RFC 8990

Internet Engineering Task Force (IETF)                        C. Bormann
Request for Comments: 8990                        Universität Bremen TZI
Category: Standards Track                              B. Carpenter, Ed.
ISSN: 2070-1721                                        Univ. of Auckland
                                                             B. Liu, Ed.
                                            Huawei Technologies Co., Ltd
                                                                May 2021

              GeneRic Autonomic Signaling Protocol (GRASP)


   This document specifies the GeneRic Autonomic Signaling Protocol
   (GRASP), which enables autonomic nodes and Autonomic Service Agents
   to dynamically discover peers, to synchronize state with each other,
   and to negotiate parameter settings with each other.  GRASP depends
   on an external security environment that is described elsewhere.  The
   technical objectives and parameters for specific application
   scenarios are to be described in separate documents.  Appendices
   briefly discuss requirements for the protocol and existing protocols
   with comparable features.

Status of This Memo

   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at

Copyright Notice

   Copyright (c) 2021 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
   (https://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.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents

   1.  Introduction
   2.  Protocol Overview
     2.1.  Terminology
     2.2.  High-Level Deployment Model
     2.3.  High-Level Design
     2.4.  Quick Operating Overview
     2.5.  GRASP Basic Properties and Mechanisms
       2.5.1.  Required External Security Mechanism
       2.5.2.  Discovery Unsolicited Link-Local (DULL) GRASP
       2.5.3.  Transport Layer Usage
       2.5.4.  Discovery Mechanism and Procedures
       2.5.5.  Negotiation Procedures
       2.5.6.  Synchronization and Flooding Procedures
     2.6.  GRASP Constants
     2.7.  Session Identifier (Session ID)
     2.8.  GRASP Messages
       2.8.1.  Message Overview
       2.8.2.  GRASP Message Format
       2.8.3.  Message Size
       2.8.4.  Discovery Message
       2.8.5.  Discovery Response Message
       2.8.6.  Request Messages
       2.8.7.  Negotiation Message
       2.8.8.  Negotiation End Message
       2.8.9.  Confirm Waiting Message
       2.8.10. Synchronization Message
       2.8.11. Flood Synchronization Message
       2.8.12. Invalid Message
       2.8.13. No Operation Message
     2.9.  GRASP Options
       2.9.1.  Format of GRASP Options
       2.9.2.  Divert Option
       2.9.3.  Accept Option
       2.9.4.  Decline Option
       2.9.5.  Locator Options
     2.10. Objective Options
       2.10.1.  Format of Objective Options
       2.10.2.  Objective Flags
       2.10.3.  General Considerations for Objective Options
       2.10.4.  Organizing of Objective Options
       2.10.5.  Experimental and Example Objective Options
   3.  Security Considerations
   4.  CDDL Specification of GRASP
   5.  IANA Considerations
   6.  References
     6.1.  Normative References
     6.2.  Informative References
   Appendix A.  Example Message Formats
     A.1.  Discovery Example
     A.2.  Flood Example
     A.3.  Synchronization Example
     A.4.  Simple Negotiation Example
     A.5.  Complete Negotiation Example
   Appendix B.  Requirement Analysis of Discovery, Synchronization,
           and Negotiation
     B.1.  Requirements for Discovery
     B.2.  Requirements for Synchronization and Negotiation Capability
     B.3.  Specific Technical Requirements
   Appendix C.  Capability Analysis of Current Protocols

   Authors' Addresses

1.  Introduction

   The success of the Internet has made IP-based networks bigger and
   more complicated.  Large-scale ISP and enterprise networks have
   become more and more problematic for human-based management.  Also,
   operational costs are growing quickly.  Consequently, there are
   increased requirements for autonomic behavior in the networks.
   General aspects of Autonomic Networks are discussed in [RFC7575] and

   One approach is to largely decentralize the logic of network
   management by migrating it into network elements.  A reference model
   for Autonomic Networking on this basis is given in [RFC8993].  The
   reader should consult this document to understand how various
   autonomic components fit together.  In order to achieve autonomy,
   devices that embody Autonomic Service Agents (ASAs, [RFC7575]) have
   specific signaling requirements.  In particular, they need to
   discover each other, to synchronize state with each other, and to
   negotiate parameters and resources directly with each other.  There
   is no limitation on the types of parameters and resources concerned,
   which can include very basic information needed for addressing and
   routing, as well as anything else that might be configured in a
   conventional non-autonomic network.  The atomic unit of discovery,
   synchronization, or negotiation is referred to as a technical
   objective, i.e., a configurable parameter or set of parameters
   (defined more precisely in Section 2.1).

