLV Type   |   TLV Flags   |       TLV Length (SDNV)       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~                           TLV Data                            ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                           Figure 6: TLV Format

   TLV Type
        Specific TLVs are defined in Section 4.3.  The TLV Type is
        encoded as an 8-bit unsigned integer in network bit order.  Each
        TLV will have fields defined that are specific to the function
        of that TLV.

   TLV Flags
        These are defined per TLV type.  Flag n corresponds to bit 15-n
        in the TLV.  Any flags that are specified as reserved in
        specific TLVs SHOULD be transmitted as 0 and ignored on receipt.






Lindgren, et al.              Experimental                     [Page 44]

RFC 6693                         PRoPHET                     August 2012


   TLV Length
        Length of the TLV in octets, including the TLV header and any
        nested TLVs.  Encoded as an SDNV.  Note that TLVs are not padded
        to any specific alignment unless explicitly required in the
        description of the TLV.  No TLVs in this document specify any
        padding.

4.3.  TLVs



   This section describes the various TLVs that can be used in PRoPHET
   messages.

4.3.1.  Hello TLV



   The Hello TLV is used to set up and maintain a link between two
   PRoPHET nodes.  Hello messages with the SYN function are transmitted
   periodically as beacons or keep-alives.  The Hello TLV is the first
   TLV exchanged between two PRoPHET nodes when they encounter each
   other.  No other TLVs can be exchanged until the first Hello sequence
   is completed.

   Once a communication link is established between two PRoPHET nodes,
   the Hello TLV will be sent once for each interval as defined in the
   interval timer.  If a node experiences the lapse of HELLO_DEAD Hello
   intervals without receiving a Hello TLV on a connection in the
   INFO_EXCH state (as defined in the state machine in Section 5.1), the
   connection SHOULD be assumed broken.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | TLV Type=0x01 |L|  Resv | HF  |       TLV Length (SDNV)       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | Timer (SDNV)  |EID Length,SDNV|  Sender EID (variable length) |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                        Figure 7: Hello TLV Format

   TLV Flags
        The TLV Flags field contains two 1-bit flags (S and L) and a
        3-bit Hello Function (HF) number that specifies one of four
        functions for the Hello TLV.  The remaining 3 bits (Resv) are
        unused and reserved:








Lindgren, et al.              Experimental                     [Page 45]

RFC 6693                         PRoPHET                     August 2012


        HF
             TLV Flags bits 0, 1, and 2 are treated as an unsigned 3-bit
             integer coded in network bit order.  The value of the
             integer specifies the Hello Function (HF) of the Hello TLV.
             Four functions are specified for the Hello TLV.

             The encoding of the Hello Function is:

                  SYN:     HF = 1
                  SYNACK:  HF = 2
                  ACK:     HF = 3
                  RSTACK:  HF = 4

   The remaining values (0, 5, 6 and 7) are unused and reserved.  If a
   Hello TLV with any of these values is received, the link should be
   reset.

        Resv
             TLV Flags bits 3, 4, 5, and 6 are reserved.  They SHOULD be
             set to 0 on transmission and ignored on reception.

        L
             The L bit flag (TLV Flags bit 7) is set (value 1) to
             request that the Bundle Offer TLV sent during the
             Information Exchange Phase contains bundle payload lengths
             for all bundles, rather than only for bundle fragments as
             when the L flag is cleared (value 0), when carried in a
             Hello TLV with Hello Function SYN or SYNACK.  The flag is
             ignored for other Hello Function values.

   TLV Data

        Timer
             The Timer field is used to inform the receiver of the timer
             value used in the Hello processing of the sender.  The
             timer specifies the nominal time between periodic Hello
             messages.  It is a constant for the duration of a session.
             The timer field is specified in units of 100 ms and is
             encoded as an SDNV.

        EID Length
             The EID Length field is used to specify the length of the
             Sender EID field in octets.  If the Endpoint Identifier
             (EID) has already been sent at least once in a message with
             the current Sender Instance, a node MAY choose to set this
             field to zero, omitting the Sender EID from the Hello TLV.
             The EID Length is encoded as an SDNV, and the field is thus
             of variable length.



Lindgren, et al.              Experimental                     [Page 46]

RFC 6693                         PRoPHET                     August 2012


        Sender EID
             The Sender EID field specifies the DTN endpoint identifier
             (EID) of the sender that is to be used in updating routing
             information and making forwarding decisions.  If a node has
             multiple EIDs, one should be chosen for PRoPHET routing.
             This field is of variable length.

4.3.2.  Error TLV



       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | TLV type=0x02 |   TLV Flags |         TLV Length (SDNV)       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~                          TLV Data                            ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                        Figure 8: Error TLV Format

   TLV Flags
        For Error TLVs, the TLV Flags field carries an identifier for
        the Error TLV type as an 8-bit unsigned integer encoded in
        network bit order.  A range of values is available for private
        and experimental use in addition to the values defined here.
        The following Error TLV types are defined:

                  Dictionary Conflict               0x00
                  Bad String ID                     0x01
                  Reserved                       0x02 - 0x7F
                  Private/Experimental Use       0x80 - 0xFF

   TLV Data
        The contents and interpretation of the TLV Data field are
        specific to the type of Error TLV.  For the Error TLVs defined
        in this document, the TLV Data is defined as follows:

        Dictionary Conflict
             The TLV Data consists of the String ID that is causing the
             conflict encoded as an SDNV followed by the EID string that
             conflicts with the previously installed value.  The
             Endpoint Identifier is NOT null terminated.  The length of
             the EID can be determined by subtracting the length of the
             TLV Header and the length of the SDNV containing the String
             ID from the TLV Length.





Lindgren, et al.              Experimental                     [Page 47]

RFC 6693                         PRoPHET                     August 2012


        Bad String ID
             The TLV Data consists of the String ID that is not found in
             the dictionary encoded as an SDNV.

4.3.3.  Routing Information Base Dictionary TLV



   The Routing Information Base Dictionary includes the list of endpoint
   identifiers used in making routing decisions.  The referents remain
   constant for the duration of a session over a link where the instance
   numbers remain the same and can be used by both the Routing
   Information Base messages and the bundle offer/response messages.
   The dictionary is a shared resource (see Section 3.2.1) built in each
   of the paired peers from the contents of one or more incoming TLVs of
   this type and from the information used to create outgoing TLVs of
   this type.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | TLV type=0xA0 |   TLV Flags   |       TLV Length (SDNV)       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                     RIBD Entry Count (SDNV)                   |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      ~                                                               ~
      ~           Variable-Length Routing Address Strings             ~
      ~                                                               ~
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

      ~ Routing Address String 1                                      ~

      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |        String ID 1 (SDNV)     |         Length (SDNV)         |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      ~            Endpoint Identifier 1 (variable length)            ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                               .                               |
      ~ Routing Address String n      .                               ~
      |                               .                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |        String ID n (SDNV)     |         Length (SDNV)         |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      ~            Endpoint Identifier n (variable length)            ~
      |                                                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

         Figure 9: Routing Information Base Dictionary TLV Format



Lindgren, et al.              Experimental                     [Page 48]

RFC 6693                         PRoPHET                     August 2012


   TLV Flags
        The encoding of the Header flag field relates to the
        capabilities of the source node sending the RIB Dictionary:

             Flag 0: Sent by Listener    0b1
             Flag 1: Reserved            0b1
             Flag 2: Reserved            0b1
             Flag 3: Unassigned          0b1
             Flag 4: Unassigned          0b1
             Flag 5: Unassigned          0b1
             Flag 6: Unassigned          0b1
             Flag 7: Unassigned          0b1

        The "Sent by Listener" flag is set to 0 if this TLV was sent by
        a node in the Initiator role and set to 1 if this TLV was sent
        by a node in the Listener role (see Section 3.2 for explanations
        of these roles).

   TLV Data

        RIBD Entry Count
             Number of entries in the database.  Encoded as SDNV.

        String ID
             SDNV identifier that is constant for the duration of a
             session.  String ID zero is predefined as the node that
             initiates the session through sending the Hello SYN
             message, and String ID one is predefined as the node that
             responds with the Hello SYNACK message.  These entries do
             not need to be sent explicitly as the EIDs are exchanged
             during the Hello procedure.

             In order to ensure that the String IDs originated by the
             two peers do not conflict, the String IDs generated in the
             node that sent the Hello SYN message MUST have their least
             significant bit set to 0 (i.e., are even numbers), and the
             String IDs generated in the node that responded with the
             Hello SYNACK message MUST have their least significant bit
             set to 1 (i.e., they are odd numbers).

        Length
             Length of Endpoint Identifier in this entry.  Encoded as
             SDNV.

        Endpoint Identifier
             Text string representing the Endpoint Identifier.  Note
             that it is NOT null terminated as the entry contains the
             length of the identifier.



Lindgren, et al.              Experimental                     [Page 49]

RFC 6693                         PRoPHET                     August 2012


4.3.4.  Routing Information Base TLV



   The Routing Information Base lists the destinations (endpoints) a
   node knows of and the delivery predictabilities it has associated
   with them.  This information is needed by the PRoPHET algorithm to
   make decisions on routing and forwarding.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | TLV Type=0xA1 |   TLV Flags   |       TLV Length (SDNV)       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                     RIB String Count (SDNV)                   |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |     RIBD String ID 1 (SDNV)   |            P-value            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |  RIB Flags 1  |               .                               ~
      +-+-+-+-+-+-+-+-+               .                               ~
      ~                               .                               ~
      ~                               .                               ~
      ~                               .                               ~
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |     RIBD String ID n (SDNV)   |            P-value            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |  RIB Flags n  |
      +-+-+-+-+-+-+-+-+

              Figure 10: Routing Information Base TLV Format

   TLV Flags
        The encoding of the Header flag field relates to the
        capabilities of the Source node sending the RIB:

             Flag 0: More RIB TLVs       0b1
             Flag 1: Reserved            0b1
             Flag 2: Reserved            0b1
             Flag 3: Unassigned          0b1
             Flag 4: Unassigned          0b1
             Flag 5: Unassigned          0b1
             Flag 6: Unassigned          0b1
             Flag 7: Unassigned          0b1

        The "More RIB TLVs" flag is set to 1 if the RIB requires more
        TLVs to be sent in order to be fully transferred.  This flag is
        set to 0 if this is the final TLV of this RIB.






Lindgren, et al.              Experimental                     [Page 50]

RFC 6693                         PRoPHET                     August 2012


   TLV Data

        RIB String Count
             Number of routing entries in the TLV.  Encoded as an SDNV.

        RIBD String ID
             String ID of the endpoint identifier of the destination for
             which this entry specifies the delivery predictability as
             predefined in a dictionary TLV.  Encoded as an SDNV.

        P-value
             Delivery predictability for the destination of this entry
             as calculated from previous encounters according to the
             equations in Section 2.1.2, encoded as a 16-bit unsigned
             integer.  The encoding of this field is a linear mapping
             from [0,1] to [0, 0xFFFF] (e.g., for a P-value of 0.75, the
             mapping would be 0.75*65535=49151=0xBFFF; thus, the P-value
             would be encoded as 0xBFFF).

        RIB Flag
             The encoding of the 8-bit RIB Flag field is:

             Flag 0: Unassigned          0b1
             Flag 1: Unassigned          0b1
             Flag 2: Unassigned          0b1
             Flag 3: Unassigned          0b1
             Flag 4: Unassigned          0b1
             Flag 5: Unassigned          0b1
             Flag 6: Unassigned          0b1
             Flag 7: Unassigned          0b1

4.3.5.  Bundle Offer and Response TLVs (Version 2)



   After the routing information has been passed, the node will ask the
   other node to review available bundles and determine which bundles it
   will accept for relay.  The source relay will determine which bundles
   to offer based on relative delivery predictabilities as explained in
   Section 3.6.

        Note: The original versions of these TLVs (TLV Types 0xA2 and
        0xA3) used in version 1 of the PRoPHET protocol have been
        deprecated, as they did not contain the complete information
        needed to uniquely identify bundles and could not handle bundle
        fragments.

   Depending on the bundles stored in the offering node, the Bundle
   Offer TLV might contain descriptions of both complete bundles and
   bundle fragments.  In order to correctly identify bundle fragments, a



Lindgren, et al.              Experimental                     [Page 51]

RFC 6693                         PRoPHET                     August 2012


   bundle fragment descriptor MUST contain the offset of the payload
   fragment in the bundle payload and the length of the payload
   fragment.  If requested by the receiving node by setting the L flag
   in the SYN or SYNACK message during the neighbor awareness phase, the
   offering node MUST include the length of the payload in the
   descriptor for complete bundles.  The appropriate flags MUST be set
   in the B_flags for the descriptor to indicate if the descriptor
   contains the payload length field (set for fragments in all cases and
   for complete bundles if the L flag was set) and if the descriptor
   contains a payload offset field (fragments only).

   The Bundle Offer TLV also lists the bundles for which a PRoPHET
   acknowledgement has been issued.  Those bundles have the PRoPHET ACK
   flag set in their entry in the list.  When a node receives a PRoPHET
   ACK for a bundle, it SHOULD, if possible, signal to the bundle
   protocol agent that this bundle is no longer required for
   transmission by PRoPHET.  Despite no longer transmitting the bundle,
   it SHOULD keep an entry for the acknowledged bundle to be able to
   further propagate the PRoPHET ACK.

   The Response TLV format is identical to the Offer TLV with the
   exception of the TLV Type field.  Bundles that are being accepted
   from the corresponding Offer are explicitly marked with a B_flag.
   Specifications for bundles that are not being accepted MAY either be
   omitted or left in but not marked as accepted.  The payload length
   field MAY be omitted for complete bundles in the Response message
   even if it was included in the Offer message.  The B_flags payload
   length flag MUST be set correctly to indicate if the length field is
   included or not.  The Response message MUST include both payload
   offset and payload length fields for bundle fragments, and the
   B_flags MUST be set to indicate that both are present.




















Lindgren, et al.              Experimental                     [Page 52]

RFC 6693                         PRoPHET                     August 2012


       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |    TLV Type   |   TLV Flags   |       TLV Length (SDNV)       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                    Bundle Offer Count (SDNV)                  |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |    B_flags    |       Bundle Source     |  Bundle Destination |
      |               |     String ID 1 (SDNV)  |  String ID 1 (SDNV) |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                 Bundle 1 Creation Timestamp Time              |
      |                             (SDNV)                            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |           Bundle 1 Creation Timestamp Sequence Number         |
      |                             (SDNV)                            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | Bundle 1 Payload Offset - only present if bundle is a fragment|
      |                             (SDNV)                            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | Bundle 1 Payload Length - only present if bundle is a fragment|
      |         or transmission of length requested (SDNV)            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      ~                               .                               ~
      ~                               .                               ~
      ~                               .                               ~
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |    B_flags    |       Bundle Source     |  Bundle Destination |
      |               |     String ID n (SDNV)  |  String ID n (SDNV) |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                 Bundle n Creation Timestamp Time              |
      |                             (SDNV)                            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |           Bundle n Creation Timestamp Sequence Number         |
      |                             (SDNV)                            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | Bundle n Payload Offset - only present if bundle is a fragment|
      |                             (SDNV)                            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | Bundle n Payload Length - only present if bundle is a fragment|
      |         or transmission of length requested (SDNV)            |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

              Figure 11: Bundle Offer and Response TLV Format








Lindgren, et al.              Experimental                     [Page 53]

RFC 6693                         PRoPHET                     August 2012


   TLV Type
        The TLV Type for a Bundle Offer is 0xA4.  The TLV Type for a
        Bundle Response is 0xA5.

