RFC 3048

Network Working Group                                         B. Whetten
Request for Comments: 3048                                      Talarian
Category: Informational                                      L. Vicisano
                                                              R. Kermode
                                                              M. Handley
                                                                 ACIRI 9
                                                                S. Floyd
                                                                 M. Luby
                                                        Digital Fountain
                                                            January 2001

      Reliable Multicast Transport Building Blocks for One-to-Many
                           Bulk-Data Transfer

Status of this Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2001).  All Rights Reserved.


   This document describes a framework for the standardization of bulk-
   data reliable multicast transport.  It builds upon the experience
   gained during the deployment of several classes of contemporary
   reliable multicast transport, and attempts to pull out the
   commonalities between these classes of protocols into a number of
   building blocks.  To that end, this document recommends that certain
   components that are common to multiple protocol classes be
   standardized as "building blocks".  The remaining parts of the
   protocols, consisting of highly protocol specific, tightly
   intertwined functions, shall be designated as "protocol cores".
   Thus, each protocol can then be constructed by merging a "protocol
   core" with a number of "building blocks" which can be re-used across
   multiple protocols.

Whetten, et al.              Informational                      [Page 1]

RFC 3048                  RMT Building Blocks               January 2001

Table of Contents

   1 Introduction ..................................................  2
   1.1 Protocol Families ...........................................  5
   2 Building Blocks Rationale .....................................  6
   2.1 Building Blocks Advantages ..................................  6
   2.2 Building Block Risks ........................................  7
   2.3 Building Block Requirements .................................  8
   3 Protocol Components ...........................................  8
   3.1 Sub-Components Definition ...................................  9
   4 Building Block Recommendations ................................ 12
   4.1 NACK-based Reliability ...................................... 13
   4.2 FEC coding .................................................. 13
   4.3 Congestion Control .......................................... 13
   4.4 Generic Router Support ...................................... 14
   4.5 Tree Configuration .......................................... 14
   4.6 Data Security ............................................... 15
   4.7 Common Headers .............................................. 15
   4.8 Protocol Cores .............................................. 15
   5 Security ...................................................... 15
   6 IANA Considerations ........................................... 15
   7 Conclusions ................................................... 16
   8 Acknowledgements .............................................. 16
   9 References .................................................... 16
   10 Authors' Addresses ........................................... 19
   11 Full Copyright Statement ..................................... 20

1.  Introduction

   RFC 2357 lays out the requirements for reliable multicast protocols
   that are to be considered for standardization by the IETF.  They

   o  Congestion Control.  The protocol must be safe to deploy in the
      widespread Internet.  Specifically, it must adhere to three
      mandates:  a) it must achieve good throughput (i.e., it must not
      consistently overload links with excess data or repair traffic),
      b) it must achieve good link utilization, and c) it must not
      starve competing flows.

   o  Scalability.  The protocol should be able to work under a variety
      of conditions that include multiple network topologies, link
      speeds, and the receiver set size.  It is more important to have a
      good understanding of how and when a protocol breaks than when it

Whetten, et al.              Informational                      [Page 2]

RFC 3048                  RMT Building Blocks               January 2001

   o  Security.  The protocol must be analyzed to show what is necessary
      to allow it to cope with security and privacy issues.  This
      includes understanding the role of the protocol in data
      confidentiality and sender authentication, as well as how the
      protocol will provide defenses against denial of service attacks.

   These requirements are primarily directed towards making sure that
   any standards will be safe for widespread Internet deployment.  The
   advancing maturity of current work on reliable multicast congestion
   control (RMCC) [HFW99] in the IRTF Reliable Multicast Research Group
   (RMRG) has been one of the events that has allowed the IETF to
   charter the RMT working group.  RMCC only addresses a subset of the
   design space for reliable multicast.  Fortuitously, the requirements
   it addresses are also the most pressing application and market

   A protocol's ability to meet the requirements of congestion control,
   scalability, and security is affected by a number of secondary
   requirements that are described in a separate document [RFC2887].  In
   summary, these are:

   o  Ordering Guarantees.  A protocol must offer at least one of either
      source ordered or unordered delivery guarantees.  Support for
      total ordering across multiple senders is not recommended, as it
      makes it more difficult to scale the protocol, and can more easily
      be implemented at a higher level.

   o  Receiver Scalability.  A protocol should be able to support a
      "large" number of simultaneous receivers per transport group.  A
      typical receiver set could be on the order of at least 1,000 -
      10,000 simultaneous receivers per group, or could even eventually
      scale up to millions of receivers in the large Internet.

   o  Real-Time Feedback.  Some versions of RMCC may require soft real-
      time feedback, so a protocol may provide some means for this
      information to be measured and returned to the sender.  While this
      does not require that a protocol deliver data in soft real-time,
      it is an important application requirement that can be provided
      easily given real-time feedback.