   Negotiation is an iterative process, requiring multiple message
   exchanges forming a closed loop between the negotiating entities.  In
   fact, these entities are ASAs, normally but not necessarily in
   different network devices.  State synchronization, when needed, can
   be regarded as a special case of negotiation without iteration.  Both
   negotiation and synchronization must logically follow discovery.
   More details of the requirements are found in Appendix B.
   Section 2.3 describes a behavior model for a protocol intended to
   support discovery, synchronization, and negotiation.  The design of
   GeneRic Autonomic Signaling Protocol (GRASP) in Section 2 is based on
   this behavior model.  The relevant capabilities of various existing
   protocols are reviewed in Appendix C.

   The proposed discovery mechanism is oriented towards synchronization
   and negotiation objectives.  It is based on a neighbor discovery
   process on the local link, but it also supports diversion to peers on
   other links.  There is no assumption of any particular form of
   network topology.  When a device starts up with no preconfiguration,
   it has no knowledge of the topology.  The protocol itself is capable
   of being used in a small and/or flat network structure such as a
   small office or home network as well as in a large, professionally
   managed network.  Therefore, the discovery mechanism needs to be able
   to allow a device to bootstrap itself without making any prior
   assumptions about network structure.

   Because GRASP can be used as part of a decision process among
   distributed devices or between networks, it must run in a secure and
   strongly authenticated environment.

   In realistic deployments, not all devices will support GRASP.
   Therefore, some Autonomic Service Agents will directly manage a group
   of non-autonomic nodes, and other non-autonomic nodes will be managed
   traditionally.  Such mixed scenarios are not discussed in this

2.  Protocol Overview

2.1.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "OPTIONAL" in this document are to be interpreted as described in
   BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

   This document uses terminology defined in [RFC7575].

   The following additional terms are used throughout this document:

      A process by which an ASA discovers peers according to a specific
      discovery objective.  The discovery results may be different
      according to the different discovery objectives.  The discovered
      peers may later be used as negotiation counterparts or as sources
      of synchronization data.

      A process by which two ASAs interact iteratively to agree on
      parameter settings that best satisfy the objectives of both ASAs.

   State Synchronization:
      A process by which ASAs interact to receive the current state of
      parameter values stored in other ASAs.  This is a special case of
      negotiation in which information is sent, but the ASAs do not
      request their peers to change parameter settings.  All other
      definitions apply to both negotiation and synchronization.

   Technical Objective (usually abbreviated as Objective):
      A technical objective is a data structure whose main contents are
      a name and a value.  The value consists of a single configurable
      parameter or a set of parameters of some kind.  The exact format
      of an objective is defined in Section 2.10.1.  An objective occurs
      in three contexts: discovery, negotiation, and synchronization.
      Normally, a given objective will not occur in negotiation and
      synchronization contexts simultaneously.

         One ASA may support multiple independent objectives.

         The parameter(s) in the value of a given objective apply to a
         specific service or function or action.  They may in principle
         be anything that can be set to a specific logical, numerical,
         or string value, or a more complex data structure, by a network
         node.  Each node is expected to contain one or more ASAs which
         may each manage subsidiary non-autonomic nodes.

         Discovery Objective:  an objective in the process of discovery.
            Its value may be undefined.

         Synchronization Objective:  an objective whose specific
            technical content needs to be synchronized among two or more
            ASAs.  Thus, each ASA will maintain its own copy of the

         Negotiation Objective:  an objective whose specific technical
            content needs to be decided in coordination with another
            ASA.  Again, each ASA will maintain its own copy of the

         A detailed discussion of objectives, including their format, is
         found in Section 2.10.

   Discovery Initiator:
      An ASA that starts discovery by sending a Discovery message
      referring to a specific discovery objective.

   Discovery Responder:
      A peer that either contains an ASA supporting the discovery
      objective indicated by the discovery initiator or caches the
      locator(s) of the ASA(s) supporting the objective.  It sends a
      Discovery Response, as described later.

   Synchronization Initiator:
      An ASA that starts synchronization by sending a request message
      referring to a specific synchronization objective.

   Synchronization Responder:
      A peer ASA that responds with the value of a synchronization

   Negotiation Initiator:
      An ASA that starts negotiation by sending a request message
      referring to a specific negotiation objective.

   Negotiation Counterpart:
      A peer with which the negotiation initiator negotiates a specific
      negotiation objective.