   TLV Flags
        The encoding of the Header flag field relates to the
        capabilities of the source node sending the RIB:

             Flag 0: More Offer/Response
                     TLVs Following      0b1
             Flag 1: Unassigned          0b1
             Flag 2: Unassigned          0b1
             Flag 3: Unassigned          0b1
             Flag 4: Unassigned          0b1
             Flag 5: Unassigned          0b1
             Flag 6: Unassigned          0b1
             Flag 7: Unassigned          0b1

        If the Bundle Offers or Bundle Responses are divided between
        several TLVs, the "More Offer/Response TLVs Following" flag MUST
        be set to 1 in all but the last TLV in the sequence where it
        MUST be set to 0.

   TLV Data

        Bundle Offer Count
             Number of bundle offer/response entries.  Encoded as an
             SDNV.  Note that 0 is an acceptable value.  In particular,
             a Bundle Response TLV with 0 entries is used to signal that
             a cycle of information exchange and bundle passing is
             completed.

        B Flags
             The encoding of the B Flags is:

             Flag 0: Bundle Accepted       0b1
             Flag 1: Bundle is a Fragment  0b1
             Flag 2: Bundle Payload Length
                     included in TLV       0b1
             Flag 3: Unassigned            0b1
             Flag 4: Unassigned            0b1
             Flag 5: Unassigned            0b1
             Flag 6: Unassigned            0b1
             Flag 7: PRoPHET ACK           0b1

        Bundle Source String ID
             String ID of the source EID of the bundle as predefined in
             a dictionary TLV.  Encoded as an SDNV.



Lindgren, et al.              Experimental                     [Page 54]

RFC 6693                         PRoPHET                     August 2012


        Bundle Destination String ID
             String ID of the destination EID of the bundle as
             predefined in a dictionary TLV.  Encoded as an SDNV.

        Bundle Creation Timestamp Time
             Time component of the Bundle Creation Timestamp for the
             bundle.  Encoded as an SDNV.

        Bundle Creation Timestamp Sequence Number
             Sequence Number component of the Bundle Creation Timestamp
             for the bundle.  Encoded as an SDNV.

        Bundle Payload Offset
             Only included if the bundle is a fragment and the fragment
             bit is set (value 1) in the bundle B Flags.  Offset of the
             start of the fragment payload in the complete bundle
             payload.  Encoded as an SDNV.

        Bundle Payload Length
             Only included if the bundle length included bit is set
             (value 1) in the bundle B Flags.  Length of the payload in
             the bundle specified.  This is either the total payload
             length if the bundle is a complete bundle or the bundle
             fragment payload length if the bundle is a fragment.
             Encoded as an SDNV.

5.  Detailed Operation



   In this section, some more details on the operation of PRoPHET are
   given along with state tables to help in implementing the protocol.

   As explained in Section 1.2, it is RECOMMENDED that "Success"
   responses should not be requested or sent when operating over a
   reliable, in-order transport protocol such as TCP.  If in the future
   PRoPHET were operated over an unreliable transport protocol, positive
   acknowledgements would be necessary to signal successful delivery of
   (sub)messages.  In this section, the phrase "send a message" should
   be read as *successful* sending of a message, signaled by receipt of
   the appropriate "Success" response if running over an unreliable
   protocol, but guaranteed by TCP or another reliable protocol
   otherwise.  Hence, the state descriptions below do not explicitly
   mention positive acknowledgements, whether they are being sent or
   not.








Lindgren, et al.              Experimental                     [Page 55]

RFC 6693                         PRoPHET                     August 2012


5.1.  High-Level State Tables



   This section gives high-level state tables for the operation of
   PRoPHET.  The following sections will describe each part of the
   operation in more detail (including state tables for the internal
   states of those procedures).

   The following main or high-level states are used in the state tables:

   WAIT_NB  This is the state all nodes start in.  Nodes remain in this
         state until they are notified that a new neighbor is available.
         At that point, the Hello procedure should be started with the
         new neighbor, and the node transitions into the HELLO state.
         Nodes SHOULD be able to handle multiple neighbors in parallel,
         maintaining separate state machines for each neighbor.  This
         could be handled by creating a new thread or process during the
         transition to the HELLO state that then takes care of the
         communication with the new neighbor while the parent remains in
         state WAIT_NB waiting for additional neighbors to communicate.
         In this case, when the neighbor can no longer be communicated
         with (described as "Neighbor Gone" in the tables below), the
         thread or process created is destroyed and, when a connection-
         oriented protocol is being used to communicate with the
         neighbor, the connection is closed.  The current version of the
         protocol is specified to use TCP for neighbor connections so
         that these will be closed when the neighbor is no longer
         accessible.

   HELLO Nodes are in the HELLO state from when a new neighbor is
         detected until the Hello procedure is completed and a link is
         established (which happens when the Hello procedure enters the
         ESTAB state as described in Section 5.2; during this procedure,
         the states ESTAB, SYNSENT, and SYNRCVD will be used, but these
         are internal to the Hello procedure and are not listed here).
         If the node is notified that the neighbor is no longer in range
         before a link has been established, it returns to the WAIT_NB
         state, and, if appropriate, any additional process or thread
         created to handle the neighbor MAY be destroyed.

   INFO_EXCH  After a link has been set up by the Hello procedure, the
         node transitions to the INFO_EXCH state in which the
         Information Exchange Phase is done.  The node remains in this
         state as long as Information Exchange Phase TLVs (Routing RIB,
         Routing RIB Dictionary, Bundle Offer, Bundle Response) are
         being received.  If the node is notified that the neighbor is
         no longer in range before all information and bundles have been
         exchanged, any associated connection is closed and the node




Lindgren, et al.              Experimental                     [Page 56]

RFC 6693                         PRoPHET                     August 2012


         returns to the WAIT_NB state to await new neighbors.  The
         Timer(keep_alive) is used to ensure that the connection remains
         active.

         In the INFO_EXCH state, the nodes at both ends of the
         established link are able to update their delivery
         predictability information using data from the connected peer
         and then make offers of bundles for exchange which may be
         accepted or not by the peer.  To manage these processes, each
         node acts both as an Initiator and a Listener for the
         Information Exchange Phase processes, maintaining subsidiary
         state machines for the two roles.  The Initiator and Listener
         terms refer to the sending of the Routing RIB information: it
         is perhaps counterintuitive that the Listener becomes the
         bundle offeror and the Initiator the bundle acceptor during the
         bundling passing part.

         The protocol is designed so that the two exchanges MAY be
         carried out independently but concurrently, with the messages
         multiplexed onto on a single bidirectional link (such as is
         provided by the TCP connection).  Alternatively, the exchanges
         MAY be carried out partially or wholly sequentially if
         appropriate for the implementation.  The Information Exchange
         Phase is explained in more detail in Section 3.2.

         When an empty Bundle Response TLV (i.e., no more bundles to
         send) is received, the node starts the Timer(next_exchange).
         When this timer expires, assuming that the neighbor is still
         connected, the Initiator reruns the Information Exchange Phase.
         If there is only one neighbor connected at this time, this will
         have the effect of further increasing the delivery
         predictability for this node in the neighbor, and changing the
         delivery predictabilities as a result of the transitive
         property (Equation 3).  If there is more than one neighbor
         connected or other communication opportunities have happened
         since the previous information exchange occurred, then the
         changes resulting from these other encounters will be passed on
         to the connected neighbor.  The next_exchange timer is
         restarted once the information exchange has completed again.

         If one or more new bundles are received by this node while
         waiting for the Timer(next_exchange) to expire and the delivery
         predictabilities indicate that it would be appropriate to
         forward some or all of the bundles to the connected node, the
         bundles SHOULD be immediately offered to the connected neighbor
         and transferred if accepted.





Lindgren, et al.              Experimental                     [Page 57]

RFC 6693                         PRoPHET                     August 2012


    State: WAIT_NB

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |   New Neighbor   | Start Hello procedure for neighbor|   HELLO   |
    |                  |  Keep waiting for more neighbors  |  WAIT_NB  |
    +==================================================================+



    State: HELLO

    +==================================================================+
    |    Condition     |               Action              | New State |
    +==================+===================================+===========+
    |  Hello TLV rcvd  |                                   |   HELLO   |
    +------------------+-----------------------------------+-----------+
    | Enter ESTAB state|  Start Information Exchange Phase | INFO_EXCH |
    +------------------+-----------------------------------+-----------+
    |  Neighbor Gone   |                                   |  WAIT_NB  |
    +==================================================================+



    State: INFO_EXCH

    +==================================================================+
    |    Condition     |               Action              | New State |
    +==================+===================================+===========+
    |     On entry     |     Start Timer(keep-alive)       |           |
    |                  |        Uses Hello Timer interval  | INFO_EXCH |
    +------------------+-----------------------------------+-----------+
    |Info Exch TLV rcvd| (processed by subsidiary state    |           |
    |                  |                         machines) | INFO_EXCH |
    +------------------+-----------------------------------+-----------+
    | No more bundles  |     Start Timer(next_exchange)    | INFO_EXCH |
    +------------------+-----------------------------------+-----------+
    | Keep-alive expiry|     Send Hello SYN message        | INFO_EXCH |
    +------------------+-----------------------------------+-----------+
    |  Hello SYN rcvd  |     Record reception              |           |
    |                  |     Restart Timer(keep-alive)     | INFO_EXCH |
    +------------------+-----------------------------------+-----------+
    |  Neighbor Gone   |                                   |  WAIT_NB  |
    +==================================================================+






Lindgren, et al.              Experimental                     [Page 58]

RFC 6693                         PRoPHET                     August 2012


   The keep-alive messages (messages with Hello SYN TLV) are processed
   by the high-level state machine in the INFO_EXCH state.  All other
   messages are delegated to the subsidiary state machines of the
   Information Exchange Phase described in Section 5.3.  The receipt of
   keep-alive messages is recorded and may be used by the subsidiary
   machines to check if the peer is still functioning.  The connection
   will be aborted (as described in Section 4.3.1) if several keep-alive
   messages are not received.

5.2.  Hello Procedure



   The Hello procedure is described by the following rules and state
   tables.  In this section, the messages sent consist of the PRoPHET
   header and a single Hello TLV (see Figure 4 and Section 4.3.1) with
   the HF (Hello Function) field set to the specified value (SYN,
   SYNACK, ACK or RSTACK).

   The state of the L flag in the latest SYN or SYNACK message is
   recorded in the node that receives the message.  If the L flag is set
   (value 1), the receiving node MUST send the payload length for each
   bundle that it offers to the peer during the Information Exchange
   Phase.

   The rules and state tables use the following operations:

   o  The "Update Peer Verifier" operation is defined as storing the
      values of the Sender Instance and Sender Local Address fields from
      a Hello SYN or Hello SYNACK function message received from the
      entity at the far end of the link.

   o  The procedure "Reset the link" is defined as:

      When using TCP or other reliable connection-oriented transport:
           Close the connection and terminate any separate thread or
           process managing the connection.

      Otherwise:

           1.   Generate a new instance number for the link.

           2.   Delete the peer verifier (set to zero the values of
                Sender Instance and Sender Local Address previously
                stored by the Update Peer Verifier operation).

           3.   Send a SYN message.

           4.   Transition to the SYNSENT state.




Lindgren, et al.              Experimental                     [Page 59]

RFC 6693                         PRoPHET                     August 2012


   o  The state tables use the following Boolean terms and operators:

      A    The Sender Instance in the incoming message matches the value
           stored from a previous message by the "Update Peer Verifier"
           operation.

      B    The Sender Instance and Sender Local Address fields in the
           incoming message match the values stored from a previous
           message by the "Update Peer Verifier" operation.

      C    The Receiver Instance and Receiver Local Address fields in
           the incoming message match the values of the Sender Instance
           and Sender Local Address used in outgoing Hello SYN, Hello
           SYNACK, and Hello ACK messages.

      SYN    A Hello SYN message has been received.

      SYNACK A Hello SYNACK message has been received.

      ACK    A Hello ACK message has been received.

      &&     Represents the logical AND operation

      ||     Represents the logical OR operation

      !      Represents the logical negation (NOT) operation.

   o  A timer is required for the periodic generation of Hello SYN,
      Hello SYNACK, and Hello ACK messages.  The value of the timer is
      announced in the Timer field.  To avoid synchronization effects,
      uniformly distributed random jitter of +/-5% of the Timer field
      SHOULD be added to the actual interval used for the timer.

      There are two independent events: the timer expires, and a packet
      arrives.  The processing rules for these events are:

             Timer Expires:  Reset Timer
                             If state = SYNSENT Send SYN message
                             If state = SYNRCVD Send SYNACK message
                             If state = ESTAB   Send ACK message











Lindgren, et al.              Experimental                     [Page 60]

RFC 6693                         PRoPHET                     August 2012


             Packet Arrives:
                 If incoming message is an RSTACK message:
                     If (A && C && !SYNSENT) Reset the link
                     Else discard the message.
                 If incoming message is a SYN, SYNACK, or ACK message:
                     Response defined by the following State Tables.
                 If incoming message is any other PRoPHET TLV and
                     state != ESTAB:
                     Discard incoming message.
                     If state = SYNSENT Send SYN message(Note 1)
                     If state = SYNRCVD Send SYNACK message(Note 1)

            Note 1: No more than two SYN or SYNACK messages should be
            sent within any time period of length defined by the timer.

   o  A connection across a link is considered to be achieved when the
      protocol reaches the ESTAB state.  All TLVs, other than Hello
      TLVs, that are received before synchronization is achieved will be
      discarded.

5.2.1.  Hello Procedure State Tables



    State: SYNSENT

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |   SYNACK && C    |       Update Peer Verifier;       |   ESTAB   |
    |                  |       Send ACK message            |           |
    +------------------+-----------------------------------+-----------+
    |   SYNACK && !C   |       Send RSTACK message         |  SYNSENT  |
    +------------------+-----------------------------------+-----------+
    |       SYN        |       Update Peer Verifier;       |  SYNRCVD  |
    |                  |       Send SYNACK message         |           |
    +------------------+-----------------------------------+-----------+
    |       ACK        |       Send RSTACK message         |  SYNSENT  |
    +==================================================================+














Lindgren, et al.              Experimental                     [Page 61]

RFC 6693                         PRoPHET                     August 2012


    State: SYNRCVD

    +==================================================================+
    |    Condition     |               Action              | New State |
    +==================+===================================+===========+
    |   SYNACK && C    |       Update Peer Verifier;       |   ESTAB   |
    |                  |       Send ACK message            |           |
    +------------------+-----------------------------------+-----------+
    |   SYNACK && !C   |       Send RSTACK message         |  SYNRCVD  |
    +------------------+-----------------------------------+-----------+
    |       SYN        |       Update Peer Verifier;       |  SYNRCVD  |
    |                  |       Send SYNACK message         |           |
    +------------------+-----------------------------------+-----------+
    |  ACK && B && C   |       Send ACK message            |   ESTAB   |
    +------------------+-----------------------------------+-----------+
    | ACK && !(B && C) |       Send RSTACK message         |  SYNRCVD  |
    +==================================================================+



    State: ESTAB

    +==================================================================+
    |    Condition    |               Action               | New State |
    +=================+====================================+===========+
    |  SYN || SYNACK  | Send ACK message (notes 2 and 3)  |   ESTAB   |
    +-----------------+------------------------------------+-----------+
    |  ACK && B && C  | Send ACK message (note 3)          |   ESTAB   |
    +-----------------+------------------------------------+-----------+
    | ACK && !(B && C)|          Send RSTACK message       |   ESTAB   |
    +==================================================================+

      Note 2: No more than two ACK messages should be sent within any
      time period of length defined by the timer.  Thus, one ACK message
      MUST be sent every time the timer expires.  In addition, one
      further ACK message may be sent between timer expirations if the
      incoming message is a SYN or SYNACK.  This additional ACK allows
      the Hello functions to reach synchronization more quickly.