   o  Delivery Guarantees.  In many applications, a logically defined
      unit or units of data is to be delivered to multiple clients,
      e.g., a file or a set of files, a software package, a stock quote
      or package of stock quotes, an event notification, a set of
      slides, a frame or block from a video.  An application data unit
      is defined to be a logically separable unit of data that is useful
      to the application.  In some cases, an application data unit may
      be short enough to fit into a single packet (e.g., an event

Whetten, et al.              Informational                      [Page 3]

RFC 3048                  RMT Building Blocks               January 2001

      notification or a stock quote), whereas in other cases an
      application data unit may be much longer than a packet (e.g., a
      software package).  A protocol must provide good throughput of
      application data units to receivers.  This means that most data
      that is delivered to receivers is useful in recovering the
      application data unit that they are trying to receive.  A protocol
      may optionally provide delivery confirmation, i.e., a mechanism
      for receivers to inform the sender when data has been delivered.
      There are two types of confirmation, at the application data unit
      level and at the packet level.  Application data unit confirmation
      is useful at the application level, e.g., to inform the
      application about receiver progress and to decide when to stop
      sending packets about a particular application data unit.  Packet
      confirmation is useful at the transport level, e.g., to inform the
      transport level when it can release buffer space being used for
      storing packets for which delivery has been confirmed.  Packet
      level confirmation may also aid in application data unit

   o  Network Topologies.  A protocol must not break the network when
      deployed in the full Internet.  However, we recognize that
      intranets will be where the first wave of deployments happen, and
      so are also very important to support.  Thus, support for
      satellite networks (including those with terrestrial return paths
      or no return paths at all) is encouraged, but not required.

   o  Group Membership.  The group membership algorithms must be
      scalable.  Membership can be anonymous (where the sender does not
      know the list of receivers), or fully distributed (where the
      sender receives a count of the number of receivers, and optionally
      a list of failures).

   o  Example Applications.  Some of the applications that a RM protocol
      could be designed to support include multimedia broadcasts, real
      time financial market data distribution, multicast file transfer,
      and server replication.

   In the rest of this document the following terms will be used with a
   specific connotation: "protocol family", "protocol component",
   "building block", "protocol core", and "protocol instantiation".  A
   "protocol family" is a broad class of RM protocols which share a
   common characteristic.  In our classification, this characteristic is
   the mechanism used to achieve reliability.  A "protocol component" is
   a logical part of the protocol that addresses a particular
   functionality.  A "building block" is a constituent of a protocol
   that implements one, more than one or a part of a component.  A
   "protocol core" is the set of functionality required for the

Whetten, et al.              Informational                      [Page 4]

RFC 3048                  RMT Building Blocks               January 2001

   instantiation of a complete protocol, that is not specified by any
   building block.  Finally a "protocol instantiation" is an actual RM
   protocol defined in term of building blocks and a protocol core.

1.1.  Protocol Families

   The design-space document [RFC2887] also provides a taxonomy of the
   most popular approaches that have been proposed over the last ten
   years.  After congestion control, the primary challenge has been that
   of meeting the requirement for ensuring good throughput in a way that
   scales to a large number of receivers.  For protocols that include a
   back-channel for recovery of lost packets, the ability to take
   advantage of support of elements in the network has been found to be
   very beneficial for supporting good throughput for a large numbers of
   receivers.  Other protocols have found it very beneficial to transmit
   coded data to achieve good throughput for large numbers of receivers.

   This taxonomy breaks proposed protocols into four families.  Some
   protocols in the family provide packet level delivery confirmation
   that may be useful to the transport level.  All protocols in all
   families can be supplemented with higher level protocols that provide
   delivery confirmation of application data units.

   1  NACK only.  Protocols such as SRM [FJM95] and MDP2 [MA99] attempt
      to limit traffic by only using NACKs for requesting packet
      retransmission.  They do not require network infrastructure.

   2  Tree based ACK.  Protocols such as RMTP [LP96, PSLM97], RMTP-II
      [WBPM98] and TRAM [KCW98], use positive acknowledgments (ACKs).
      ACK based protocols reduce the need for supplementary protocols
      that provide delivery confirmation, as the ACKS can be used for
      this purpose.  In order to avoid ACK implosion in scaled up
      deployments, the protocol can use servers placed in the network.

   3  Asynchronous Layered Coding (ALC).  These protocols (examples
      include [RV97] and [BLMR98]) use sender-based Forward Error
      Correction (FEC) methods with no feedback from receivers or the
      network to ensure good throughput.  These protocols also used
      sender-based layered multicast and receiver-driven protocols to
      join and leave these layers with no feedback to the sender to
      achieve scalable congestion control.