   GRASP Instance:
      This refers to an instantiation of a GRASP protocol engine, likely
      including multiple threads or processes as well as dynamic data
      structures such as a discovery cache, running in a given security
      environment on a single device.

   GRASP Core:
      This refers to the code and shared data structures of a GRASP
      instance, which will communicate with individual ASAs via a
      suitable Application Programming Interface (API).

   Interface or GRASP Interface:
      Unless otherwise stated, this refers to a network interface, which
      might be physical or virtual, that a specific instance of GRASP is
      currently using.  A device might have other interfaces that are
      not used by GRASP and which are outside the scope of the Autonomic

2.2.  High-Level Deployment Model

   A GRASP implementation will be part of the Autonomic Networking
   Infrastructure (ANI) in an autonomic node, which must also provide an
   appropriate security environment.  In accordance with [RFC8993], this
   SHOULD be the Autonomic Control Plane (ACP) [RFC8994].  As a result,
   all autonomic nodes in the ACP are able to trust each other.  It is
   expected that GRASP will access the ACP by using a typical socket
   programming interface, and the ACP will make available only network
   interfaces within the Autonomic Network.  If there is no ACP, the
   considerations described in Section 2.5.1 apply.

   There will also be one or more Autonomic Service Agents (ASAs).  In
   the minimal case of a single-purpose device, these components might
   be fully integrated with GRASP and the ACP.  A more common model is
   expected to be a multipurpose device capable of containing several
   ASAs, such as a router or large switch.  In this case it is expected
   that the ACP, GRASP and the ASAs will be implemented as separate
   processes, which are able to support asynchronous and simultaneous
   operations, for example by multithreading.

   In some scenarios, a limited negotiation model might be deployed
   based on a limited trust relationship such as that between two
   administrative domains.  ASAs might then exchange limited information
   and negotiate some particular configurations.

   GRASP is explicitly designed to operate within a single addressing
   realm.  Its discovery and flooding mechanisms do not support
   autonomic operations that cross any form of address translator or
   upper-layer proxy.

   A suitable Application Programming Interface (API) will be needed
   between GRASP and the ASAs.  In some implementations, ASAs would run
   in user space with a GRASP library providing the API, and this
   library would in turn communicate via system calls with core GRASP
   functions.  Details of the API are out of scope for the present
   document.  For further details of possible deployment models, see

   An instance of GRASP must be aware of the network interfaces it will
   use, and of the appropriate global-scope and link-local addresses.
   In the presence of the ACP, such information will be available from
   the adjacency table discussed in [RFC8993].  In other cases, GRASP
   must determine such information for itself.  Details depend on the
   device and operating system.  In the rest of this document, the terms
   'interfaces' or 'GRASP interfaces' refers only to the set of network
   interfaces that a specific instance of GRASP is currently using.

   Because GRASP needs to work with very high reliability, especially
   during bootstrapping and during fault conditions, it is essential
   that every implementation continues to operate in adverse conditions.
   For example, discovery failures, or any kind of socket exception at
   any time, must not cause irrecoverable failures in GRASP itself, and
   must return suitable error codes through the API so that ASAs can
   also recover.

   GRASP must not depend upon nonvolatile data storage.  All runtime
   error conditions, and events such as address renumbering, network
   interface failures, and CPU sleep/wake cycles, must be handled in
   such a way that GRASP will still operate correctly and securely
   afterwards (Section 2.5.1).

   An autonomic node will normally run a single instance of GRASP, which
   is used by multiple ASAs.  Possible exceptions are mentioned below.

2.3.  High-Level Design

   This section describes the behavior model and general design of
   GRASP, supporting discovery, synchronization, and negotiation, to act
   as a platform for different technical objectives.

   A generic platform:
      The protocol design is generic and independent of the
      synchronization or negotiation contents.  The technical contents
      will vary according to the various technical objectives and the
      different pairs of counterparts.

   Multiple instances:
      Normally, a single main instance of the GRASP protocol engine will
      exist in an autonomic node, and each ASA will run as an
      independent asynchronous process.  However, scenarios where
      multiple instances of GRASP run in a single node, perhaps with
      different security properties, are possible (Section 2.5.2).  In
      this case, each instance MUST listen independently for GRASP link-
      local multicasts, and all instances MUST be woken by each such
      multicast in order for discovery and flooding to work correctly.

   Security infrastructure:
      As noted above, the protocol itself has no built-in security
      functionality and relies on a separate secure infrastructure.

   Discovery, synchronization, and negotiation are designed together:
      The discovery method and the synchronization and negotiation
      methods are designed in the same way and can be combined when this
      is useful, allowing a rapid mode of operation described in
      Section 2.5.4.  These processes can also be performed
      independently when appropriate.