      Note 3: No more than one ACK message should be sent within any
      time period of length defined by the timer.

5.3.  Information Exchange Phase



   After the Hello messages have been exchanged, and the nodes are in
   the ESTAB state, the Information Exchange Phase, consisting of the
   RIB Exchange and Bundle Passing Sub-Phases, is initiated.  This
   section describes the procedure and shows the state transitions



Lindgren, et al.              Experimental                     [Page 62]

RFC 6693                         PRoPHET                     August 2012


   necessary in these sub-phases; the following sections describe in
   detail the various TLVs passed in these phases.  On reaching the
   ESTAB state in the high-level HELLO state, there is an automatic
   transition to the INFO_EXCH high-level state.

   PRoPHET runs over a bidirectional transport as documented in
   Section 1.2 so that when a pair of nodes (A and B) have reached the
   ESTAB state, they are able to perform the Information Exchange Phase
   processes for both the A-to-B and B-to-A directions over the link
   that has just been established.  In principle, these two processes
   are independent of each other and can be performed concurrently.
   However, complete concurrency may not be the most efficient way to
   implement the complete process.  As explained in Section 3.2.1, the
   Routing Information Base Dictionary is a shared resource assembled
   from a combination of information generated locally on each node and
   information passed from the peer node.  Overlaps in this information,
   and hence the amount of information that has to be passed between the
   nodes, can be minimized by sequential rather than concurrent
   operation of the dictionary generation and update processes.  It may
   also be possible to reduce the number of bundles that need to be
   offered by the second offeror by examining the offers received from
   the first offeror -- there is no need for the second offeror to offer
   a bundle that is already present in the first offeror's offer list,
   as it will inevitably be refused.

   All implementations MUST be capable of operating in a fully
   concurrent manner.  Each implementation needs to define a policy,
   which SHOULD be configurable, as to whether it will operate in a
   concurrent or sequential manner during the Information Exchange
   Phase.  If it is to operate sequentially, then further choices can be
   made as to whether to interleave dictionary, offer, and response
   exchange parts, or to complete all parts in one direction before
   initiating the other direction.

   Sequential operation will generally minimize the amount of data
   transferred across the PRoPHET link and is especially appropriate if
   the link is half-duplex.  However it is probably not desirable to
   postpone starting the information exchange in the second direction
   until the exchange of bundles has completed.  If the contact between
   the nodes ends before all possible bundles have been exchanged, it is
   possible that postponing the start of bundle exchange in the second
   direction can lead to bundle exchange being skewed in favor of one
   direction over the other.  It may be preferable to share the
   available contact time and bandwidth between directions by
   overlapping the Information Exchange Phases and running the actual
   bundle exchanges concurrently if possible.  Also, if encounters
   expected in the current PRoPHET zone are expected to be relatively
   short, it MAY not be appropriate to use sequential operation.



Lindgren, et al.              Experimental                     [Page 63]

RFC 6693                         PRoPHET                     August 2012


   One possible interleaving strategy is to alternate between sending
   from the two nodes.  For example, if the Hello SYN node sends its
   initial dictionary entries while the Hello SYNACK node waits until
   this is complete, the Hello SYNACK node can then prune its proposed
   dictionary entries before sending in order to avoid duplication.
   This approach can be repeated for the second tranche of dictionary
   entries needed for the Bundle Offers and Responses, and also for the
   Bundle Offers, where any bundles that are offered by the Hello SYN
   node that are already present in the Hello SYNACK node need not be
   offered to the Hello SYN node.  This approach is well suited to a
   transport protocol and physical medium that is effectively half-
   duplex.

   At present, the decision to operate concurrently or sequentially is
   purely a matter of local policy in each node.  If nodes have
   inconsistent policies, the behavior at each encounter will depend on
   which node takes the SYN role; this is a matter of chance depending
   on random timing of the start of communications during the encounter.

   To manage the information transfer, two subsidiary state machines are
   created in each node to control the stages of the RIB Exchange Sub-
   Phase and Bundle Passing Sub-Phase processes within the INFO_EXCH
   high-level state as shown in Figure 12.  Each subsidiary state
   machine consists of two essentially independent components known as
   the "Initiator role" and the "Listener role".  One of these
   components is instantiated in each node.  The Initiator role starts
   the Information Exchange Phase in each node and the Listener role
   responds to the initial messages, but it is not a passive listener as
   it also originates messages.  The transition from the ESTAB state is
   a "forking" transition in that it starts both subsidiary state
   machines.  The two subsidiary state machines operate in parallel for
   as long as the neighbor remains in range and connected.



















Lindgren, et al.              Experimental                     [Page 64]

RFC 6693                         PRoPHET                     August 2012


   + - - - - - - - - +                              + - - - - - - - - +

   |    SYN node     |    PRoPHET messages with:    |   SYNACK node   |

   | +-------------+ | A. Delivery Predictabilities | +-------------+ |
     | Subsidiary  |--->---->---->---->---->---->---->| Subsidiary  |
   | |   State     | | C. Bundle Responses          | |   State     | |
     | Machine 1:  |                                  | Machine 1:  |
   | |  Initiator  | | B. Bundle Offers             | |  Listener   | |
     |    Role     |<----<----<----<----<----<----<---|    Role     |
   | +-------------+ | D. Requested Bundles         | +-------------+ |

   | +-------------+ | A. Delivery Predictabilities | +-------------+ |
     | Subsidiary  |<----<----<----<----<----<----<---| Subsidiary  |
   | |   State     | | C. Bundle Responses          | |   State     | |
     | Machine 2:  |                                  | Machine 2:  |
   | |  Listener   | | B. Bundle Offers             | |  Initiator  | |
     |    Role     |--->---->---->---->---->---->---->|    Role     |
   | +-------------+ | D. Requested Bundles         | +-------------+ |

   + - - - - - - - - +                              + - - - - - - - - +

         The letters (A - D) indicate the sequencing of messages.

      Figure 12: Information Exchange Phase Subsidiary State Machines

   These subsidiary state machines can be thought of as mirror images:
   for each state machine, one node takes on the Initiator role while
   the other node takes on the Listener role.  TLVs sent by a node from
   the Initiator role will be processed by the peer node in the Listener
   role and vice versa.  As indicated in Figure 12, the Initiator role
   handles sending that node's current set of delivery predictabilities
   for known destinations to the Listener role node.  The Listener role
   node uses the supplied values to update its delivery predictabilities
   according to the update algorithms described in Section 2.1.2.  It
   then decides which bundles that it has in store should be offered for
   transfer to the Initiator role node as a result of comparing the
   local predictabilities and those supplied by the Initiator node.
   When these offers are delivered to the Initiator role node, it
   decides which ones to accept and supplies the Listener role node with
   a prioritized list of bundles that it wishes to accept.  The Listener
   role node then sends the requested bundles.

   These exchanges are repeated periodically for as long as the nodes
   remain in contact.  Additionally, if new bundles arrive from other
   sources, they may be offered, accepted, and sent in between these
   exchanges.




Lindgren, et al.              Experimental                     [Page 65]

RFC 6693                         PRoPHET                     August 2012


   The PRoPHET protocol is designed so that in most cases the TLV type
   determines the role in which it will be processed on reception.  The
   only exception to this is that both roles may send RIB Dictionary
   TLVs: the Initiator role sends dictionary entries for use in the
   subsequent RIB TLV(s), and the Listener role may send additional
   dictionary entries for use in subsequent Bundle Offer TLVs.  The two
   cases are distinguished by a TLV flag to ensure that they are
   processed in the right role context on reception.  If this flag was
   not provided, there are states where both roles could accept the RIB
   Dictionary TLV, making it impossible to ensure that the correct role
   state machine accepts the RIB Dictionary TLV.  Note that the correct
   updates would be made to the dictionary whichever role processed the
   TLV and that the ambiguity would not arise if the roles are adopted
   completely sequentially, i.e., if the RIB Exchange Sub-Phase and
   associated Bundle Passing Sub-Phase run to completion in one
   direction before the process for the reverse direction is started.

   If sequential operation is selected, the node that sent the Hello SYN
   function message MUST be the node that sends the first message in the
   Information Exchange Phase process.  This ensures that there is a
   well-defined order of events with the Initiator role in the Hello SYN
   node (i.e., the node identified by String ID 0) starting first.  The
   Hello SYNACK node MAY then postpone sending its first message until
   the Listener role state machine in the Hello SYNACK node has reached
   any of a number of points in its state progression according to
   locally configured policy and the nature of the physical link for the
   current encounter between the nodes as described above.  If
   concurrent operation is selected, the Hello SYNACK node can start
   sending messages immediately without waiting to receive messages from
   the peer.

   The original design of the PRoPHET protocol allowed it to operate
   over unreliable datagram-type transports as well as the reliable, in-
   order delivery transport of TCP that is currently specified.  When
   running over TCP, protocol errors and repeated timeouts during the
   Information Exchange Phase SHOULD result in the connection being
   terminated.

5.3.1.  State Definitions for the Initiator Role



   The state machine component with the Initiator role in each node
   starts the transfer of information from one node to its peer during
   the Information Exchange Phase.  The process from the Initiator's
   point of view does the following:

   o  The Initiator role determines the set of delivery predictabilities
      to be sent to the peer node and sends RIB dictionary entries
      necessary to interpret the set of RIB predictability values that



Lindgren, et al.              Experimental                     [Page 66]

RFC 6693                         PRoPHET                     August 2012


      are sent after the dictionary updates.  On second and subsequent
      executions of this state machine during a single session with the
      same peer, there may be no RIB Dictionary entries to send.  Either
      an empty TLV can be sent or the TLV can be omitted.

   o  The Initiator then waits to receive any RIB Dictionary updates
      followed by bundle offers from the Listener role on the peer node.

   o  The Initiator determines which of the bundle offers should be
      accepted and, if necessary, reorders the offers to suit its own
      priorities.  The possibly reordered list of accepted bundles is
      sent to the peer node using one or more bundle responses.

   o  The peer then sends the accepted bundles to the Initiator in turn.

   o  Assuming that the link remains open during the bundle sending
      process, the Initiator signals that the Bundle Passing Sub-Phase
      is complete by sending a message with an empty Bundle Response TLV
      (i.e, with the Bundle Offer Count set to 0 and no bundle offers
      following the TLV header).

   o  When the bundle transfer is complete, the Initiator starts the
      Timer(next_exchange).  Assuming that the connection to the
      neighbor remains open, when the timer expires, the Initiator
      restarts the Information Exchange Phase.  During this period,
      Hello SYN messages are exchanged as keep-alives to check that the
      neighbor is still present.  The keep-alive mechanism is common to
      the Initiator and Listener machines and is handled in the high-
      level state machine (see Section 5.1.

   A timer is provided that restarts the Initiator role state machine if
   Bundle Offers are not received after sending the RIB.  If this node
   receives a Hello ACK message containing an Error TLV indicating there
   has been a protocol problem, then the connection MUST be terminated.

   The following states are used:

   CREATE_DR
      The initial transition to this state from the ESTAB state is
      immediate and automatic for the node that sent the Hello SYN
      message.  For the peer (Hello SYNACK sender) node, it may be
      immediate for nodes implementing a fully concurrent process or may
      be postponed until the corresponding Listener has reached a
      specified state if a sequential process is configured in the node
      policy.






Lindgren, et al.              Experimental                     [Page 67]

RFC 6693                         PRoPHET                     August 2012


      The local dictionary is initialized when this state is entered for
      the first time from the ESTAB state.  The initial state of the
      dictionary contains two entries: the EID of the node that sent the
      Hello SYN (String ID 0) and the EID of the node that sent the
      Hello SYNACK (String ID 1).  If the peer reports via a Hello ACK
      message containing an Error TLV reporting a Dictionary Conflict or
      Bad String ID error, then the connection MUST be terminated.

      The CREATE_DR state will be entered in the same way from the
      REQUEST state when the Timer(next_exchange) expires, signaling the
      start of a new round of information exchange and bundle passing.

      When in this state:

      *  Determine the destination EIDs for which delivery
         predictabilities will be sent to the peer in a RIB TLV, if any.
         Record the prior state of the local dictionary (assuming that
         String IDs are numbers allocated sequentially, the state
         information needed is just the highest ID used before this
         process started) so that the process can be restarted if
         necessary.  Update the local dictionary if any new EIDS are
         required; format one or more RIB Dictionary TLVs and one or
         more RIB TLVs and send them to the peer.  If there are no
         dictionary entries to send, TLVs with zero entries MAY be sent,
         or the TLV can be omitted, but an empty RIB TLV MUST be sent if
         there is no data to send.  The RIB Dictionary TLVs generated
         here MUST have the Sent by Listener flag set to 0 to indicate
         that they were sent by the Initiator.

      *  If an Error TLV indicating a Dictionary Conflict or
         Bad String ID is received during or after sending the RIB
         Dictionary TLVs and/or the RIB TLVs, abort any in-progress
         Initiator or Listener process, and terminate the connection to
         the peer.

      *  Start a timer (known as Timer(info)) and transition to the
         SEND_DR state.

      Note that when (and only when) running over a transport protocol
      such as TCP, both the RIB Dictionary and RIB information MAY be
      spread across multiple TLVs and messages if required by known
      constraints of the transport protocol or to reduce the size of
      memory buffers.  Alternatively, the information can be formatted
      using a single RIB Dictionary TLV and a single RIB TLV.  These
      TLVs may be quite large, so it may be necessary to segment the
      message either using the PRoPHET submessage capability or, if the
      transport protocol has appropriate capabilities, using those
      inherent capabilities.  This discussion of segmentation applies to



Lindgren, et al.              Experimental                     [Page 68]

RFC 6693                         PRoPHET                     August 2012


      the other states and the bundle offer and bundle response messages
      and will not be repeated.

      If more than one RIB TLV is to be used, all but the last one have
      the "More RIB TLVs" flag set to 1 in the TLV flags.  It is not
      necessary to distinguish the last RIB Dictionary TLV because the
      actions taken at the receiver are essentially passive (recording
      the contents), and the sequence is ended by the sending of the
      first RIB TLV.

   SEND_DR
      In this state, the Initiator node expects to be receiving Bundle
      Offers and sending Bundle Responses.  The Initiator node builds a
      list of bundles offered by the peer while in this state:

      *  Clear the set of bundles offered by the peer on entry to the
         state.

      *  If the Timer(info) expires, re-send the RIB Dictionary and RIB
         information sent in the previous CREATE_DR state using the
         stored state to re-create the information.  The RIB dictionary
         update process in the peer is idempotent provided that the
         mappings between the EID and the String ID in the re-sent RIB
         Dictionary TLVs are the same as in the original.  This means
         that it does not matter if some of the RIB Dictionary TLVs had
         already been processed in the peer.  Similarly, re-sending RIB
         TLVs will not cause a problem.

      *  If a message with a RIB Dictionary TLV marked as sent by a
         Listener is received, update the local dictionary based on the
         received TLV.  If any of the entries in the RIB Dictionary TLV
         conflict with existing entries (i.e., an entry is received that
         uses the same String ID as some previously received entry but
         the EID in the entry is different), send a Response message
         with an Error TLV containing a Dictionary Conflict indicator,
         abort any in-progress Initiator or Listener process, and
         terminate the connection to the peer.  Note that in some
         circumstances no dictionary updates are needed, and the first
         message received in this state will carry a Bundle Offer TLV.