   4  Router assist.  Like SRM, protocols such as PGM [FLST98] and
      [LG97] also use negative acknowledgments for packet recovery.
      These protocols take advantage of new router software to do
      constrained negative acknowledgments and retransmissions.  Router
      assist protocols can also provide other functionality more
      efficiently than end to end protocols.  For example, [LVS99] shows

Whetten, et al.              Informational                      [Page 5]

RFC 3048                  RMT Building Blocks               January 2001

      how router assist can provide fine grained congestion control for
      ALC protocols.  Router assist protocols can be designed to
      complement all protocol families described above.

   Note that the distinction in protocol families in not necessarily
   precise and mutually exclusive.  Actual protocols may use a
   combination of the mechanisms belonging to different classes.  For
   example, hybrid NACK/ACK based protocols (such as [WBPM98]) are
   possible.  Other examples are protocols belonging to class 1 through
   3 that take advantage of router support.

2.  Building Blocks Rationale

   As specified in RFC 2357 [MRBP98], no single reliable multicast
   protocol will likely meet the needs of all applications.  Therefore,
   the IETF expects to standardize a number of protocols that are
   tailored to application and network specific needs.  This document
   concentrates on the requirements for "one-to-many bulk-data
   transfer", but in the future, additional protocols and building-
   blocks are expected that will address the needs of other types of
   applications, including "many-to- many" applications.  Note that
   bulk-data transfer does not refer to the timeliness of the data,
   rather it states that there is a large amount of data to be
   transferred in a session.  The scope and approach taken for the
   development of protocols for these additional scenarios will depend
   upon large part on the success of the "building-block" approach put
   forward in this document.

2.1.  Building Blocks Advantages

   Building a large piece of software out of smaller modular components
   is a well understood technique of software engineering.  Some of the
   advantages that can come from this include:

   o  Specification Reuse.  Modules can be used in multiple protocols,
      which reduces the amount of development time required.

   o  Reduced Complexity.  To the extent that each module can be easily
      defined with a simple API, breaking a large protocol in to smaller
      pieces typically reduces the total complexity of the system.

   o  Reduced Verification and Debugging Time.  Reduced complexity
      results in reduced time to debug the modules.  It is also usually
      faster to verify a set of smaller modules than a single larger

Whetten, et al.              Informational                      [Page 6]

RFC 3048                  RMT Building Blocks               January 2001

   o  Easier Future Upgrades.  There is still ongoing research in
      reliable multicast, and we expect the state of the art to continue
      to evolve.  Building protocols with smaller modules allows them to
      be more easily upgraded to reflect future research.

   o  Common Diagnostics.  To the extent that multiple protocols share
      common packet headers, packet analyzers and other diagnostic tools
      can be built which work with multiple protocols.

   o  Reduces Effort for New Protocols.  As new application requirements
      drive the need for new standards, some existing modules may be
      reused in these protocols.

   o  Parallelism of Development.  If the APIs are defined clearly, the
      development of each module can proceed in parallel.

2.2.  Building Block Risks

   Like most software specification, this technique of breaking down a
   protocol in to smaller components also brings tradeoffs.  After a
   certain point, the disadvantages outweigh the advantages, and it is
   not worthwhile to further subdivide a problem.  These risks include:

   o  Delaying Development.  Defining the API for how each two modules
      inter-operate takes time and effort.  As the number of modules
      increases, the number of APIs can increase at more than a linear
      rate.  The more tightly coupled and complex a component is, the
      more difficult it is to define a simple API, and the less
      opportunity there is for reuse.  In particular, the problem of how
      to build and standardize fine grained building blocks for a
      transport protocol is a difficult one, and in some cases requires
      fundamental research.

   o  Increased Complexity.  If there are too many modules, the total
      complexity of the system actually increases, due to the
      preponderance of interfaces between modules.

   o  Reduced Performance.  Each extra API adds some level of processing
      overhead.  If an API is inserted in to the "common case" of packet
      processing, this risks degrading total protocol performance.

   o  Abandoning Prior Work.  The development of robust transport
      protocols is a long and time intensive process, which is heavily
      dependent on feedback from real deployments.  A great deal of work
      has been done over the past five years on components of protocols
      such as RMTP-II, SRM, and PGM.  Attempting to dramatically re-
      engineer these components risks losing the benefit of this prior

Whetten, et al.              Informational                      [Page 7]

RFC 3048                  RMT Building Blocks               January 2001

2.3.  Building Block Requirements

   Given these tradeoffs, we propose that a building block must meet the
   following requirements:

   o  Wide Applicability.  In order to have confidence that the
      component can be reused, it should apply across multiple protocol
      families and allow for the component's evolution.

   o  Simplicity.  In order to have confidence that the specification of
      the component APIs will not dramatically slow down the standards
      process, APIs must be simple and straight forward to define.  No
      new fundamental research should be done in defining these APIs.

   o  Performance.  To the extent possible, the building blocks should
      attempt to avoid breaking up the "fast track", or common case
      packet processing.