         Thus, for some objectives, especially those concerned with
         application-layer services, another discovery mechanism such as
         DNS-based Service Discovery [RFC7558] MAY be used.  The choice
         is left to the designers of individual ASAs.

   A uniform pattern for technical objectives:
      The synchronization and negotiation objectives are defined
      according to a uniform pattern.  The values that they contain
      could be carried either in a simple binary format or in a complex
      object format.  The basic protocol design uses the Concise Binary
      Object Representation (CBOR) [RFC8949], which is readily
      extensible for unknown, future requirements.

   A flexible model for synchronization:
      GRASP supports synchronization between two nodes, which could be
      used repeatedly to perform synchronization among a small number of
      nodes.  It also supports an unsolicited flooding mode when large
      groups of nodes, possibly including all autonomic nodes, need data
      for the same technical objective.

         There may be some network parameters for which a more
         traditional flooding mechanism such as the Distributed Node
         Consensus Protocol (DNCP) [RFC7787] is considered more
         appropriate.  GRASP can coexist with DNCP.

   A simple initiator/responder model for negotiation:
      Multiparty negotiations are very complicated to model and cannot
      readily be guaranteed to converge.  GRASP uses a simple bilateral
      model and can support multiparty negotiations by indirect steps.

   Organizing of synchronization or negotiation content:
      The technical content transmitted by GRASP will be organized
      according to the relevant function or service.  The objectives for
      different functions or services are kept separate because they may
      be negotiated or synchronized with different counterparts or have
      different response times.  Thus a normal arrangement is a single
      ASA managing a small set of closely related objectives, with a
      version of that ASA in each relevant autonomic node.  Further
      discussion of this aspect is out of scope for the current

   Requests and responses in negotiation procedures:
      The initiator can negotiate a specific negotiation objective with
      relevant counterpart ASAs.  It can request relevant information
      from a counterpart so that it can coordinate its local
      configuration.  It can request the counterpart to make a matching
      configuration.  It can request simulation or forecast results by
      sending some dry-run conditions.

      Beyond the traditional yes/no answer, the responder can reply with
      a suggested alternative value for the objective concerned.  This
      would start a bidirectional negotiation ending in a compromise
      between the two ASAs.

   Convergence of negotiation procedures:
      To enable convergence when a responder suggests a new value or
      condition in a negotiation step reply, it should be as close as
      possible to the original request or previous suggestion.  The
      suggested value of later negotiation steps should be chosen
      between the suggested values from the previous two steps.  GRASP
      provides mechanisms to guarantee convergence (or failure) in a
      small number of steps, namely a timeout and a maximum number of

      GRASP intentionally does not have a version number, and it can be
      extended by adding new message types and options.  The Invalid
      message (M_INVALID) will be used to signal that an implementation
      does not recognize a message or option sent by another
      implementation.  In normal use, new semantics will be added by
      defining new synchronization or negotiation objectives.

2.4.  Quick Operating Overview

   An instance of GRASP is expected to run as a separate core module,
   providing an API (such as [RFC8991]) to interface to various ASAs.
   These ASAs may operate without special privilege, unless they need it
   for other reasons (such as configuring IP addresses or manipulating
   routing tables).

   The GRASP mechanisms used by the ASA are built around GRASP
   objectives defined as data structures containing administrative
   information such as the objective's unique name and its current
   value.  The format and size of the value is not restricted by the
   protocol, except that it must be possible to serialize it for
   transmission in CBOR, which is no restriction at all in practice.

   GRASP provides the following mechanisms:

   *  A discovery mechanism (M_DISCOVERY, M_RESPONSE) by which an ASA
      can discover other ASAs supporting a given objective.

   *  A negotiation request mechanism (M_REQ_NEG) by which an ASA can
      start negotiation of an objective with a counterpart ASA.  Once a
      negotiation has started, the process is symmetrical, and there is
      a negotiation step message (M_NEGOTIATE) for each ASA to use in
      turn.  Two other functions support negotiating steps (M_WAIT,

   *  A synchronization mechanism (M_REQ_SYN) by which an ASA can
      request the current value of an objective from a counterpart ASA.
      With this, there is a corresponding response function (M_SYNCH)
      for an ASA that wishes to respond to synchronization requests.

   *  A flood mechanism (M_FLOOD) by which an ASA can cause the current
      value of an objective to be flooded throughout the Autonomic
      Network so that any ASA can receive it.  One application of this
      is to act as an announcement, avoiding the need for discovery of a
      widely applicable objective.