      *  If a message with a Bundle Offer TLV is received, restart the
         Timer(info) if the "More Offer/Response TLVs Following" flag is
         set in the TLV; otherwise, stop the Timer(info).  Then process
         any PRoPHET ACKs in the TLV by informing the bundle protocol
         agent, and add the bundles offered in the TLV to the set of
         bundles offered.  If the "More Offer/Response TLVs Following"
         flag is set in the TLV, wait for further Bundle Offer TLVs.  If
         a Bundle Offer TLV is received with a String ID that is not in



Lindgren, et al.              Experimental                     [Page 69]

RFC 6693                         PRoPHET                     August 2012


         the dictionary, send a message with an Error TLV containing a
         Bad String ID indicator, abort any in-progress Initiator or
         Listener process, and terminate the connection to the peer.

      *  If the "More Offer/Response TLVs Following" flag is clear in
         the last Bundle Offer TLV received, inspect the set of bundles
         offered to determine the set of bundles that are to be accepted
         using the configured queueing policy.  Record the set of
         bundles accepted so that reception can be checked in the Bundle
         Passing Sub-Phase.  Format one or more Bundle Response TLVs
         flagging the accepted offers and send them to the peer.  If
         more than one Bundle Response TLV is sent, all but the last one
         should have the "More Offer/Response TLVs Following" flag set
         to 1.  At least one Bundle Response TLV MUST be sent even if
         the node does not wish to accept any of the offers.  In this
         case, the Bundle Response TLV contains an empty set of
         acceptances.

      *  If an Error TLV indicating a Bad String ID is received during
         or after sending the Bundle Response TLVs, abort any in-
         progress Initiator or Listener process, re-initialize the local
         dictionary, and terminate the connection to the peer.

      *  Restart the Timer(info) timer in case the peer does not start
         sending the requested bundles.

      *  Transition to state REQUEST.

   REQUEST
      In this state, the Initiator node expects to be receiving the
      bundles accepted in the Bundle Response TLV(s):

      *  Keep track of the bundles received and delete them from the set
         of bundles accepted.

      *  If the Timer(info) expires while waiting for bundles, format
         and send one or more Bundle Response TLVs listing the bundles
         previously accepted but not yet received.  If more than one
         Bundle Response TLV is sent, all but the last one should have
         the "More Offer/Response TLVs Following" flag set to 1.

      *  If an Error TLV indicating a Bad String ID is received during
         or after sending the Bundle Response TLVs, abort any in-
         progress Initiator or Listener process, re-initialize the local
         dictionary, and terminate the connection to the peer.

      *  Restart the Timer(info) timer after each bundle is received in
         case the peer does not continue sending the requested bundles.



Lindgren, et al.              Experimental                     [Page 70]

RFC 6693                         PRoPHET                     August 2012


      *  When all the requested bundles have been received, format a
         Bundle Response TLV with the Bundle Offer Count set to zero and
         with the "More Offer/Response TLVs Following" flag cleared to 0
         to signal completion to the peer node.  Also, signal the
         Listener in this node that the Initiator has completed.  If the
         peer node is using a sequential policy, the Listener may still
         be in the initial state, in which case, it needs to start a
         timer to ensure that it detects if the peer fails to start the
         Initiator state machine.  Thereafter, coordinate with the
         Listener state machine in the same node: when the Listener has
         received the completion notification from the peer node and
         this Initiator has sent its completion notification, start
         Timer(next_exchange).

      *  If the Timer(next_exchange) expires, transition to state
         CREATE_DR to restart the Information Exchange Phase.

      Note that if Timer(info) timeout occurs a number of times
      (configurable, typically 3) without any bundles being received,
      then this SHOULD generally be interpreted as the problem that the
      link to the peer is no longer functional and the session should be
      terminated.  However, some bundles may be very large and take a
      long time to transmit.  Before terminating the session, this state
      machine needs to check if a large bundle is actually being
      received although no new completed bundles have been received
      since the last expiry of the timer.  In this case the timer should
      be restarted without sending the Bundle Response TLV.  Also, if
      the bundles are being exchanged over a transport protocol that can
      detect link failure, then the session MUST be terminated if the
      bundle exchange link is shut down because it has failed.

5.3.2.  State Definitions for the Listener Role



   The state machine component with the Listener role in each node
   initially waits to receive a RIB Dictionary update followed by a set
   of RIB delivery predictabilities during the Information Exchange
   Phase.  The process from the point of view of the Listener does the
   following:

   o  Receive RIB Dictionary updates and RIB values from the peer.  Note
      that in some circumstances no dictionary updates are needed, and
      the RIBD TLV will contain no entries or may be omitted completely.

   o  When all RIB messages have been received, the delivery
      predictability update algorithms are run (see Section 2.1.2) using
      the values received from the Initiator node and applying any of
      the optional optimizations configured for this node (see
      Section 2.1.3).



Lindgren, et al.              Experimental                     [Page 71]

RFC 6693                         PRoPHET                     August 2012


   o  Using the updated delivery predictabilities and the queueing
      policy and forwarding strategy configured for this node (see
      Section 2.1.4) examine the set of bundles currently stored in the
      Listener node to determine the set of bundles to be offered to the
      Initiator and order the list according to the forwarding strategy
      in use.  The Bundle Offer TLVs are also used to notify the peer of
      any PRoPHET ACKs that have been received by the Listener role
      node.

   o  Send the list of bundles in one or more bundle offers, preceded if
      necessary by one or more RIB dictionary updates to add any EIDs
      required for the source or destination EIDs of the offered
      bundles.  These updates MUST be marked as being sent by the
      Listener role so that they will be processed by the Initiator role
      in the peer.

   o  Wait for the Initiator to send bundle responses indicating which
      bundles should be sent and possibly a modified order for the
      sending.  Send the accepted bundles in the specified order.  The
      bundle sending will normally be carried out over a separate
      connection using a suitable DTN convergence layer.

   o  On completion of the sending, wait for a message with an empty
      Bundle Response TLV indicating correct completion of the process.

   o  The Listener process will be notified if any new bundles or
      PRoPHET ACKs are received by the node after the completion of the
      bundle sending that results from this information exchange.  The
      forwarding policy and the current delivery predictabilities will
      then be applied to determine if this information should be sent to
      the peer.  If it is determined that one or more bundles and/or
      ACKs ought to be forwarded, a new set of bundle offers are sent to
      the peer.  If the peer accepts them by sending bundle responses,
      the bundles and/or ACKS are transferred as previously.

   o  Periodically, the Initiator in the peer will restart the complete
      information exchange by sending a RIB TLV that may be, optionally,
      preceded by RIB Dictionary entries if they are required for the
      updated RIB.

   Timers are used to ensure that the Listener does not lock up if
   messages are not received from the Initiator in a timely fashion.
   The Listener is restarted if the RIB is not received, and a Hello ACK
   message is sent to force the Initiator to restart.  If bundle
   response messages are not received in a timely fashion, the Listener
   re-sends the bundle offers and associated dictionary updates.  The
   following states are used:




Lindgren, et al.              Experimental                     [Page 72]

RFC 6693                         PRoPHET                     August 2012


   WAIT_DICT
      The Listener subsidiary state machine transitions to this state
      automatically and immediately from the state ESTAB in both peers.
      This state will be entered in the same way if the
      Timer(next_exchange) expires in the peer, signaling the start of a
      new round of information exchange and bundle passing.  This will
      result in one or more RIB TLVs being sent to the Listener by the
      peer node's Initiator.

      *  When a RIB Dictionary TLV is received, use the TLV to update
         the local dictionary, start or (if it is running) restart the
         Timer(peer) and transition to state WAIT_RIB.  If any of the
         entries in the RIB Dictionary TLV conflict with existing
         entries (i.e., an entry is received that uses the same String
         ID as some previously received entry, but the EID in the entry
         is different), send a Response message with an Error TLV
         containing a Dictionary Conflict indicator, abort any in-
         progress Initiator or Listener process, and terminate the
         connection to the peer.

      *  If a Hello ACK message is received from the peer node,
         transition to state WAIT_DICT and restart the process.

      If multiple timeouts occur (configurable, typically 3), assume
      that the link is broken and terminate the session.  Note that the
      RIB Dictionary and RIB TLVs may be combined into a single message.
      The RIB TLV should be passed on to be processed in the WAIT_RIB
      state.

   WAIT_RIB
      In this state, the Listener expects to be receiving one or more
      RIB TLVs and possibly additional RIB Dictionary TLVs.

      *  On entry to this state, clear the set of received delivery
         predictabilities.

      *  Whenever a new message is received, restart the Timer(peer)
         timer.

      *  If a RIB dictionary TLV is received, use it to update the local
         dictionary and remain in this state.  If any of the entries in
         the RIB Dictionary TLV conflict with existing entries (i.e., an
         entry is received that uses the same String ID as some
         previously received entry, but the EID in the entry is
         different), send a message with an Error TLV containing a
         Dictionary Conflict indicator, abort any in-progress Initiator
         or Listener process, and terminate the connection to the peer.




Lindgren, et al.              Experimental                     [Page 73]

RFC 6693                         PRoPHET                     August 2012


      *  If a RIB TLV is received, record the received delivery
         predictabilities for use in recalculating the local delivery
         predictabilities.  If a delivery predictability value is
         received for an EID that is already in the set of received
         delivery predictabilities, overwrite the previously received
         value with the latest value.  If a delivery predictability
         value is received with a String ID that is not in the
         dictionary, send a message with an Error TLV containing a
         Bad String ID indicator, abort any in-progress Initiator or
         Listener process, and terminate the connection to the peer.

      *  When a RIB TLV is received with the "More RIB TLVs" flag
         cleared, initiate the recalculation of delivery
         predictabilities and stop the Timer(peer).  Use the revised
         delivery predictabilities and the configured queueing and
         forwarding strategies to create a list of bundles to be offered
         to the peer node.

      *  Record the state of the local dictionary in case the offer
         procedure has to be restarted.  Determine if any new dictionary
         entries are required for use in the Bundle Offer TLV(s).  If
         so, record them in the local dictionary, then format and send
         RIB Dictionary entries in zero or more RIB Dictionary TLV
         messages to update the dictionary in the peer if necessary.

      *  Format and send Bundle Offer TLV(s) carrying the identifiers of
         the bundles to be offered together with any PRoPHET ACKs
         received or generated by this node.  If more than one Bundle
         Offer TLV is sent, all but the last Bundle Offer TLV sent MUST
         have the "More Offer/Response TLVs Following" flag set to 1.

      *  When all Bundle Offer TLVs have been sent, start the
         Timer(info) and transition to state OFFER.

      *  If the Timer(peer) expires, send a Hello ACK TLV to the peer,
         restart the timer, and transition to state WAIT_DICT.

      *  If an Error TLV indicating a Dictionary Conflict or
         Bad String ID is received during or after sending the RIB
         Dictionary TLVs and/or the Bundle Offer TLVs, abort any in-
         progress Initiator or Listener process, and terminate the
         connection to the peer.

      *  If a Hello ACK message is received from the peer node,
         transition to state WAIT_DICT and restart the process.






Lindgren, et al.              Experimental                     [Page 74]

RFC 6693                         PRoPHET                     August 2012


   OFFER
      In this state, the Listener expects to be receiving one or more
      Bundle Response TLVs detailing the bundles accepted by the
      Initiator node.  The ordered list of accepted bundles is
      communicated to the bundle protocol agent, which controls sending
      them to the peer node over a separate connection.

      *  When a Bundle Response TLV is received with a non-zero count of
         Bundle Offers, extract the list of accepted bundles and send
         the list to the bundle protocol agent so that it can start
         transmission to the peer node.  Ensure that the order of offers
         from the TLV is maintained.  Restart the Timer(info) unless the
         last Bundle Response TLV received has the "More Offer/
         Response TLVs Following" flag set to 0.  If a Bundle Response
         TLV is received with a String ID that is not in the dictionary,
         send a message with an Error TLV containing a Bad String ID
         indicator, abort any in-progress Initiator or Listener process,
         and terminate the connection to the peer.

      *  After receiving a Bundle Response TLV with the "More Offer/
         Response TLVs Following" flag set to 0 stop the Timer(info) and
         transition to state SND_BUNDLE.

      *  If the Timer(info) expires, send a Hello ACK TLV to the peer,
         restart the timer and transition to state WAIT_DICT.

      *  If a Hello ACK message is received from the peer node,
         transition to state WAIT_DICT and restart the process.

   SND_BUNDLE
      In this state the Listener monitors the sending of bundles to the
      Initiator peer node.  In the event of disruption in transmission,
      the Initiator node will, if possible, re-send the list of bundles
      that were accepted but have not yet been received.  The bundle
      protocol agent has to be informed of any updates to the list of
      bundles to send (this is likely to involve re-sending one or more
      bundles).  Otherwise, the Listener is quiescent in this state.

      *  When a Bundle Response TLV is received with a non-zero count of
         Bundle Offers, extract the list of accepted bundles and update
         the list previously passed to the bundle protocol agent so that
         it can (re)start transmission to the peer node.  Ensure that
         the order of offers from the TLV is maintained so far as is
         possible.  Restart the Timer(info) unless the last Bundle
         Response TLV received has the "More Offer/Response TLVs
         Following" flag set to 0.  If a Bundle Response TLV is received
         with a String ID that is not in the dictionary, send a message
         with an Error TLV containing a Bad String ID indicator, abort



Lindgren, et al.              Experimental                     [Page 75]

RFC 6693                         PRoPHET                     August 2012


         any in-progress Initiator or Listener process, re-initialize
         the local dictionary, and restart the Information Exchange
         Phase as if the ESTAB state had just been reached.

      *  After receiving a Bundle Response TLV with the "More Offer/
         Response TLVs Following" flag set to 0, stop the Timer(info)
         and wait for completion of bundle sending.

      *  If the Timer(info) expires, send a Hello ACK TLV to the peer,
         restart the timer, and transition to state WAIT_DICT.

      *  If a Hello ACK message is received from the peer node,
         transition to state WAIT_DICT and restart the process.

      *  When a Bundle Response TLV is received with a zero count of
         Bundle Offers, the Bundle Passing Sub-Phase is complete.
         Notify the Initiator that the Listener process is complete and
         transition to state WAIT_MORE.

      As explained in the Initiator state REQUEST description, depending
      on the transport protocol (convergence layer) used to send the
      bundles to the peer node, it may be necessary during the bundle
      sending process to monitor the liveness of the connection to the
      peer node in the Initiator process using a timer.

   WAIT_MORE
      In this state, the Listener monitors the reception of new bundles
      that might be received from a number of sources, including

      *  local applications on the node,

      *  other mobile nodes that connect to the node while this
         connection is open, and

      *  permanent connections such as might occur at an Internet
         gateway.

      When the Listener is notified of received bundles, it determines
      if they should be offered to the peer.  The peer may also re-
      initiate the Information Exchange Phase periodically.

      *  When the bundle protocol agent notifies the Listener that new
         bundles and/or new PRoPHET ACKs have been received, the
         Listener applies the selected forwarding policy and the current
         delivery predictabilities to determine if any of the items
         ought to be offered to the connected peer.  If so, it carries





Lindgren, et al.              Experimental                     [Page 76]

RFC 6693                         PRoPHET                     August 2012


         out the same operations as are described in the WAIT_RIB state
         to build and send any necessary RIB Dictionary TLVs and RIB
         TLVs to the Initiator in the peer.

      *  When all Bundle Offer TLVs have been sent, start the
         Timer(info) and transition to state OFFER.