3.  Protocol Components

   This section proposes a functional decomposition of RM bulk-data
   protocols from the perspective of the functional components provided
   to an application by a transport protocol.  It also covers some
   components that while not necessarily part of the transport protocol,
   are directly impacted by the specific requirements of a reliable
   multicast transport.  The next section then specifies recommended
   building blocks that can implement these components.

   Although this list tries to cover all the most common transport-
   related needs of one-to-many bulk-data transfer applications, new
   application requirements may arise during the process of
   standardization, hence this list must not be interpreted as a
   statement of what the transport layer should provide and what it
   should not.  Nevertheless, it must be pointed out that some
   functional components have been deliberately omitted since they have
   been deemed irrelevant to the type of application considered (i.e.,
   one-to-many bulk-data transfer).  Among these are advanced message
   ordering (i.e., those which cannot be implemented through a simple
   sequence number) and atomic delivery.

   It is also worth mentioning that some of the functional components
   listed below may be required by other functional components and not
   directly by the application (e.g., membership knowledge is usually
   required to implement ACK-based reliability).

   The following list covers various transport functional components and
   splits them in sub-components.

Whetten, et al.              Informational                      [Page 8]

RFC 3048                  RMT Building Blocks               January 2001

   o  Data Reliability (ensuring good throughput)    |
                          | - Loss Detection/Notification
                          | - Loss Recovery
                          | - Loss Protection

   o  Congestion Control  |
                          | - Congestion Feedback
                          | - Rate Regulation
                          | - Receiver Controls

   o  Security

   o  Group membership    |
                          | - Membership Notification
                          | - Membership Management

   o  Session Management  |
                          | - Group Membership Tracking
                          | - Session Advertisement
                          | - Session Start/Stop
                          | - Session Configuration/Monitoring

   o  Tree Configuration

   Note that not all components are required by all protocols, depending
   upon the fully defined service that is being provided by the
   protocol.  In particular, some minimal service models do not require
   many of these functions, including loss notification, loss recovery,
   and group membership.

3.1.  Sub-Components Definition

   Loss Detection/Notification.  This includes how missing packets are
   detected during transmission and how knowledge of these events are
   propagated to one or more agents which are designated to recover from
   the transmission error.  This task raises major scalability issues
   and can lead to feedback implosion and poor throughput if not
   properly handled.  Mechanisms based on TRACKs (tree-based positive
   acknowledgements) or NACKs (negative acknowledgements) are the most
   widely used to perform this function.  Mechanisms based on a
   combination of TRACKs and NACKs are also possible.

   Loss Recovery.  This function responds to loss notification events
   through the transmission of additional packets, either in the form of
   copies of those packets lost or in the form of FEC packets.  The
   manner in which this function is implemented can significantly affect
   the scalability of a protocol.

Whetten, et al.              Informational                      [Page 9]

RFC 3048                  RMT Building Blocks               January 2001

   Loss Protection.  This function attempts to mask packet-losses so
   that they don't become Loss Notification events.  This function can
   be realized through the pro-active transmission of FEC packets.  Each
   FEC packet is created from an entire application data unit [LMSSS97]
   or a portion of an application data unit [RV97], [BKKKLZ95], a fact
   that allows a receiver to recover from some packet loss without
   further retransmissions.  The number of losses that can be recovered
   from without requiring retransmission depends on the amount of FEC
   packets sent in the first place.  Loss protection can also be pushed
   to the extreme when good throughput is achieved without any Loss
   Detection/Notification and Loss Recovery functionality, as in the ALC
   family of protocols defined above.

   Congestion Feedback.  For sender driven congestion control protocols,
   the receiver must provide some type of feedback on congestion to the
   sender.  This typically involves loss rate and round trip time

   Rate Regulation.  Given the congestion feedback, the sender then must
   adjust its rate in a way that is fair to the network.  One proposal
   that defines this notion of fairness and other congestion control
   requirements is [Whetten99].