   Some example messages and simple message flows are provided in
   Appendix A.

2.5.  GRASP Basic Properties and Mechanisms

2.5.1.  Required External Security Mechanism

   GRASP does not specify transport security because it is meant to be
   adapted to different environments.  Every solution adopting GRASP
   MUST specify a security and transport substrate used by GRASP in that

   The substrate MUST enforce sending and receiving GRASP messages only
   between members of a mutually trusted group running GRASP.  Each
   group member is an instance of GRASP.  The group members are nodes of
   a connected graph.  The group and graph are created by the security
   and transport substrate and are called the GRASP domain.  The
   substrate must support unicast messages between any group members and
   (link-local) multicast messages between adjacent group members.  It
   must deny messages between group members and non-group members.  With
   this model, security is provided by enforcing group membership, but
   any member of the trusted group can attack the entire network until

   Substrates MUST use cryptographic member authentication and message
   integrity for GRASP messages.  This can be end to end or hop by hop
   across the domain.  The security and transport substrate MUST provide
   mechanisms to remove untrusted members from the group.

   If the substrate does not mandate and enforce GRASP message
   encryption, then any service using GRASP in such a solution MUST
   provide protection and encryption for message elements whose exposure
   could constitute an attack vector.

   The security and transport substrate for GRASP in the ANI is the ACP.
   Unless otherwise noted, we assume this security and transport
   substrate in the remainder of this document.  The ACP does mandate
   the use of encryption; therefore, GRASP in the ANI can rely on GRASP
   messages being encrypted.  The GRASP domain is the ACP: all nodes in
   an autonomic domain connected by encrypted virtual links formed by
   the ACP.  The ACP uses hop-by-hop security (authentication and
   encryption) of messages.  Removal of nodes relies on standard PKI
   certificate revocation or expiry of sufficiently short-lived
   certificates.  Refer to [RFC8994] for more details.

   As mentioned in Section 2.3, some GRASP operations might be performed
   across an administrative domain boundary by mutual agreement, without
   the benefit of an ACP.  Such operations MUST be confined to a
   separate instance of GRASP with its own copy of all GRASP data
   structures running across a separate GRASP domain with a security and
   transport substrate.  In the most simple case, each point-to-point
   interdomain GRASP peering could be a separate domain, and the
   security and transport substrate could be built using transport or
   network-layer security protocols.  This is subject to future

   An exception to the requirements for the security and transport
   substrate exists for highly constrained subsets of GRASP meant to
   support the establishment of a security and transport substrate,
   described in the following section.

2.5.2.  Discovery Unsolicited Link-Local (DULL) GRASP

   Some services may need to use insecure GRASP discovery, response, and
   flood messages without being able to use preexisting security
   associations, for example, as part of discovery for establishing
   security associations such as a security substrate for GRASP.

   Such operations being intrinsically insecure, they need to be
   confined to link-local use to minimize the risk of malicious actions.
   Possible examples include discovery of candidate ACP neighbors
   [RFC8994], discovery of bootstrap proxies [RFC8995], or perhaps
   initialization services in networks using GRASP without being fully
   autonomic (e.g., no ACP).  Such usage MUST be limited to link-local
   operations on a single interface and MUST be confined to a separate
   insecure instance of GRASP with its own copy of all GRASP data
   structures.  This instance is nicknamed DULL -- Discovery Unsolicited

   The detailed rules for the DULL instance of GRASP are as follows:

   *  An initiator MAY send Discovery or Flood Synchronization link-
      local multicast messages that MUST have a loop count of 1, to
      prevent off-link operations.  Other unsolicited GRASP message
      types MUST NOT be sent.

   *  A responder MUST silently discard any message whose loop count is
      not 1.

   *  A responder MUST silently discard any message referring to a GRASP
      objective that is not directly part of a service that requires
      this insecure mode.

   *  A responder MUST NOT relay any multicast messages.

   *  A Discovery Response MUST indicate a link-local address.

   *  A Discovery Response MUST NOT include a Divert option.

   *  A node MUST silently discard any message whose source address is
      not link-local.

   To minimize traffic possibly observed by third parties, GRASP traffic
   SHOULD be minimized by using only Flood Synchronization to announce
   objectives and their associated locators, rather than by using
   Discovery and Discovery Response messages.  Further details are out
   of scope for this document.