      *  If a RIB dictionary TLV is received, use it to update the local
         dictionary and transition to state WAIT_RIB.  If any of the
         entries in the RIB Dictionary TLV conflict with existing
         entries (i.e., an entry is received that uses the same String
         ID as some previously received entry, but the EID in the entry
         is different), send a message with an Error TLV containing a
         Dictionary Conflict indicator, abort any in-progress Initiator
         or Listener process, and terminate the connection to the peer.

      Note that the RIB Dictionary and RIB TLVs may be combined into a
      single message.  The RIB TLV should be passed on to be processed
      in the WAIT_RIB state.

5.3.3.  Recommendations for Information Exchange Timer Periods



   The Information Exchange Phase (IEP) state definitions include a
   number of timers.  This section provides advice and recommendations
   for the periods that are appropriate for these timers.

   Both Timer(info) and Timer(peer) are used to ensure that the state
   machines do not become locked into inappropriate states if the peer
   node does not apparently respond to messages sent in a timely fashion
   either because of message loss in the network or unresponsiveness
   from the peer.  The appropriate values are to some extent dependent
   on the speed of the network connection between the nodes and the
   capabilities of the nodes executing the PRoPHET implementations.
   Values in the range 1 to 10 seconds SHOULD be used, with a value of 5
   seconds RECOMMENDED as default.  The period should not be set to too
   low a value, as this might lead to inappropriate restarts if the
   hardware is relatively slow or there are large numbers of pieces of
   information to process before responding.  When using a reliable
   transport protocol such as TCP, these timers effectively provide a
   keep-alive mechanism and ensure that a failed connection is detected
   as rapidly as possible so that remedial action can be taken (if
   possible) or the connection shut down tidily if the peer node has
   moved out of range.

   Timer(next_exchange) is used to determine the maximum frequency of
   (i.e., minimum period between) successive re-executions of the
   information exchange state machines during a single session between a
   pair of nodes.  Selection of the timer period SHOULD reflect the



Lindgren, et al.              Experimental                     [Page 77]

RFC 6693                         PRoPHET                     August 2012


   trade-off between load on the node processor and desire for timely
   forwarding of bundles received from other nodes.  It is RECOMMENDED
   that the timer periods used should be randomized over a range from
   50% to 150% of the base value in order to avoid synchronization
   between multiple nodes.  Consideration SHOULD be given to the
   expected length of typical encounters and the likelihood of
   encounters between groups of nodes when setting this period.  Base
   values in the range of 20 to 60 seconds are RECOMMENDED.

5.3.4.  State Tables for Information Exchange



   This section shows the state transitions that nodes go through during
   the Information Exchange Phase.  State tables are given for the
   Initiator role and for the Listener role of the subsidiary state
   machines.  Both nodes will be running machines in each role during
   the Information Exchange Phase, and this can be done either
   concurrently or sequentially, depending on the implementation, as
   explained in Section 5.3.  The state tables in this section should be
   read in conjunction with the state descriptions in Sections 5.3.1 and
   5.3.2.

5.3.4.1.  Common Notation, Operations and Events



   The following notation is used:

   nS            Node that sent the Hello SYN message.

   nA            Node that sent the Hello SYNACK message.

   The following events are common to the Initiator and Listener state
   tables:

   ErrDC         Dictionary Conflict Error TLV received.

   ErrBadSI      Bad String ID Error TLV received.

   HelloAck      Hello ACK TLV received.  This message is delivered to
                 both Initiator and Listener roles in order to cause a
                 restart of the Information Exchange Phase in the event
                 of message loss or protocol problems.











Lindgren, et al.              Experimental                     [Page 78]

RFC 6693                         PRoPHET                     August 2012


   InitStart     Sent by Listener role to Initiator role to signal the
                 Initiator role to commence sending messages to peer.
                 If the Listener instance is running in the node that
                 sent the Hello SYN (nS), then InitStart is signaled
                 immediately when the state is entered.  For the node
                 that sent the Hello SYNACK (nA), InitStart may be
                 signaled immediately if the operational policy requires
                 concurrent operation of the Initiator and Listener
                 roles or postponed until the Listener role state
                 machine has reached a state defined by the configured
                 policy.

   RIBnotlast    RIB TLV received with "More RIB TLVs" flag set to 1.

   RIBlast       RIB TLV received with "More RIB TLVs" flag set to 0.

   REQnotlast    Bundle Response TLV received with More Offer/Response
                 TLVs Following flag set to 1.

   REQlast       Bundle Response TLV received with More Offer/Response
                 TLVs Following flag set to 0.

   RIBDi         RIBD TLV received with Sent by Listener flag set to 0
                 (i.e., it was sent by Initiator role).

   RIBDl         RIBD TLV received with Sent by Listener flag set to 1
                 (i.e., it was sent by Listener role).

   Timeout(info) The Timer(info) has expired.

   Timeout(peer) The Timer(peer) has expired.

   Both the Initiator and Listener state tables use the following common
   operations:

   o  The "Initialize Dictionary" operation is defined as emptying any
      existing local dictionary and inserting the two initial entries:
      the EID of the node that sent the Hello SYN (String ID 0) and the
      EID of the node that sent the Hello SYNACK (String ID 1).

   o  The "Send RIB Dictionary Updates" operation is defined as:

      1.  Determining what dictionary updates will be needed for any
          extra EIDs in the previously selected RIB entries set that are
          not already in the dictionary and updating the local
          dictionary with these EIDs.  The set of dictionary updates may
          be empty if no extra EIDs are needed.  The set may be empty
          even on the first execution if sequential operation has been



Lindgren, et al.              Experimental                     [Page 79]

RFC 6693                         PRoPHET                     August 2012


          selected, this is the second node to start and the necessary
          EIDs were in the set previously sent by the first node to
          start.

      2.  Formatting zero or more RIBD TLVs for the set of dictionary
          updates identified in the "Build RIB Entries" operation and
          sends them to the peer.  The RIBD TLVs MUST have the "Sent by
          Listener" flag set to 0 if the updates are sent by the
          Initiator role and to 1 if sent by the Listener role.  In the
          case of the Initiator role, an empty RIBD TLV MUST be sent
          even if the set of updates is empty in order to trigger the
          Listener state machine.

   o  The "Update Dictionary" operation uses received RIBD TLV entries
      to update the local dictionary.  The received entries are checked
      against the existing dictionary.  If the String ID in the entry is
      already in use, the entry is accepted if the EID in the received
      entry is identical to that stored in the dictionary previously.
      If it is identical, the entry is unchanged, but if it is not a
      Response message with an Error TLV indicating Dictionary Conflict
      is sent to the peer in an Error Response message, the whole
      received RIBD TLV is ignored, and the Initiator and Listener
      processes are restarted as if the ESTAB state has just been
      reached.

   o  The "Abort Exchange" operation is defined as aborting any in-
      progress information exchange state machines and terminating the
      connection to the peer.

   o  The "Start TI" operation is defined as (re)starting the
      Timer(info) timer.

   o  The "Start TP" operation is defined as (re)starting the
      Timer(peer) timer.

   o  The "Cancel TI" operation is defined as canceling the Timer(info)
      timer.

   o  The "Cancel TP" operation is defined as canceling the Timer(info)
      timer.











Lindgren, et al.              Experimental                     [Page 80]

RFC 6693                         PRoPHET                     August 2012


5.3.4.2.  State Tables for the Initiator Role



   The rules and state tables for the Initiator role use the following
   operations:

   o  The "Build RIB Entries" operation is defined as:

      1.  Recording the state of the local dictionary.

      2.  Determining the set of EIDs for which RIB entries should be
          sent during this execution of the Initiator role state machine
          component.  If this is a second or subsequent run of the state
          machine in this node during the current session with the
          connected peer, then the set of EIDs may be empty if no
          changes have occurred since the previous run of the state
          machine.

      3.  Determining and extracting the current delivery predictability
          information for the set of EIDs selected.

   o  The "Send RIB Entries" operation formats one or more RIB TLVs with
      the set of RIB entries identified in the "Build RIB Entries"
      operation and sends them to the peer.  If the set is empty, a
      single RIB TLV with zero entries is sent.  If more than one RIB
      TLV is sent, all but the last one MUST have the "More RIB TLVs"
      flag set to 1; the last or only one MUST have the flag set to 0.

   o  The "Clear Bundle Lists" operation is defined as emptying the
      lists of bundles offered by the peer and bundles requested from
      the peer.

   o  The "Notify ACKs" operation is defined as informing the bundle
      protocol agent that PRoPHET ACKs has been received for one or more
      bundles in a Bundle Offer TLV using the Bundle Delivered interface
      (see Section 2.2).

   o  The "Record Offers" operation is defined as recording all the
      bundles offered in a Bundle Offer TLV in the list of bundles
      offers.

   o  The "Select for Request" operation prunes and sorts the list of
      offered bundles held into the list of requested bundles according
      to policy and the available resources ready for sending to the
      offering node.







Lindgren, et al.              Experimental                     [Page 81]

RFC 6693                         PRoPHET                     August 2012


   o  The "Send Requests" operation is defined as formatting one or more
      non-empty Bundle Response TLVs and sending them to the offering
      node.  If more than one Bundle Offer TLV is sent, all but the last
      one MUST have the "More Offer/Response TLVs Following" flag set to
      1; the last or only one MUST have the flag set to 0.

   o  The "Record Bundle Received" operation deletes a successfully
      received bundle from the list of requests.

   o  The "All Requests Done" operation is defined as formatting and
      sending an empty Bundle Offer TLV, with the "More Offer/Response
      TLVs Following" flag set to 0, to the offering node.

   o  The "Check Receiving" operation is defined as checking with the
      node bundle protocol agent if bundle reception from the peer node
      is currently in progress.  This is needed in case a timeout occurs
      while waiting for bundle reception and a very large bundle is
      being processed.

   o  The "Start NE" operation is defined as (re)starting the
      Timer(next_exchange).

   The following events are specific to the Initiator role state
   machine:

   LastBndlRcvd  Bundle received from peer that is the only remaining
                 bundle in Bundle Requests List.

   NotLastBndlRcvd  Bundle received from peer that is not the only
                 remaining bundle in Bundle Requests List.

   OFRnotlast    Bundle Offer TLV received with "More Offer/Response
                 TLVs Following" flag set to 1.

   OFRlast       Bundle Offer TLV received with "More Offer/Response
                 TLVs Following" flag set to 0

   Timeout(next_exch)  The Timer(next_exchange) has expired













Lindgren, et al.              Experimental                     [Page 82]

RFC 6693                         PRoPHET                     August 2012


    State: CREATE_DR

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |     On Entry     |    If previous state was ESTAB:   |           |
    |                  |         Initialize Dictionary     |           |
    |                  |    Always:                        |           |
    |                  |         Build RIB Entries         |           |
    |                  |         Wait for Init Start       | CREATE_DR |
    +------------------+-----------------------------------+-----------+
    |    InitStart     |    Send RIB Dictionary Updates    |           |
    |                  |    Send RIB Entries               |           |
    |                  |    Start TI                       |  SEND_DR  |
    +------------------+-----------------------------------+-----------+
    |      ErrDC       |           Abort Exchange          |(finished) |
    +------------------+-----------------------------------+-----------+
    |     ErrBadSI     |           Abort Exchange          |(finished) |
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |           Abort Exchange          | CREATE_DR |
    +==================================================================+






























Lindgren, et al.              Experimental                     [Page 83]

RFC 6693                         PRoPHET                     August 2012


    State: SEND_DR

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |     On Entry     |         Clear Bundle Lists        |  SEND_DR  |
    +------------------+-----------------------------------+-----------+
    |  Timeout(info)   |   Send RIB Dictionary Updates     |           |
    |                  |   Send RIB Entries (note 1)       |  SEND_DR  |
    +------------------+-----------------------------------+-----------+
    |  RIBDl received  |   Update Dictionary (note 2)      |           |
    |                  |   If Dictionary Conflict found:   |           |
    |                  |           Abort Exchange          | CREATE_DR |
    |                  |   Else:                           |           |
    |                  |           Start TI                |  SEND_DR  |
    +------------------+-----------------------------------+-----------+
    |    OFRnotlast    |           Notify ACKs             |           |
    |                  |           Record Offers           |           |
    |                  |           Start TI                |  SEND_DR  |
    +------------------+-----------------------------------+-----------+
    |     OFRlast      |           Cancel TI               |           |
    |                  |           Notify ACKs             |           |
    |                  |           Record Offers           |           |
    |                  |           Select for Request      |           |
    |                  |           Send Requests           |           |
    |                  |           Start TI                |  REQUEST  |
    +------------------+-----------------------------------+-----------+
    |      ErrDC       |           Abort Exchange          |(finished) |
    +------------------+-----------------------------------+-----------+
    |     ErrBadSI     |           Abort Exchange          |(finished) |
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |           Abort Exchange          | CREATE_DR |
    +==================================================================+


















Lindgren, et al.              Experimental                     [Page 84]

RFC 6693                         PRoPHET                     August 2012


    State: REQUEST

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |  Timeout(info)   |  Check Receiving                  |           |
    |                  |  If bundle reception in progress: |           |
    |                  |         Start TI                  |  REQUEST  |
    |                  |  Otherwise:                       |           |
    |                  |         Send Requests             |           |
    |                  |         Start TI (note 3)         |  REQUEST  |
    +------------------+-----------------------------------+-----------+
    | NotLastBndlRcvd  |     Record Bundle Received        |           |
    |                  |     Start TI                      |  REQUEST  |
    +------------------+-----------------------------------+-----------+
    |   LastBndlRcvd   |     Cancel TI                     |           |
    |                  |     All Requests Done             |           |
    |                  |     Start NE                      |  REQUEST  |
    +------------------+-----------------------------------+-----------+
    |Timeout(next_exch)|                                   | CREATE_DR |
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |     Abort Exchange                | CREATE_DR |
    +==================================================================+

   Note 1:
      No response to the RIB has been received before the timer expired,
      so we re-send the dictionary and RIB TLVs.  If the timeout occurs
      repeatedly, it is likely that communication has failed and the
      connection MUST be terminated.

   Note 2:
      If a Dictionary Conflict error has to be sent, the state machine
      will be aborted.  If this event occurs repeatedly, it is likely
      that there is either a serious software problem or a security
      issue.  The connection MUST be terminated.

   Note 3:
      Remaining requested bundles have not arrived before the timer
      expired, so we re-send the list of outstanding requests.  If the
      timeout occurs repeatedly, it is likely that communication has
      failed and the connection MUST be terminated.










Lindgren, et al.              Experimental                     [Page 85]

RFC 6693                         PRoPHET                     August 2012


5.3.4.3.  State Tables for the Listener Role



   The rules and state tables for the Listener role use the following
   operations:

   o  The "Clear Supplied RIBs" operation is defined as setting up an
      empty container to hold the set of RIBs supplied by the peer node.

   o  The "Record RIBs Supplied" operation is defined as:

      1.  Taking the RIB entries from a received RIB TLV.

      2.  Verifying that the String ID used in each entry is present in
          the dictionary.  If not, an Error TLV containing the offending
          String ID is sent to the peer, and the Initiator and Listener
          processes are aborted and restarted as if the ESTAB state had
          just been reached.

      3.  If all the String IDs are present in the dictionary, record
          the delivery predictabilities for each EID in the entries.

   o  The "Recalc Dlvy Predictabilities" operation uses the algorithms
      defined in Section 2.1.2 to update the local set of delivery
      predictabilities using the using the set of delivery
      predictabilities supplied by the peer in RIB TLVs.

   o  The "Determine Offers" operation determines the set of bundles to
      be offered to the peer.  The local delivery predictabilities and
      the delivery predictabilities supplied by the peer are compared,
      and a prioritized choice of the bundles stored in this node to be
      offered to the peer is made according to the configured queueing
      policy and forwarding strategy.

   o  The "Determine ACKs" operation is defined as obtaining the set of
      PRoPHET ACKs recorded by the bundle protocol agent that need to be
      forwarded to the peer.  The list of PRoPHET ACKs is maintained
      internally by the PRoPHET protocol implementation rather than the
      main bundle protocol agent (see Section 3.5).

   o  The "Determine Offer Dict Updates" operation is defined as
      determining any extra EIDs that are not already in the dictionary,
      recording the previous state of the local dictionary, and then
      adding the required extra entries to the dictionary.