   Receiver Controls.  In order to avoid allowing a receiver that has an
   extremely slow connection to the sender to stop all progress within
   single rate schemes, a congestion control algorithm will often
   require receivers to leave groups.  For multiple rate approaches,
   receivers of all connection speeds can have data delivered to them
   according to the rate of their connection without slowing down other

   Security.  Security for reliable multicast contains a number of
   complex and tricky issues that stem in large part from the IP
   multicast service model.  In this service model, hosts do not send
   traffic to another host, but instead elect to receive traffic from a
   multicast group. This means that any host may join a group and
   receive its traffic.  Conversely, hosts may also leave a group at any
   time.  Therefore, the protocol must address how it impacts the
   following security issues:

   o  Sender Authentication (since any host can send to a group),

   o  Data Encryption (since any host can join a group)

   o  Transport Protection (denial of service attacks, through
      corruption of transport state, or requests for unauthorized

Whetten, et al.              Informational                     [Page 10]

RFC 3048                  RMT Building Blocks               January 2001

   o  Group Key Management (since hosts may join and leave a group at
      any time) [WHA98]

   In particular, a transport protocol needs to pay particular attention
   to how it protects itself from denial of service attacks, through
   mechanisms such as lightweight authentication of control packets

   With Source Specific Multicast service model (SSM), a host joins
   specifically to a sender and group pair.  Thus, SSM offers more
   security against hosts receiving traffic from a denial of service
   attack where an arbitrary sender sends packets that hosts did not
   specifically request to receive.  Nevertheless, it is recommended
   that additional protections against such attacks should be provided
   when using SSM, because the protection offered by SSM against such
   attacks may not be enough.

   Sender Authentication, Data Encryption, and Group Key Management.
   While these functions are not typically part of the transport layer
   per se, a protocol needs to understand what ramifications it has on
   data security, and may need to have special interfaces to the
   security layer in order to accommodate these ramifications.

   Transport Protection.  The primary security task for a transport
   layer is that of protecting the transport layer itself from attack.
   The most important function for this is typically lightweight
   authentication of control packets in order to prevent corruption of
   state and other denial of service attacks.

   Membership Notification.  This is the function through which the data
   source--or upper level agent in a possible hierarchical
   organization--learns about the identity and/or number of receivers or
   lower level agents.  To be scalable, this typically will not provide
   total knowledge of the identity of each receiver.

   Membership Management.  This implements the mechanisms for members to
   join and leave the group, to accept/refuse new members, or to
   terminate the membership of existing members.

   Group Membership Tracking.  As an optional feature, a protocol may
   interface with a component which tracks the identity of each receiver
   in a large group.  If so, this feature will typically be implemented
   out of band, and may be implemented by an upper level protocol.  This
   may be useful for services that require tracking of usage of the
   system, billing, and usage reports.

Whetten, et al.              Informational                     [Page 11]

RFC 3048                  RMT Building Blocks               January 2001

   Session Advertisement.  This publishes the session name/contents and
   the parameters needed for its reception. This function is usually
   performed by an upper layer protocol (e.g., [HPW99] and [HJ98]).

   Session Start/Stop.  These functions determine the start/stop time of
   sender and/or receivers.  In many cases this is implicit or performed
   by an upper level application or protocol.  In some protocols,
   however, this is a task best performed by the transport layer due to
   scalability requirements.

   Session Configuration/Monitoring.  Due to the potentially far
   reaching scope of a multicast session, it is particularly important
   for a protocol to include tools for configuring and monitoring the
   protocol's operation.

   Tree Configuration.  For protocols which include hierarchical
   elements (such as PGM and RMTP-II), it is important to configure
   these elements in a way that has approximate congruence with the
   multicast routing topology.  While tree configuration could be
   included as part of the session configuration tools, it is clearly
   better if this configuration can be made automatic.

4.  Building Block Recommendations

   The families of protocols introduced in section 1.1 generally use
   different mechanisms to implement the protocol functional components
   described in section 3.  This section tries to group these mechanisms
   in macro components that define protocol building blocks.

   A building block is defined as
      "a logical protocol component that results in explicit APIs for use
      by other building blocks or by the protocol client."

   Building blocks are generally specified in terms of the set of
   algorithms and packet formats that implement protocol functional
   components.  A building block may also have API's through which it
   communicates to applications and/or other building blocks.  Most
   building blocks should also have a management API, through which it
   communicates to SNMP and/or other management protocols.

   In the following section we will list a number of building blocks
   which, at this stage, seem to cover most of the functional components
   needed to implement the protocol families presented in section 1.1.
   Nevertheless this list represents the "best current guess", and as
   such it is not meant to be exhaustive.  The actual building block
   decomposition, i.e., the division of functional components into
   building blocks, may also have to be revised in the future.

Whetten, et al.              Informational                     [Page 12]

RFC 3048                  RMT Building Blocks               January 2001

4.1.  NACK-based Reliability

   This building block defines NACK-based loss detection/notification
   and recovery.  The major issues it addresses are implosion prevention
   (suppression) and NACK semantics (i.e., how packets to be
   retransmitted should be specified, both in the case of selective and
   FEC loss repair).  Suppression mechanisms to be considered are:

   o  Multicast NACKs
   o  Unicast NACKs and Multicast confirmation

   These suppression mechanisms primarily need to both minimize delay
   while also minimizing redundant messages.  They may also need to have
   special weighting to work with Congestion Feedback.