2.5.3.  Transport Layer Usage

   All GRASP messages, after they are serialized as a CBOR byte string,
   are transmitted as such directly over the transport protocol in use.
   The transport protocol(s) for a GRASP domain are specified by the
   security and transport substrate as introduced in Section 2.5.1.

   GRASP discovery and flooding messages are designed for GRASP domain-
   wide flooding through hop-by-hop link-local multicast forwarding
   between adjacent GRASP nodes.  The GRASP security and transport
   substrate needs to specify how these link-local multicasts are
   transported.  This can be unreliable transport (UDP) but it SHOULD be
   reliable transport (e.g., TCP).

   If the substrate specifies an unreliable transport such as UDP for
   discovery and flooding messages, then it MUST NOT use IP
   fragmentation because of its loss characteristic, especially in
   multi-hop flooding.  GRASP MUST then enforce at the user API level a
   limit to the size of discovery and flooding messages, so that no
   fragmentation can occur.  For IPv6 transport, this means that the
   size of those messages' IPv6 packets must be at most 1280 bytes
   (unless there is a known larger minimum link MTU across the whole
   GRASP domain).

   All other GRASP messages are unicast between group members of the
   GRASP domain.  These MUST use a reliable transport protocol because
   GRASP itself does not provide for error detection, retransmission, or
   flow control.  Unless otherwise specified by the security and
   transport substrate, TCP MUST be used.

   The security and transport substrate for GRASP in the ANI is the ACP.
   Unless otherwise noted, we assume this security and transport
   substrate in the remainder of this document when describing GRASP's
   message transport.  In the ACP, TCP is used for GRASP unicast
   messages.  GRASP discovery and flooding messages also use TCP: these
   link-local messages are forwarded by replicating them to all adjacent
   GRASP nodes on the link via TCP connections to those adjacent GRASP
   nodes.  Because of this, GRASP in the ANI has no limitations on the
   size of discovery and flooding messages with respect to fragmentation
   issues.  While the ACP is being built using a DULL instance of GRASP,
   native UDP multicast is used to discover ACP/GRASP neighbors on

   For link-local UDP multicast, GRASP listens to the well-known GRASP
   Listen Port (Section 2.6).  Transport connections for discovery and
   flooding on relay nodes must terminate in GRASP instances (e.g.,
   GRASP ASAs) so that link-local multicast, hop-by-hop flooding of
   M_DISCOVERY and M_FLOOD messages and hop-by-hop forwarding of
   M_RESPONSE responses and caching of those responses along the path
   work correctly.

   Unicast transport connections used for synchronization and
   negotiation can terminate directly in ASAs that implement objectives;
   therefore, this traffic does not need to pass through GRASP
   instances.  For this, the ASA listens on its own dynamically assigned
   ports, which are communicated to its peers during discovery.
   Alternatively, the GRASP instance can also terminate the unicast
   transport connections and pass the traffic from/to the ASA if that is
   preferable in some implementations (e.g., to better decouple ASAs
   from network connections).

2.5.4.  Discovery Mechanism and Procedures  Separated Discovery and Negotiation Mechanisms

   Although discovery and negotiation or synchronization are defined
   together in GRASP, they are separate mechanisms.  The discovery
   process could run independently from the negotiation or
   synchronization process.  Upon receiving a Discovery message
   (Section 2.8.4), the recipient node should return a Discovery
   Response message in which it either indicates itself as a discovery
   responder or diverts the initiator towards another more suitable ASA.
   However, this response may be delayed if the recipient needs to relay
   the Discovery message onward, as described in Section

   The discovery action (M_DISCOVERY) will normally be followed by a
   negotiation (M_REQ_NEG) or synchronization (M_REQ_SYN) action.  The
   discovery results could be utilized by the negotiation protocol to
   decide which ASA the initiator will negotiate with.

   The initiator of a discovery action for a given objective need not be
   capable of responding to that objective as a negotiation counterpart,
   as a synchronization responder, or as source for flooding.  For
   example, an ASA might perform discovery even if it only wishes to act
   as a synchronization initiator or negotiation initiator.  Such an ASA
   does not itself need to respond to Discovery messages.