Lindgren, et al.              Experimental                     [Page 86]

RFC 6693                         PRoPHET                     August 2012


   o  The "Send Offers" operation is defined as formatting one or more
      non-empty Bundle Offer TLVs, incorporating the sets of Offers and
      PRoPHET ACKs previously determined, and sending them to the peer
      node.  If more than one Bundle Offer TLV is sent, all but the last
      one MUST have the "More Offer/Response TLVs Following" flag set to
      1; the last or only one MUST have the flag set to 0.

   o  The "Record Requests" operation is defined as recording all the
      bundles offered in a Bundle Offer TLV in the list of bundles
      offers.  Duplicates MUST be ignored.  The order of requests in the
      TLVs MUST be maintained so far as is possible (it is possible that
      a bundle has to be re-sent, and this may result in out-of-order
      delivery).

   o  The "Send Bundles" operation is defined as sending, in the order
      requested, the bundles in the requested list.  This requires the
      list to be communicated to the bundle protocol agent (see
      Section 2.2).

   o  The "Check Initiator Start Point" operation is defined as checking
      the configured sequential operation policy to determine if the
      Listener role has reached the point where the Initiator role
      should be started.  If so, the InitStart notification is sent to
      the Initiator role in the same node.

   The following events are specific to the Listener role state machine:

   RIBnotlast    RIB TLV received with "More RIB TLVs" flag set to 1.

   RIBlast       RIB TLV received with "More RIB TLVs" flag set to 0 and
                 a non-zero count of RIB Entries.

   REQnotlast    Bundle Response TLV received with More Offer/Response
                 TLVs Following flag set to 1.

   REQlast       Bundle Response TLV received with More Offer/Response
                 TLVs Following flag set to 0 and a non-zero count of
                 bundle offers.

   REQempty      Bundle Response TLV received with More Offer/Response
                 TLVs Following flag set to 0 and a zero count of bundle
                 offers.









Lindgren, et al.              Experimental                     [Page 87]

RFC 6693                         PRoPHET                     August 2012


    State: WAIT_DICT

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |     On Entry     |     Check Initiator Start Point   | WAIT_DICT |
    +------------------+-----------------------------------+-----------+
    |       RIBDi      |     Update Dictionary (note 1)    |           |
    |                  |     If Dictionary Conflict found: |           |
    |                  |           Abort Exchange          |(finished) |
    |                  |     Else:                         |           |
    |                  |           Start TP                | WAIT_RIB  |
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |     Abort Exchange                | WAIT_DICT |
    +==================================================================+




































Lindgren, et al.              Experimental                     [Page 88]

RFC 6693                         PRoPHET                     August 2012


    State: WAIT_RIB

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |     On Entry     |   Clear Supplied RIBS             | WAIT_RIB  |
    +------------------+-----------------------------------+-----------+
    |       RIBDi      |   Update Dictionary (note 1)      |           |
    |                  |   If Dictionary Conflict found:   |           |
    |                  |         Abort Exchange            |(finished) |
    |                  |   Else:                           |           |
    |                  |         Start TP                  | WAIT_RIB  |
    +------------------+-----------------------------------+-----------+
    |    RIBnotlast    |   Record RIBS Supplied (note 2)   |           |
    |                  |   If EID missing in dictionary:   |           |
    |                  |         Abort Exchange            |(finished) |
    |                  |   Else:                           |           |
    |                  |         Start TP                  | WAIT_RIB  |
    +------------------+-----------------------------------+-----------
    |     RIBlast      |   Check Initiator Start Point     |           |
    |                  |   Record RIBS Supplied (note 2)   |           |
    |                  |   If EID missing in dictionary:   |           |
    |                  |         Abort Exchange            |(finished) |
    |                  |   Otherwise                       |           |
    |                  |         Recalc Dlvy               |           |
    |                  |               Predictabilities    |           |
    |                  |         Cancel TP                 |           |
    |                  |         Determine Offers          |           |
    |                  |         Determine ACKs            |           |
    |                  |         Determine Offer           |           |
    |                  |               Dict Updates        |           |
    |                  |         Send RIB Dictionary       |           |
    |                  |               Updates             |           |
    |                  |         Send Offers               |           |
    |                  |         Start TI                  |   OFFER   |
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |     Abort Exchange                | WAIT_DICT |
    +------------------+-----------------------------------+-----------+
    |Any Other TLV rcvd|     Abort Exchange                |(finished) |
    +------------------+-----------------------------------+-----------+
    |  Timeout(peer)   |     Send RIB Dictionary Updates   |           |
    |                  |     Send Offers                   |           |
    |                  |     Start TI (note 3)             |   OFFER   |
    +==================================================================+







Lindgren, et al.              Experimental                     [Page 89]

RFC 6693                         PRoPHET                     August 2012


    State: OFFER

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |    REQnotlast    |      Send Bundles                 |           |
    |                  |      Start TI                     |   OFFER   |
    +------------------+-----------------------------------+-----------+
    |     REQlast      |      Cancel TI                    |           |
    |                  |      Check Initiator Start Point  |           |
    |                  |      Send Bundles                 | SND_BUNDLE|
    +------------------+-----------------------------------+-----------+
    |     REQempty     |      Cancel TI                    |           |
    |                  |      Check Initiator Start Point  | WAIT_MORE|
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |      Abort Exchange               | WAIT_DICT |
    +------------------+-----------------------------------+-----------+
    |  Timeout(info)   |      Send RIB Dictionary Updates  |           |
    |                  |      Send Offers                  |           |
    |                  |      Start TI (note 3)            |   OFFER   |
    +==================================================================+



    State: SND_BUNDLE

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    |    REQnotlast    |      Send Bundles                 |           |
    |                  |      Start TI                     | SND_BUNDLE|
    +------------------+-----------------------------------+-----------+
    |     REQlast      |      Cancel TI                    |           |
    |                  |      Send Bundles                 | SND_BUNDLE|
    +------------------+-----------------------------------+-----------+
    |     REQempty     |      Cancel TI                    |           |
    |                  |      Check Initiator Start Point  | WAIT_MORE|
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |      Abort Exchange               | WAIT_DICT |
    +------------------+-----------------------------------+-----------+
    |  Timeout(info)   |      Send RIB Dictionary Updates  |           |
    |                  |      Send Offers                  |           |
    |                  |      Start TI (note 3)            |   OFFER   |
    +==================================================================+







Lindgren, et al.              Experimental                     [Page 90]

RFC 6693                         PRoPHET                     August 2012


    State: WAIT_MORE

    +==================================================================+
    |     Condition    |               Action              | New State |
    +==================+===================================+===========+
    | More Bundles     |         Determine Offers          |           |
    |                  |         Determine ACKs            |           |
    |                  |         Determine Offer           |           |
    |                  |               Dict Updates        |           |
    |                  |         Send RIB Dictionary       |           |
    |                  |               Updates             |           |
    |                  |         Send Offers               |           |
    |                  |         Start TI                  |   OFFER   |
    +------------------+-----------------------------------+-----------+
    |       RIBDi      |   Update Dictionary (note 1)      |           |
    |                  |   If Dictionary Conflict found:   |           |
    |                  |         Abort Exchange            |(finished) |
    |                  |   Else:                           |           |
    |                  |         Start TP                  | WAIT_RIB  |
    +------------------+-----------------------------------+-----------+
    |    REQnotlast    |      Send Bundles                 |           |
    |                  |      Start TI                     | SND_BUNDLE|
    +------------------+-----------------------------------+-----------+
    |     REQlast      |      Cancel TI                    |           |
    |                  |      Send Bundles                 | SND_BUNDLE|
    +------------------+-----------------------------------+-----------+
    |     REQempty     |      Cancel TI                    |           |
    |                  |      Check Initiator Start Point  | SND_BUNDLE|
    +------------------+-----------------------------------+-----------+
    |     HelloAck     |      Abort Exchange               | WAIT_DICT |
    +------------------+-----------------------------------+-----------+
    |  Timeout(info)   |      Send RIB Dictionary Updates  |           |
    |                  |      Send Offers                  |           |
    |                  |      Start TI (note 3)            |   OFFER   |
    +==================================================================+

   Note 1:
      Both the dictionary and the RIB TLVs may come in the same PRoPHET
      message.  In that case, the state will change to WAIT_RIB, and the
      RIB will then immediately be processed.

   Note 2:
      Send an ACK if the timer for the peering node expires.  Either the
      link has been broken, and then the link setup will restart, or it
      will trigger the Information Exchange Phase to restart.






Lindgren, et al.              Experimental                     [Page 91]

RFC 6693                         PRoPHET                     August 2012


   Note 3:
      When the RIB is received, it is possible for the PRoPHET agent to
      update its delivery predictabilities according to Section 2.1.2.
      The delivery predictabilities and the RIB is then used together
      with the forwarding strategy in use to create a bundle offer TLV.
      This is sent to the peering node.

   Note 4:
      No more bundles are requested by the other node; transfer is
      complete.

   Note 5:
      No response to the bundle offer has been received before the timer
      expired, so we re-send the bundle offer.

5.4.  Interaction with Nodes Using Version 1 of PRoPHET



   There are existing implementations of PRoPHET based on draft versions
   of this specification that use version 1 of the protocol.  There are
   a number of significant areas of difference between version 1 and
   version 2 as described in this document:

   o  In version 1, the delivery predictability update equations were
      significantly different, and in the case of the transitivity
      equation (Equation 3) could lead to degraded performance or non-
      delivery of bundles in some circumstances.

   o  In the current version , constraints were placed on the String IDs
      generated by each node to ensure that it was not possible for
      there to be a conflict if the IDs were generated concurrently and
      independently in the two nodes.

   o  In the current version, a flag has been added to the Routing
      Information Base Dictionary TLV to distinguish dictionary updates
      sent by the Initiator role and by the Listener role.

   o  In the current version, the Bundle Offer and Response TLVs have
      been significantly revised.  The version 2 TLVs have been
      allocated new TLV Type numbers, and the version 1 TLVs (types 0xA2
      and 0xA3) are now deprecated.  For each bundle specifier, the
      source EID is transmitted in addition to the creation timestamp by
      version 2 to ensure that the bundle is uniquely identified.
      Version 2 also transmits the fragment payload offset and length
      when the offered bundle is a bundle fragment.  The payload length
      can optionally be transmitted for each bundle (whether or not it
      is a fragment) to give the receiver additional information that
      can be useful when determining which bundle offers to accept.




Lindgren, et al.              Experimental                     [Page 92]

RFC 6693                         PRoPHET                     August 2012


   o  The behavior of the system after the first Information Exchange
      Phase has been better defined.  The state machine has been altered
      to better describe how the ongoing operations work.  This has
      involved the removal of the high-level state WAIT_INFO and the
      addition of two states in the Listener role subsidiary state
      machine (SND_BUNDLE and WAIT_MORE).  The protocol on the wire has
      not been altered by this change to the description of the state
      machine.  However, the specification of the later stages of
      operation was slightly vague and might have been interpreted
      differently by various implementers.

   A node implementing version 2 of the PRoPHET protocol as defined in
   this document MAY ignore a communication opportunity with a node that
   sends a HELLO message indicating that it uses version 1, or it MAY
   partially downgrade and respond to messages as if it were a version 1
   node.  This means that the version field in all message headers MUST
   contain 1.

   It is RECOMMENDED that the version 2 node use the metric update
   equations defined in this document even when communicating with a
   version 1 node as this will partially inhibit the problems with the
   transitivity equation in version 1, and that the version 2 node
   modify any received metrics that are greater than (1 - delta) to be
   (1 - delta) to avoid becoming a "sink" for bundles that are not
   destined for this node.  Also version 1 nodes cannot be explicitly
   offered bundle fragments, and an exchange with a node supporting
   version 1 MUST use the, now deprecated, previous versions of the
   Bundle Offer and Response TLVs.

   Generally, nodes using version 1 should be upgraded if at all
   possible because of problems that have been identified.

6.  Security Considerations



   Currently, PRoPHET does not specify any special security measures.
   As a routing protocol for intermittently connected networks, PRoPHET
   is a target for various attacks.  The various known possible
   vulnerabilities are discussed in this section.

   The attacks described here are not problematic if all nodes in the
   network can be trusted and are working towards a common goal.  If
   there exist such a set of nodes, but there also exist malicious
   nodes, these security problems can be solved by introducing an
   authentication mechanism when two nodes meet, for example, using a
   public key system.  Thus, only nodes that are known to be members of
   the trusted group of nodes are allowed to participate in the routing.
   This of course introduces the additional problem of key distribution,
   but that is not addressed here.



Lindgren, et al.              Experimental                     [Page 93]

RFC 6693                         PRoPHET                     August 2012


   Where suitable, the mechanisms (such as key management and bundle
   authentication or integrity checks) and terminology specified by the
   Bundle Security Protocol [RFC6257] are to be used.

6.1.  Attacks on the Operation of the Protocol



   There are a number of kinds of attacks on the operation of the
   protocol that it would be possible to stage on a PRoPHET network.
   The attacks and possible remedies are listed here.

6.1.1.  Black-Hole Attack



   A malicious node sets its delivery predictabilities for all
   destinations to a value close to or exactly equal to 1 and/or
   requests all bundles from nodes it meets, and does not forward any
   bundles.  This has two effects, both causing messages to be drawn
   towards the black hole instead of to their correct destinations.

   1.  A node encountering a malicious node will try to forward all its
       bundles to the malicious node, creating the belief that the
       bundle has been very favorably forwarded.  Depending on the
       forwarding strategy and queueing policy in use, this might hamper
       future forwarding of the bundle and/or lead to premature dropping
       of the bundle.

   2.  Due to the transitivity, the delivery predictabilities reported
       by the malicious node will affect the delivery predictabilities
       of other nodes.  This will create a gradient for all destinations
       with the black hole as the "center of gravity" towards which all
       bundles traverse.  This should be particularly severe in
       connected parts of the network.

6.1.1.1.  Attack Detection



   A node receiving a set of delivery predictabilities that are all at
   or close to 1 should be suspicious.  Similarly, a node that accepts
   all bundles and offers none might be considered suspicious.  However,
   these conditions are not impossible in normal operation.

6.1.1.2.  Attack Prevention/Solution



   To prevent this attack, authentication between nodes that meet needs
   to be present.  Nodes can also inspect the received metrics and
   bundle acceptances/offers for suspicious patterns and terminate
   communications with nodes that appear suspicious.  The natural
   evolution of delivery predictabilities should mean that a genuine
   node would not be permanently ostracized even if the values lead to




Lindgren, et al.              Experimental                     [Page 94]

RFC 6693                         PRoPHET                     August 2012


   termination of a communication opportunity on one occasion.  The
   epidemic nature of PRoPHET would mean that such a termination rarely
   leads to non-delivery of bundles.

6.1.2.  Limited Black-Hole Attack / Identity Spoofing



   A malicious node misrepresents itself by claiming to be someone else.
   The effects of this attack are:

   1.  The effects of the black-hole attack listed above hold for this
       attack as well, with the exception that only the delivery
       predictabilities and bundles for one particular destination are
       affected.  This could be used to "steal" the data that should be
       going to a particular node.