4.2.  FEC coding

   This building block is concerned with packet level FEC information
   when FEC codes are used either proactively or as repairs in reaction
   to lost packets.  It specifies the FEC codec selection and the FEC
   packet naming (indexing) for both reactive FEC repair and pro-active

4.3.  Congestion Control

   There will likely be multiple versions of this building block,
   corresponding to different design policies in addressing congestion
   control.  Two main approaches are considered for the time being: a
   source-based rate regulation with a single rate provided to all the
   receivers in the session, and a multiple rate receiver-driven
   approach with different receivers receiving at different rates in the
   same session.  The multiple rate approach may use multiple layers of
   multicast traffic [VRC98] or router filtering of a single layer
   [LVS99].  The multiple rate approach is most applicable for ALC

   Both approaches are still in the phase of study, however the first
   seems to be mature enough [HFW99] to allow the standardization
   process to begin.

   At the time of writing this document, a third class of congestion
   control algorithm based on router support is beginning to emerge in
   the IRTF RMRG [LVS99].  This work may lead to the future
   standardization of one or more additional building blocks for
   congestion control.

Whetten, et al.              Informational                     [Page 13]

RFC 3048                  RMT Building Blocks               January 2001

4.4.  Generic Router Support

   The task of designing RM protocols can be made much easier by the
   presence of some specific support in routers.  In some application-
   specific cases, the increased benefits afforded by the addition of
   special router support can justify the resulting additional
   complexity and expense [FLST98].

   Functional components which can take advantage of router support
   include feedback aggregation/suppression (both for loss notification
   and congestion control) and constrained retransmission of repair
   packets.  Another component that can take advantage of router support
   is intentional packet filtering to provide different rates of
   delivery of packets to different receivers from the same multicast
   packet stream.  This could be most advantageous when combined with
   ALC protocols [LVS99].

   The process of designing and deploying these mechanisms inside
   routers can be much slower than the one required for end-host
   protocol mechanisms.  Therefore, it would be highly advantageous to
   define these mechanisms in a generic way that multiple protocols can
   use if it is available, but do not necessarily need to depend on.

   This component has two halves, a signaling protocol and actual router
   algorithms.  The signaling protocol allows the transport protocol to
   request from the router the functions that it wishes to perform, and
   the router algorithms actually perform these functions.  It is more
   urgent to define the signaling protocol, since it will likely impact
   the common case protocol headers.

   An important component of the signaling protocol is some level of
   commonality between the packet headers of multiple protocols, which
   allows the router to recognize and interpret the headers.

4.5.  Tree Configuration

   It has been shown that the scalability of RM protocols can be greatly
   enhanced by the insertion of some kind of retransmission or feedback
   aggregation agents between the source and receivers.  These agents
   are then used to form a tree with the source at (or near) the root,
   the receivers at the leaves of the tree, and the aggregation/local
   repair nodes in the middle.  The internal nodes can either be
   dedicated software for this task, or they may be receivers that are
   performing dual duty.

   The effectiveness of these agents to assist in the delivery of data
   is highly dependent upon how well the logical tree they use to
   communicate matches the underlying routing topology.  The purpose of

Whetten, et al.              Informational                     [Page 14]

RFC 3048                  RMT Building Blocks               January 2001

   this building block would be to construct and manage the logical tree
   connecting the agents.  Ideally, this building block would perform
   these functions in a manner that adapts to changes in session
   membership, routing topology, and network availability.

4.6.  Data Security

   At the time of writing, the security issues are the subject of
   research within the IRTF Secure Multicast Group (SMuG).  Solutions
   for these requirements will be standardized within the IETF when

4.7.  Common Headers

   As pointed out in the generic router support section, it is important
   to have some level of commonality across packet headers.  It may also
   be useful to have common data header formats for other reasons.  This
   building block would consist of recommendations on fields in their
   packet headers that protocols should make common across themselves.

4.8.  Protocol Cores

   The above building blocks consist of the functional components listed
   in section 3 that appear to meet the requirements for being
   implemented as building blocks presented in section 2.

   The other functions from section 3, which are not covered above,
   should be implemented as part of "protocol cores", specific to each
   protocol standardized.

5.  Security Considerations

   RFC 2357 specifically states that "reliable multicast Internet-Drafts
   reviewed by the Transport Area Directors must explicitly explore the
   security aspects of the proposed design."  Specifically, RMT building
   block works in progress must examine the denial-of-service attacks
   that can be made upon building blocks and affected by building blocks
   upon the Internet at large.  This requirement is in addition to any
   discussions regarding data-security, that is the manipulation of or
   exposure of session information to unauthorized receivers.  Readers
   are referred to section 5.e of RFC 2357 for further details.