   It is also entirely possible to use GRASP discovery without any
   subsequent negotiation or synchronization action.  In this case, the
   discovered objective is simply used as a name during the discovery
   process, and any subsequent operations between the peers are outside
   the scope of GRASP.  Discovery Overview

   A complete discovery process will start with a multicast Discovery
   message (M_DISCOVERY) on the local link.  On-link neighbors
   supporting the discovery objective will respond directly with
   Discovery Response (M_RESPONSE) messages.  A neighbor with multiple
   interfaces may respond with a cached Discovery Response.  If it has
   no cached response, it will relay the Discovery message on its other
   GRASP interfaces.  If a node receiving the relayed Discovery message
   supports the discovery objective, it will respond to the relayed
   Discovery message.  If it has a cached response, it will respond with
   that.  If not, it will repeat the discovery process, which thereby
   becomes iterative.  The loop count and timeout will ensure that the
   process ends.  Further details are given in Section

   A Discovery message MAY be sent unicast to a peer node, which SHOULD
   then proceed exactly as if the message had been multicast, except
   that when TCP is used, the response will be on the same socket as the
   query.  However, this mode does not guarantee successful discovery in
   the general case.  Discovery Procedures

   Discovery starts as an on-link operation.  The Divert option can tell
   the discovery initiator to contact an off-link ASA for that discovery
   objective.  If the security and transport substrate of the GRASP
   domain (see Section 2.5.3) uses UDP link-local multicast, then the
   discovery initiator sends these to the ALL_GRASP_NEIGHBORS link-local
   multicast address (Section 2.6), and all GRASP nodes need to listen
   to this address to act as discovery responders.  Because this port is
   unique in a device, this is a function of the GRASP instance and not
   of an individual ASA.  As a result, each ASA will need to register
   the objectives that it supports with the local GRASP instance.

   If an ASA in a neighbor device supports the requested discovery
   objective, the device SHOULD respond to the link-local multicast with
   a unicast Discovery Response message (Section 2.8.5) with locator
   option(s) (Section 2.9.5) unless it is temporarily unavailable.
   Otherwise, if the neighbor has cached information about an ASA that
   supports the requested discovery objective (usually because it
   discovered the same objective before), it SHOULD respond with a
   Discovery Response message with a Divert option pointing to the
   appropriate discovery responder.  However, it SHOULD NOT respond with
   a cached response on an interface if it learned that information from
   the same interface because the peer in question will answer directly
   if still operational.

   If a device has no information about the requested discovery
   objective and is not acting as a discovery relay (see
   Section, it MUST silently discard the Discovery message.

   The discovery initiator MUST set a reasonable timeout on the
   discovery process.  A suggested value is 100 milliseconds multiplied
   by the loop count embedded in the objective.

   If no Discovery Response is received within the timeout, the
   Discovery message MAY be repeated with a newly generated Session ID
   (Section 2.7).  An exponential backoff SHOULD be used for subsequent
   repetitions to limit the load during busy periods.  The details of
   the backoff algorithm will depend on the use case for the objective
   concerned but MUST be consistent with the recommendations in
   [RFC8085] for low data-volume multicast.  Frequent repetition might
   be symptomatic of a denial-of-service attack.

   After a GRASP device successfully discovers a locator for a discovery
   responder supporting a specific objective, it SHOULD cache this
   information, including the interface index [RFC3493] via which it was
   discovered.  This cache record MAY be used for future negotiation or
   synchronization, and the locator SHOULD be passed on when appropriate
   as a Divert option to another discovery initiator.

   The cache mechanism MUST include a lifetime for each entry.  The
   lifetime is derived from a time-to-live (ttl) parameter in each
   Discovery Response message.  Cached entries MUST be ignored or
   deleted after their lifetime expires.  In some environments,
   unplanned address renumbering might occur.  In such cases, the
   lifetime SHOULD be short compared to the typical address lifetime.
   The discovery mechanism needs to track the node's current address to
   ensure that Discovery Responses always indicate the correct address.

   If multiple discovery responders are found for the same objective,
   they SHOULD all be cached unless this creates a resource shortage.
   The method of choosing between multiple responders is an
   implementation choice.  This choice MUST be available to each ASA,
   but the GRASP implementation SHOULD provide a default choice.

   Because discovery responders will be cached in a finite cache, they
   might be deleted at any time.  In this case, discovery will need to
   be repeated.  If an ASA exits for any reason, its locator might still
   be cached for some time, and attempts to connect to it will fail.
   ASAs need to be robust in these circumstances.  Discovery Relaying

   A GRASP instance with multiple link-layer interfaces (typically
   running in a router) MUST support discovery on all GRASP interfaces.
   We refer to this as a 'relaying instance'.

   DULL instances (Section 2.5.2) are always single-interface instances
   and therefore MUST NOT perform discovery relaying.

   If a relaying instance receives a Discovery message on a given
   interface for a specific objective that it does not support and for
   which it has not previously cached a discovery responder, it MUST
   relay the query by reissuing a new Discovery message as a link-local
   multicast on its other GRASP interfaces.