   2.  In addition to the above problems, PRoPHET ACKs will be issued
       for the bundles that are delivered to the malicious node.  This
       will cause these bundles to be removed from the network, reducing
       the chance that they will reach their real destination.

6.1.2.1.  Attack Detection



   The destination can detect that this kind of attack has occurred (but
   it cannot prevent the attack) when it receives a PRoPHET ACK for a
   bundle destined to itself but for which it did not receive the
   corresponding bundle.

6.1.2.2.  Attack Prevention/Solution



   To prevent this attack, authentication between nodes that meet needs
   to be present.

6.1.3.  Fake PRoPHET ACKs



   A malicious node may issue fake PRoPHET ACKs for all bundles (or only
   bundles for a certain destination if the attack is targeted at a
   single node) carried by nodes it met.  The affected bundles will be
   deleted from the network, greatly reducing their probability of being
   delivered to the destination.

6.1.3.1.  Attack Prevention/Solution



   If a public key cryptography system is in place, this attack can be
   prevented by mandating that all PRoPHET ACKs be signed by the
   destination.  Similarly to other solutions using public key
   cryptography, this introduces the problem of key distribution.





Lindgren, et al.              Experimental                     [Page 95]

RFC 6693                         PRoPHET                     August 2012


6.1.4.  Bundle Store Overflow



   After encountering and receiving the delivery predictability
   information from the victim, a malicious node may generate a large
   number of fake bundles for the destination for which the victim has
   the highest delivery predictability.  This will cause the victim to
   most likely accept these bundles, filling up its bundle storage,
   possibly at the expense of other, legitimate, bundles.  This problem
   is transient as the messages will be removed when the victim meets
   the destination and delivers the messages.

6.1.4.1.  Attack Detection



   If it is possible for the destination to figure out that the bundles
   it is receiving are fake, it could report that malicious actions are
   underway.

6.1.4.2.  Attack Prevention/Solution



   This attack could be prevented by requiring sending nodes to sign all
   bundles they send.  By doing this, intermediate nodes could verify
   the integrity of the messages before accepting them for forwarding.

6.1.5.  Bundle Store Overflow with Delivery Predictability Manipulation



   A more sophisticated version of the attack in the previous section
   can be attempted.  The effect of the previous attack was lessened
   since the destination node of the fake bundles existed.  This caused
   fake bundles to be purged from the network when the destination was
   encountered.  The malicious node may now use the transitive property
   of the protocol to boost the victim's delivery predictabilities for a
   non-existent destination.  After this, it creates a large number of
   fake bundles for this non-existent destination and offers them to the
   victim.  As before, these bundles will fill up the bundle storage of
   the victim.  The impact of this attack will be greater as there is no
   probability of the destination being encountered and the bundles
   being acknowledged.  Thus, they will remain in the bundle storage
   until they time out (the malicious node may set the timeout to a
   large value) or until they are evicted by the queueing policy.

   The delivery predictability for the fake destination may spread in
   the network due to the transitivity, but this is not a problem, as it
   will eventually age and fade away.

   The impact of this attack could be increased if multiple malicious
   nodes collude, as network resources can be consumed at a greater
   speed and at many different places in the network simultaneously.




Lindgren, et al.              Experimental                     [Page 96]

RFC 6693                         PRoPHET                     August 2012


6.2.  Interactions with External Routing Domains



   Users may opt to connect two regions of sparsely connected nodes
   through a connected network such as the Internet where another
   routing protocol is running.  To this network, PRoPHET traffic would
   look like any other application-layer data.  Extra care must be taken
   in setting up these gateway nodes and their interconnections to make
   sure that malicious nodes cannot use them to launch attacks on the
   infrastructure of the connected network.  In particular, the traffic
   generated should not be significantly more than what a single regular
   user end host could create on the network.

7.  IANA Considerations



   Following the policies outlined in "Guidelines for Writing an IANA
   Considerations Section in RFCs" (RFC 5226 [RFC5226]), the following
   name spaces are defined in PRoPHET.

   o  For fields in the PRoPHET message header (Section 4.1):

      *  DTN Routing Protocol Number

      *  PRoPHET Protocol Version

      *  PRoPHET Header Flags

      *  PRoPHET Result Field

      *  PRoPHET Codes for Success and Codes for Failure

   o  Identifiers for TLVs carried in PRoPHET messages:

      *  PRoPHET TLV Type (Section 4.2)

   o  Definitions of TLV Flags and other flag fields in TLVs:

      *  Hello TLV Flags (Section 4.3.1)

      *  Error TLV Flags (Section 4.3.2)

      *  Routing Information Base (RIB) Dictionary TLV Flags
         (Section 4.3.3)

      *  Routing Information Base (RIB) TLV Flags (Section 4.3.4)

      *  Routing Information Base (RIB) Flags per entry (Section 4.3.4)

      *  Bundle Offer and Response TLV Flags (Section 4.3.5)



Lindgren, et al.              Experimental                     [Page 97]

RFC 6693                         PRoPHET                     August 2012


      *  Bundle Offer and Response B Flags per offer or response
         (Section 4.3.5)

   The following subsections list the registries that have been created.
   Initial values for the registries are given below; future assignments
   for unassigned values are to be made through the Specification
   Required policy.  Where specific values are defined in the IANA
   registries according to the specifications in the subsections below,
   the registry refers to this document as defining the allocation.

7.1.  DTN Routing Protocol Number



   The encoding of the Protocol Number field in the PRoPHET header
   (Section 4.1) is:

         +--------------------------+-----------+---------------+
         |         Protocol         |   Value   |   Reference   |
         +--------------------------+-----------+---------------+
         |     PRoPHET Protocol     |    0x00   | This document |
         |        Unassigned        | 0x01-0xEF |               |
         | Private/Experimental Use | 0xF0-0xFF | This document |
         +--------------------------+-----------+---------------+

7.2.  PRoPHET Protocol Version



   The encoding of the PRoPHET Version field in the PRoPHET header
   (Section 4.1) is:

        +----------------------------+-----------+---------------+
        |           Version          |   Value   |   Reference   |
        +----------------------------+-----------+---------------+
        | Reserved (do not allocate) |    0x00   | This document |
        |         PRoPHET v1         |    0x01   | This document |
        |         PRoPHET v2         |    0x02   | This document |
        |         Unassigned         | 0x03-0xEF |               |
        |  Private/Experimental Use  | 0xF0-0xFE | This document |
        |          Reserved          |    0xFF   |               |
        +----------------------------+-----------+---------------+













Lindgren, et al.              Experimental                     [Page 98]

RFC 6693                         PRoPHET                     August 2012


7.3.  PRoPHET Header Flags



   The following Flags are defined for the PRoPHET Header (Section 4.1):

                 +------------+--------------+-----------+
                 |   Meaning  | Bit Position | Reference |
                 +------------+--------------+-----------+
                 | Unassigned |     Bit 0    |           |
                 | Unassigned |     Bit 1    |           |
                 | Unassigned |     Bit 2    |           |
                 | Unassigned |     Bit 3    |           |
                 +------------+--------------+-----------+

7.4.  PRoPHET Result Field



   The encoding of the Result field in the PRoPHET header (Section 4.1)
   is:

        +--------------------------+-------------+---------------+
        |       Result Value       |    Value    |   Reference   |
        +--------------------------+-------------+---------------+
        |         Reserved         |     0x00    | This document |
        |       NoSuccessAck       |     0x01    | This document |
        |          AckAll          |     0x02    | This document |
        |          Success         |     0x03    | This document |
        |          Failure         |     0x04    | This document |
        |       ReturnReceipt      |     0x05    | This document |
        |        Unassigned        | 0x06 - 0x7F |               |
        | Private/Experimental Use | 0x80 - 0xFF | This document |
        +--------------------------+-------------+---------------+

7.5.  PRoPHET Codes for Success and Codes for Failure



   The encoding for Code field in the PRoPHET header (Section 4.1) for
   "Success" messages is:

        +--------------------------+-------------+---------------+
        |         Code Name        |    Values   |   Reference   |
        +--------------------------+-------------+---------------+
        |      Generic Success     |     0x00    | This document |
        |    Submessage Received   |     0x01    | This document |
        |        Unassigned        | 0x02 - 0x7F |               |
        | Private/Experimental Use | 0x80 - 0xFF | This document |
        +--------------------------+-------------+---------------+







Lindgren, et al.              Experimental                     [Page 99]

RFC 6693                         PRoPHET                     August 2012


   The encoding for Code in the PRoPHET header (Section 4.1) for
   "Failure" messages is:

       +----------------------------+-------------+---------------+
       |          Code Name         |    Values   |   Reference   |
       +----------------------------+-------------+---------------+
       | Reserved (do not allocate) | 0x00 - 0x01 | This document |
       |     Unspecified Failure    |     0x02    | This document |
       |         Unassigned         | 0x03 - 0x7F |               |
       |  Private/Experimental Use  | 0x80 - 0xFE | This document |
       |    Error TLV in Message    |     0xFF    | This document |
       +----------------------------+-------------+---------------+

7.6.  PRoPHET TLV Type



   The TLV Types defined for PRoPHET (Section 4.2) are:

      +------------------------------+-------------+---------------+
      |             Type             |    Value    |   Reference   |
      +------------------------------+-------------+---------------+
      |  Reserved (do not allocate)  |     0x00    | This document |
      |           Hello TLV          |     0x01    | This document |
      |           Error TLV          |     0x02    | This document |
      |          Unsassigned         | 0x03 - 0x9F |               |
      |      RIB dictionary TLV      |     0xA0    | This document |
      |            RIB TLV           |     0xA1    | This document |
      |   Bundle Offer (deprecated)  |     0xA2    | This document |
      | Bundle Response (deprecated) |     0xA3    | This document |
      |       Bundle Offer (v2)      |     0xA4    | This document |
      |     Bundle Response (v2)     |     0xA5    | This document |
      |          Unassigned          | 0xA6 - 0xCF |               |
      |   Private/Experimental Use   | 0xD0 - 0xFF | This document |
      +------------------------------+-------------+---------------+


















Lindgren, et al.              Experimental                    [Page 100]

RFC 6693                         PRoPHET                     August 2012


7.7.  Hello TLV Flags



   The following TLV Flags are defined for the Hello TLV
   (Section 4.3.1).  Flag numbers 0, 1, and 2 are treated as a 3-bit
   unsigned integer with 5 of the 8 possible values allocated, and the
   other 3 reserved.  The remaining bits are treated individually:

   +----------------------------+---------------------+---------------+
   |           Meaning          |        Value        |   Reference   |
   +----------------------------+---------------------+---------------+
   |                            | (Flags 0, 1, and 2) |               |
   | Reserved (do not allocate) |        0b000        | This document |
   |             SYN            |        0b001        | This document |
   |           SYNACK           |        0b010        | This document |
   |             ACK            |        0b011        | This document |
   |           RSTACK           |        0b100        | This document |
   |         Unassigned         |    0b101 - 0b111    |               |
   |                            |    (Flags 3 - 7)    |               |
   |         Unassigned         |        Flag 3       |               |
   |         Unassigned         |        Flag 4       |               |
   |         Unassigned         |        Flag 5       |               |
   |         Unassigned         |        Flag 6       |               |
   |           L Flag           |        Flag 7       | This document |
   +----------------------------+---------------------+---------------+

7.8.  Error TLV Flags



   The TLV Flags field in the Error TLV (Section 4.3.2) is treated as an
   unsigned 8-bit integer encoding the Error TLV number.  The following
   values are defined:

      +--------------------------+------------------+---------------+
      |      Error TLV Name      | Error TLV Number |   Reference   |
      +--------------------------+------------------+---------------+
      |    Dictionary Conflict   |       0x00       | This document |
      |       Bad String ID      |       0x01       | This document |
      |        Unassigned        |    0x02 - 0x7F   |               |
      | Private/Experimental Use |    0x80 - 0xFF   | This document |
      +--------------------------+------------------+---------------+












Lindgren, et al.              Experimental                    [Page 101]

RFC 6693                         PRoPHET                     August 2012


7.9.  RIB Dictionary TLV Flags



   The following TLV Flags are defined for the RIB Base Dictionary TLV
   (Section 4.3.3):

       +----------------------------+--------------+---------------+
       |           Meaning          | Bit Position |   Reference   |
       +----------------------------+--------------+---------------+
       |      Sent by Listener      |    Flag 0    | This document |
       | Reserved (do not allocate) |    Flag 1    | This document |
       | Reserved (do not allocate) |    Flag 2    | This document |
       |         Unassigned         |    Flag 3    |               |
       |         Unassigned         |    Flag 4    |               |
       |         Unassigned         |    Flag 5    |               |
       |         Unassigned         |    Flag 6    |               |
       |         Unassigned         |    Flag 7    |               |
       +----------------------------+--------------+---------------+

7.10.  RIB TLV Flags



   The following TLV Flags are defined for the RIB TLV (Section 4.3.4):

       +----------------------------+--------------+---------------+
       |           Meaning          | Bit Position |   Reference   |
       +----------------------------+--------------+---------------+
       |        More RIB TLVs       |    Flag 0    | This document |
       | Reserved (do not allocate) |    Flag 1    | This document |
       | Reserved (do not allocate) |    Flag 2    | This document |
       |         Unassigned         |    Flag 3    |               |
       |         Unassigned         |    Flag 4    |               |
       |         Unassigned         |    Flag 5    |               |
       |         Unassigned         |    Flag 6    |               |
       |         Unassigned         |    Flag 7    |               |
       +----------------------------+--------------+---------------+

















Lindgren, et al.              Experimental                    [Page 102]

RFC 6693                         PRoPHET                     August 2012


7.11.  RIB Flags



   The following RIB Flags are defined for the individual entries in the
   RIB TLV (Section 4.3.4):

                 +------------+--------------+-----------+
                 |   Meaning  | Bit Position | Reference |
                 +------------+--------------+-----------+
                 | Unassigned |    Flag 0    |           |
                 | Unassigned |    Flag 1    |           |
                 | Unassigned |    Flag 2    |           |
                 | Unassigned |    Flag 3    |           |
                 | Unassigned |    Flag 4    |           |
                 | Unassigned |    Flag 5    |           |
                 | Unassigned |    Flag 6    |           |
                 | Unassigned |    Flag 7    |           |
                 +------------+--------------+-----------+

7.12.  Bundle Offer and Response TLV Flags



   The following TLV Flags are defined for the Bundle Offer and Response
   TLV (Section 4.3.5):

   +------------------------------------+--------------+---------------+
   |               Meaning              | Bit Position |   Reference   |
   +------------------------------------+--------------+---------------+
   | More Offer/Response TLVs Following |    Flag 0    | This document |
   |             Unassigned             |    Flag 1    |               |
   |             Unassigned             |    Flag 2    |               |
   |             Unassigned             |    Flag 3    |               |
   |             Unassigned             |    Flag 4    |               |
   |             Unassigned             |    Flag 5    |               |
   |             Unassigned             |    Flag 6    |               |
   |             Unassigned             |    Flag 7    |               |
   +------------------------------------+--------------+---------------+
















Lindgren, et al.              Experimental                    [Page 103]

RFC 6693                         PRoPHET                     August 2012


7.13.  Bundle Offer and Response B Flags



   The following B Flags are defined for each Bundle Offer in the Bundle
   Offer and Response TLV (Section 4.3.5):

   +------------------------------------+--------------+---------------+
   |               Meaning              | Bit Position |   Reference   |
   +------------------------------------+--------------+---------------+
   |           Bundle Accepted          |    Flag 0    | This document |
   |        Bundle is a Fragment        |    Flag 1    | This document |
   |  Bundle Payload Length Included in |    Flag 2    | This document |
   |                 TLV                |              |               |
   |             Unassigned             |    Flag 3    |               |
   |             Unassigned             |    Flag 4    |               |
   |             Unassigned             |    Flag 5    |               |
   |             Unassigned             |    Flag 6    |               |
   |             PRoPHET ACK            |    Flag 7    | This document |
   +------------------------------------+--------------+---------------+

8.  Implementation Experience



   Multiple independent implementations of the PRoPHET protocol exist.