6.  IANA Considerations

   There will be more than one building block, and possibly multiple
   versions of individual building blocks as their designs are refined.
   For this reason, the creation of new building blocks and new building
   block versions will be administered via a building block registry

Whetten, et al.              Informational                     [Page 15]

RFC 3048                  RMT Building Blocks               January 2001

   that will be administered by IANA.  Initially, this registry will be
   empty, since the building blocks described in sections 4.1 to 4.3 are
   presented for example and design purposes.  The requested IANA
   building block registry will be populated from specifications as they
   are approved for RFC publication (using the "Specification Required"
   policy as described in RFC 2434 [RFC2434]).  A registration will
   consist of a building block name, a version number, a brief text
   description, a specification RFC number, and a responsible person, to
   which IANA will assign the type number.

7.  Conclusions

   In this document, we briefly described a number of building blocks
   that may be used for the generation of reliable multicast protocols
   to be used in the application space of one-to-many reliable bulk-data
   transfer.  The list of building blocks presented was derived from
   considering the functions that a protocol in this space must perform
   and how these functions should be grouped.  This list is not intended
   to be all-inclusive but instead to act as guide as to which building
   blocks are considered during the standardization process within the
   Reliable Multicast Transport WG.

8.  Acknowledgements

   This document represents an overview of a number of building blocks
   for one to many bulk data transfer that may be ready for
   standardization within the RMT working group.  The ideas presented
   are not those of the authors, rather they are a summarization of many
   years of research into multicast transport combined with the varied
   presentations and discussions in the IRTF Reliable Multicast Research
   Group.  Although they are too numerous to list here, we thank
   everyone who has participated in these discussions for their

9.  References

   [BKKKLZ95]  J. Bloemer, M. Kalfane, M. Karpinski, R. Karp, M. Luby,
               D.  Zuckerman, "An XOR-based Erasure Resilient Coding
               Scheme," ICSI Technical Report No. TR-95-048, August

   [BLMR98]    J. Byers, M. Luby, M. Mitzenmacher, A. Rege, "A Digital
               Fountain Approach to Reliable Distribution of Bulk Data,"
               Proc ACM SIGCOMM 98.

   [FJM95]     S. Floyd, V. Jacobson, S. McCanne, "A Reliable Multicast
               Framework for Light-weight Sessions and Application Level
               Framing," Proc ACM SIGCOMM 95, Aug 1995 pp. 342-356.

Whetten, et al.              Informational                     [Page 16]

RFC 3048                  RMT Building Blocks               January 2001

   [FLST98]    D. Farinacci, S. Lin, T. Speakman, and A. Tweedly, "PGM
               reliable transport protocol specification," Work in

   [HFW99]     M. Handley, S. Floyd, B. Whetten, "Strawman Specification
               for TCP Friendly (Reliable) Multicast Congestion Control
               (TFMCC)," Work in Progress.

   [HJ98]      Handley, M. and V. Jacobson, "SDP: Session Description
               Protocol", RFC 2327, April 1998.

   [HPW99]     M. Handley, C. Perkins, E. Whelan, "Session Announcement
               Protocol," Work in Progress, June 1999.

   [HW99]      T. Hardjorno, B. Whetten,  "Security Requirements for
               RMTP-II," Work in Progress, June 1999.

   [RFC2887]   Handley, M., Whetten, B., Kermode, R., Floyd, S.,
               Vicisano, L. and M. Luby, "The Reliable Multicast Design
               Space for Bulk Data Transfer", RFC 2887, August 2000.

   [KCW98]     M. Kadansky, D. Chiu, and J. Wesley, "Tree-based reliable
               multicast (TRAM)," Work in Progress.

   [Kermode98] R. Kermode, "Scoped Hybrid Automatic Repeat Request with
               Forward Error Correction," Proc ACM SIGCOMM 98, Sept

   [LDW98]     M. Lucas, B. Dempsey, A. Weaver, "MESH: Distributed Error
               Recovery for Multimedia Streams in Wide-Area Multicast

   [LESZ97]    C-G. Liu, D. Estrin, S. Shenkar, L. Zhang, "Local Error
               Recovery in SRM: Comparison of Two Approaches," USC
               Technical Report 97-648, Jan 1997.

   [LG97]      B.N. Levine, J.J. Garcua-Luna-Aceves, "Improving Internet
               Multicast Routing with Routing Labels," IEEE
               International Conference on Network Protocols (ICNP-97),
               Oct 28-31, 1997, p.241-250.

   [LP96]      K. Lin and S. Paul. "RMTP: A Reliable Multicast Transport
               Protocol," IEEE INFOCOMM 1996, March 1996, pp. 1414-1424.

   [LMSSS97]   M. Luby, M. Mitzenmacher, A. Shokrollahi, D. Spielman, V.
               Stemann, "Practical Loss-Resilient Codes", Proc ACM
               Symposium on Theory of Computing, 1997.

Whetten, et al.              Informational                     [Page 17]

RFC 3048                  RMT Building Blocks               January 2001

   [LVS99]     M. Luby, L. Vicisano, T. Speakman. "Heterogeneous
               multicast congestion control based on router packet
               filtering", RMT working group, June 1999, Pisa, Italy.

   [MA99]      J. Macker, B. Adamson. "Multicast Dissemination Protocol
               version 2 (MDPv2)," Work in Progress,

   [MRBP98]    Mankin, A., Romanow, A., Brander, S. and V.Paxson, "IETF
               Criteria for Evaluating Reliable Multicast Transport and
               Application Protocols", RFC 2357, June 1998.

   [RFC2434]   Narten, T. and H. Alvestrand, "Guidelines for Writing an
               IANA Considerations Section in RFCs", BCP 26, RFC 2434,
               October 1998.

   [OXB99]     O. Ozkasap, Z. Xiao, K. Birman.  "Scalability of Two
               Reliable Multicast Protocols", Work in Progress, May

   [PSLB97]    "Reliable Multicast Transport Protocol (RMTP)," S. Paul,
               K. K. Sabnani, J. C. Lin, and S. Bhattacharyya, IEEE
               Journal on Selected Areas in Communications, Vol. 15, No.
               3, April 1997.

   [RV97]      L. Rizzo, L. Vicisano, "A Reliable Multicast Data
               Distribution Protocol Based on Software FEC Techniques,"
               Proc. of The Fourth IEEE Workshop on the Architecture and
               Implementation of High Performance Communication Systems
               (HPCS'97), Sani Beach, Chalkidiki, Greece June 23-25,

   [VRC98]     L. Vicisano, L. Rizzo, J. Crowcroft, "TCP-Like Congestion
               Control for Layered Multicast Data Transfer", Proc. of
               IEEE Infocom'98, March 1998.

   [WBPM98]    B. Whetten, M. Basavaiah, S. Paul, T. Montgomery, N.
               Rastogi, J. Conlan, and T. Yeh, "THE RMTP-II PROTOCOL,"
               Work in Progress.

   [WHA98]     D. Wallner, E. Hardler, R. Agee, "Key Management for
               Multicast: Issues and Architectures," Work in Progress.

   [Whetten99] B. Whetten,  "A Proposal for Reliable Multicast
               Congestion Control Requirements," Work in Progress.

Whetten, et al.              Informational                     [Page 18]

RFC 3048                  RMT Building Blocks               January 2001

10.  Authors' Addresses

   Brian Whetten
   Talarian Corporation,
   333 Distel Circle,
   Los Altos, CA 94022, USA

   EMail: whetten@talarian.com

   Lorenzo Vicisano
   Cisco Systems,
   170 West Tasman Dr.
   San Jose, CA 95134, USA

   EMail: lorenzo@cisco.com

   Roger Kermode
   Motorola Australian Research Centre
   Level 3, 12 Lord St,
   Botany  NSW  2019, Australia

   EMail: Roger.Kermode@motorola.com

   Mark Handley, Sally Floyd
   ATT Center for Internet Research at ICSI,
   International Computer Science Institute,
   1947 Center Street, Suite 600,
   Berkeley, CA 94704, USA

   EMail: mjh@aciri.org, floyd@aciri.org

   Michael Luby
   600 Alabama Street
   San Francisco, CA  94110
   Digital Fountain, Inc.

   EMail: luby@digitalfountain.com

Whetten, et al.              Informational                     [Page 19]

RFC 3048                  RMT Building Blocks               January 2001

11.  Full Copyright Statement

   Copyright (C) The Internet Society (2001).  All Rights Reserved.

   This document and translations of it may be copied and furnished to
   others, and derivative works that comment on or otherwise explain it
   or assist in its implementation may be prepared, copied, published
   and distributed, in whole or in part, without restriction of any
   kind, provided that the above copyright notice and this paragraph are
   included on all such copies and derivative works.  However, this
   document itself may not be modified in any way, such as by removing
   the copyright notice or references to the Internet Society or other
   Internet organizations, except as needed for the purpose of
   developing Internet standards in which case the procedures for
   copyrights defined in the Internet Standards process must be
   followed, or as required to translate it into languages other than

   The limited permissions granted above are perpetual and will not be
   revoked by the Internet Society or its successors or assigns.

   This document and the information contained herein is provided on an


   Funding for the RFC Editor function is currently provided by the
   Internet Society.

Whetten, et al.              Informational                     [Page 20]