   The first implementation is written in Java, and has been optimized
   to run on the Lego MindStorms platform that has very limited
   resources.  Due to the resource constraints, some parts of the
   protocol have been simplified or omitted, but the implementation
   contains all the important mechanisms to ensure proper protocol
   operation.  The implementation is also highly modular and can be run
   on another system with only minor modifications (it has currently
   been shown to run on the Lego MindStorms platform and on regular
   laptops).

   Another implementation is written in C++ and runs in the OmNet++
   simulator to enable testing and evaluation of the protocol and new
   features.  Experience and feedback from the implementers on early
   versions of the protocol have been incorporated into the current
   version.

   An implementation compliant to an Internet-Draft (which was posted in
   2006 and eventually evolved into this RFC) has been written at Baylor
   University.  This implementation has been integrated into the DTN2
   reference implementation.

   An implementation of the protocol in C++ was developed by one of the
   authors (Samo Grasic) at Lulea University of Technology (LTU) as part
   of the Saami Networking Connectivity project (see Section 9) and
   continues to track the development of the protocol.  This work is now



Lindgren, et al.              Experimental                    [Page 104]

RFC 6693                         PRoPHET                     August 2012


   part of the Networking for Communications Challenged Communities
   (N4C) project and is used in N4C testbeds.

9.  Deployment Experience



   During a week in August 2006, a proof-of-concept deployment of a DTN
   system, using the LTU PRoPHET implementation for routing was made in
   the Swedish mountains -- the target area for the Saami Network
   Connectivity project [ccnc07] [doria_02].  Four fixed camps with
   application gateways, one Internet gateway, and seven mobile relays
   were deployed.  The deployment showed PRoPHET to be able to route
   bundles generated by different applications such as email and web
   caching.

   Within the realms of the SNC and N4C projects, multiple other
   deployments, both during summer and winter conditions, have been done
   at various scales during 2007-2010 [winsdr08].

   An implementation has been made for Android-based mobile telephones
   in the Bytewalla project [bytewalla].

10.  Acknowledgements



   The authors would like to thank Olov Schelen and Kaustubh S. Phanse
   for contributing valuable feedback regarding various aspects of the
   protocol.  We would also like to thank all other reviewers and the
   DTNRG chairs for the feedback in the process of developing the
   protocol.  The Hello TLV mechanism is loosely based on the Adjacency
   message developed for RFC 3292.  Luka Birsa and Jeff Wilson have
   provided us with feedback from doing implementations of the protocol
   based on various preliminary versions of the document.  Their
   feedback has helped us make the document easier to read for an
   implementer and has improved the protocol.

11.  References



11.1.  Normative References



   [RFC2119]      Bradner, S., "Key words for use in RFCs to Indicate
                  Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC5050]      Scott, K. and S. Burleigh, "Bundle Protocol
                  Specification", RFC 5050, November 2007.








Lindgren, et al.              Experimental                    [Page 105]

RFC 6693                         PRoPHET                     August 2012


11.2.  Informative References



   [CLAYER]       Demmer, M., Ott, J., and S. Perreault, "Delay Tolerant
                  Networking TCP Convergence Layer Protocol", Work
                  in Progress, August 2012.

   [RFC1058]      Hedrick, C., "Routing Information Protocol", RFC 1058,
                  June 1988.

   [RFC4838]      Cerf, V., Burleigh, S., Hooke, A., Torgerson, L.,
                  Durst, R., Scott, K., Fall, K., and H. Weiss, "Delay-
                  Tolerant Networking Architecture", RFC 4838,
                  April 2007.

   [RFC5226]      Narten, T. and H. Alvestrand, "Guidelines for Writing
                  an IANA Considerations Section in RFCs", BCP 26,
                  RFC 5226, May 2008.

   [RFC6257]      Symington, S., Farrell, S., Weiss, H., and P. Lovell,
                  "Bundle Security Protocol Specification", RFC 6257,
                  May 2011.

   [bytewalla]    Prasad, M., "Bytewalla 3: Network architecture and
                  PRoPHET implementation", Bytewalla Project, KTH Royal
                  Institute of Technology, Stockholm, Sweden, October
                   2010,
                  <http://www.bytewalla.org/sites/bytewalla.org/files/
                  Bytewalla3_Network_architecture_and_PRoPHET_v1.0.pdf>.

   [ccnc07]       Lindgren, A. and A. Doria, "Experiences from Deploying
                  a Real-life DTN System", Proceedings of the 4th Annual
                  IEEE Consumer Communications and Networking Conference
                  (CCNC 2007), Las Vegas, Nevada, USA, January 2007.

   [doria_02]     Doria, A., Uden, M., and D. Pandey, "Providing
                  connectivity to the Saami nomadic community",
                  Proceedings of the 2nd International Conference on
                  Open Collaborative Design for Sustainable Innovation
                  (dyd 02), Bangalore, India, December 2002.

   [lindgren_06]  Lindgren, A. and K. Phanse, "Evaluation of Queueing
                  Policies and Forwarding Strategies for Routing in
                  Intermittently Connected Networks", Proceedings of
                  COMSWARE 2006, January 2006.

   [vahdat_00]    Vahdat, A. and D. Becker, "Epidemic Routing for
                  Partially Connected Ad Hoc Networks", Duke University
                  Technical Report CS-200006, April 2000.



Lindgren, et al.              Experimental                    [Page 106]

RFC 6693                         PRoPHET                     August 2012


   [winsdr08]     Lindgren, A., Doria, A., Lindblom, J., and M. Ek,
                  "Networking in the Land of Northern Lights - Two Years
                  of Experiences from DTN System Deployments",
                  Proceedings of the ACM Wireless Networks and Systems
                  for Developing Regions Workshop (WiNS-DR), San
                  Francisco, California, USA, September 2008.













































Lindgren, et al.              Experimental                    [Page 107]

RFC 6693                         PRoPHET                     August 2012


Appendix A.  PRoPHET Example



   To help grasp the concepts of PRoPHET, an example is provided to give
   an understanding of the transitive property of the delivery
   predictability and the basic operation of PRoPHET.  In Figure 13, we
   revisit the scenario where node A has a message it wants to send to
   node D.  In the bottom right corner of subfigures a-c, the delivery
   predictability tables for the nodes are shown.  Assume that nodes C
   and D encounter each other frequently (Figure 13a), making the
   delivery predictability values they have for each other high.  Now
   assume that node C also frequently encounters node B (Figure 13b).
   Nodes B and C will get high delivery predictability values for each
   other, and the transitive property will also increase the value B has
   for D to a medium level.  Finally, node B meets node A (Figure 13c),
   which has a message for node D.  Figure 13d shows the message
   exchange between node A and node B.  Summary vectors and delivery
   predictability information is exchanged, delivery predictabilities
   are updated, and node A then realizes that P_(b,d) > P_(a,d), and
   thus forwards the message for node D to node B.
































Lindgren, et al.              Experimental                    [Page 108]

RFC 6693                         PRoPHET                     August 2012


   +----------------------------+   +----------------------------+
   |                            |   |                            |
   |                  C         |   |                       D    |
   |                   D        |   |                            |
   |       B                    |   |       B C                  |
   |                            |   |                            |
   |                            |   |                            |
   |                            |   |                            |
   |                            |   |                            |
   | A*                         |   | A*                         |
   +-------------+--------------+   +-------------+--------------+
   |   A  |   B  |   C   |  D   |   |   A  |   B  |   C   |  D   |
   |B:low |A:low |A:low  |A:low |   |B:low |A:low |A:low  |A:low |
   |C:low |C:low |B:low  |B:low |   |C:low |C:high|B:high |B:low |
   |D:low |D:low |D:high |C:high|   |D:low |D:med |D:high |C:high|
   +-------------+--------------+   +-------------+--------------+
                (a)                              (b)
   +----------------------------+   A                            B
   |                            |   |                            |
   |                       D    |   |Summary vector&delivery pred|
   |                            |   |--------------------------->|
   |         C                  |   |Summary vector&delivery pred|
   |                            |   |<---------------------------|
   |                            |   |                            |
   |   B*                       |  Update delivery predictabilities
   |  A                         |   |                            |
   |                            |  Packet for D not in SV        |
   +-------------+--------------+  P(b,d)>P(a,d)                 |
   |   A  |   B  |   C   |  D   |  Thus, send                    |
   |B:low |A:low |A:low  |A:low |   |                            |
   |C:med |C:high|B:high |B:low |   |      Packet for D          |
   |D:low+|D:med |D:high |C:high|   |--------------------------->|
   +-------------+--------------+   |                            |
                (c)                              (d)

                        Figure 13: PRoPHET example















Lindgren, et al.              Experimental                    [Page 109]

RFC 6693                         PRoPHET                     August 2012


Appendix B.  Neighbor Discovery Example



   This section outlines an example of a simple neighbor discovery
   protocol that can be run in-between PRoPHET and the underlying layer
   in case lower layers do not provide methods for neighbor discovery.
   It assumes that the underlying layer supports broadcast messages as
   would be the case if a wireless infrastructure was involved.

   Each node needs to maintain a list of its active neighbors.  The
   operation of the protocol is as follows:

   1.  Every BEACON_INTERVAL milliseconds, the node does a local
       broadcast of a beacon that contains its identity and address, as
       well as the BEACON_INTERVAL value used by the node.

   2.  Upon reception of a beacon, the following can happen:

       A.  The sending node is already in the list of active neighbors.
           Update its entry in the list with the current time, and
           update the node's BEACON_INTERVAL if it has changed.

       B.  The sending node is not in the list of active neighbors.  Add
           the node to the list of active neighbors and record the
           current time and the node's BEACON_INTERVAL.  Notify the
           PRoPHET agent that a new neighbor is available ("New
           Neighbor", as described in Section 2.4).

   3.  If a beacon has not been received from a node in the list of
       active neighbors within a time period of NUM_ACCEPTED_LOSSES *
       BEACON_INTERVAL (for the BEACON_INTERVAL used by that node), it
       should be assumed that this node is no longer a neighbor.  The
       entry for this node should be removed from the list of active
       neighbors, and the PRoPHET agent should be notified that a
       neighbor has left ("Neighbor Gone", as described in Section 2.4).

Appendix C.  PRoPHET Parameter Calculation Example



   The evolution of the delivery predictabilities in a PRoPHET node is
   controlled by three main equations defined in Section 2.1.2.  These
   equations use a number of parameters that need to be appropriately
   configured to ensure that the delivery predictabilities evolve in a
   way that mirrors the mobility model that applies in the PRoPHET zone
   where the node is operating.

   When trying to describe the mobility model, it is more likely that
   the model will be couched in terms of statistical distribution of
   times between encounters and times to deliver a bundle in the zone.
   In this section, one possible way of deriving the PRoPHET parameters



Lindgren, et al.              Experimental                    [Page 110]

RFC 6693                         PRoPHET                     August 2012


   from a more usual description of the model is presented.  It should
   be remembered that this may not be the only solution, and its
   appropriateness will depend both on the overall mobility model and
   the distribution of the times involved.  There is an implicit
   assumption in this work that these distributions can be characterized
   by a normal-type distribution with a well-defined first moment
   (mean).  The exact form of the distribution is not considered here,
   but more detailed models may wish to use more specific knowledge
   about the distributions to refine the derivation of the parameters.

   To characterize the model, we consider the following parameters:

   P1  The time resolution of the model.

   P2  The average time between encounters between nodes, I_typ, where
       the identity of the nodes is not taken into account.

   P3  The average number of encounters that a node has between meeting
       a particular node and meeting the same node again.

   P4  The average number of encounters needed to deliver a bundle in
       this zone.

   P5  The multiple of the average number of encounters needed to
       deliver a bundle (P4) after which it can be assumed that a node
       is not going to encounter a particular node again in the
       foreseeable future so that the delivery predictability ought to
       be decayed below P_first_threshold.

   P6  The number of encounters between a particular pair of nodes that
       should result in the delivery predictability of the encountered
       node getting close to the maximum possible delivery
       predictability (1 - delta).

   We can use these parameters to derive appropriate values for gamma
   and P_encounter_max, which are the key parameters in the evolution of
   the delivery predictabilities.  The values of the other parameters
   P_encounter_first (0.5), P_first_threshold (0.1), and delta (0.01),
   with the default values suggested in Figure 3, generally are not
   specific to the mobility model, although in special cases
   P_encounter_first may be different if extra information is available.

   To select a value for gamma:
   After a single, unrepeated encounter, the delivery predictability of
   the encountered node should decay from P_encounter_first to
   P_first_threshold in the expected time for P4 * P5 encounters.  Thus:





Lindgren, et al.              Experimental                    [Page 111]

RFC 6693                         PRoPHET                     August 2012


   P_first_threshold = P_encounter_first * gamma ^ ((P2 * P4 * P5)/P1)

   which can be rearranged as

   gamma =
   exp(ln(P_first_threshold/P_encounter_first) * P1 / (P2* P4 * P5)).

   Typical values of gamma will be less than 1, but very close to 1
   (usually greater than 0.99).  The value has to be stored to several
   decimal places of accuracy, but implementations can create a table of
   values for specific intervals to reduce the amount of on-the-fly
   calculation required.

   Selecting a value for P_encounter_max:
   Once gamma has been determined, the decay factor for the average time
   between encounters between a specific pair of nodes can be
   calculated:
   Decay_typ = gamma ^ ((P2 * P3)/P1)

   Starting with P_encounter_first, using Decay_typ and applying
   Equation 1 from Section 2.1.2 (P6 - 1) times, we can calculate the
   typical delivery predictability for the encountered node after P6
   encounters.  The nature of Equation 1 is such that it is not easy to
   produce a closed form that generates a value of P_encounter_max from
   the parameter values, but using a spreadsheet to apply the equation
   repeatedly and tabulate the results will allow a suitable value of
   P_encounter_max to be chosen very simply.  The evolution is not very
   sensitive to the value of P_encounter_max, and values in the range
   0.4 to 0.8 will generally be appropriate.  A value of 0.7 is
   recommended as a default.

   Once a PRoPHET zone has been in operation for some time, the logs of
   the actual encounters can and should be used to check that the
   selected parameters were appropriate and to tune them as necessary.
   In the longer term, it may prove possible to install a learning mode
   in nodes so that the parameters can be adjusted dynamically to
   maintain best congruence with the mobility model that may itself
   change over time.













Lindgren, et al.              Experimental                    [Page 112]

RFC 6693                         PRoPHET                     August 2012


Authors' Addresses



   Anders F. Lindgren
   Swedish Institute of Computer Science
   Box 1263
   Kista  SE-164 29
   SE

   Phone: +46707177269
   EMail: andersl@sics.se
   URI:   http://www.sics.se/~andersl


   Avri Doria
   Technicalities
   Providence  RI
   US

   EMail: avri@acm.org
   URI:   http://psg.com/~avri


   Elwyn Davies
   Folly Consulting
   Soham
   UK

   EMail: elwynd@folly.org.uk


   Samo Grasic
   Lulea University of Technology
   Lulea  SE-971 87
   SE

   EMail: samo.grasic@ltu.se















Lindgren, et al.              Experimental                    [Page 113]