   The relayed Discovery message MUST have the same Session ID and
   'initiator' field as the incoming message (see Section 2.8.4).  The
   IP address in the 'initiator' field is only used to disambiguate the
   Session ID and is never used to address Response packets.  Response
   packets are sent back to the relaying instance, not the original

   The M_DISCOVERY message does not encode the transport address of the
   originator or relay.  Response packets must therefore be sent to the
   transport-layer address of the connection on which the M_DISCOVERY
   message was received.  If the M_DISCOVERY was relayed via a reliable
   hop-by-hop transport connection, the response is simply sent back via
   the same connection.

   If the M_DISCOVERY was relayed via link-local (e.g., UDP) multicast,
   the response is sent back via a reliable hop-by-hop transport
   connection with the same port number as the source port of the link-
   local multicast.  Therefore, if link-local multicast is used and
   M_RESPONSE messages are required (which is the case in almost all
   GRASP instances except for the limited use of DULL instances in the
   ANI), GRASP needs to be able to bind to one port number on UDP from
   which to originate the link-local multicast M_DISCOVERY messages and
   the same port number on the reliable hop-by-hop transport (e.g., TCP
   by default) to be able to respond to transport connections from
   responders that want to send M_RESPONSE messages back.  Note that
   this port does not need to be the GRASP_LISTEN_PORT.

   The relaying instance MUST decrement the loop count within the
   objective, and MUST NOT relay the Discovery message if the result is
   zero.  Also, it MUST limit the total rate at which it relays
   Discovery messages to a reasonable value in order to mitigate
   possible denial-of-service attacks.  For example, the rate limit
   could be set to a small multiple of the observed rate of Discovery
   messages during normal operation.  The relaying instance MUST cache
   the Session ID value and initiator address of each relayed Discovery
   message until any Discovery Responses have arrived or the discovery
   process has timed out.  To prevent loops, it MUST NOT relay a
   Discovery message that carries a given cached Session ID and
   initiator address more than once.  These precautions avoid discovery
   loops and mitigate potential overload.

   Since the relay device is unaware of the timeout set by the original
   initiator, it SHOULD set a suitable timeout for the relayed Discovery
   message.  A suggested value is 100 milliseconds multiplied by the
   remaining loop count.

   The discovery results received by the relaying instance MUST in turn
   be sent as a Discovery Response message to the Discovery message that
   caused the relay action.  Rapid Mode (Discovery with Negotiation or Synchronization)

   A Discovery message MAY include an objective option.  This allows a
   rapid mode of negotiation (Section or synchronization
   (Section  Rapid mode is currently limited to a single
   objective for simplicity of design and implementation.  A possible
   future extension is to allow multiple objectives in rapid mode for
   greater efficiency.

2.5.5.  Negotiation Procedures

   A negotiation initiator opens a transport connection to a counterpart
   ASA using the address, protocol, and port obtained during discovery.
   It then sends a negotiation request (using M_REQ_NEG) to the
   counterpart, including a specific negotiation objective.  It may
   request the negotiation counterpart to make a specific configuration.
   Alternatively, it may request a certain simulation or forecast result
   by sending a dry-run configuration.  The details, including the
   distinction between a dry run and a live configuration change, will
   be defined separately for each type of negotiation objective.  Any
   state associated with a dry-run operation, such as temporarily
   reserving a resource for subsequent use in a live run, is entirely a
   matter for the designer of the ASA concerned.

   Each negotiation session as a whole is subject to a timeout (default
   GRASP_DEF_TIMEOUT milliseconds, Section 2.6), initialized when the
   request is sent (see Section 2.8.6).  If no reply message of any kind
   is received within the timeout, the negotiation request MAY be
   repeated with a newly generated Session ID (Section 2.7).  An
   exponential backoff SHOULD be used for subsequent repetitions.  The
   details of the backoff algorithm will depend on the use case for the
   objective concerned.

   If the counterpart can immediately apply the requested configuration,
   it will give an immediate positive (O_ACCEPT) answer using the
   Negotiation End (M_END) message.  This will end the negotiation phase
   immediately.  Otherwise, it will negotiate (using M_NEGOTIATE).  It
   will reply with a proposed alternative configuration that it can
   apply (typically, a configuration that uses fewer resources than
   requested by the negotiation initiator).  This will start a
   bidirectional negotiation using the Negotiate (M_NEGOTIATE) message
   to reach a compromise between the two ASAs.

   The negotiation procedure is ended when one of the negotiation peers
   sends a Negotiation End (M_END) message, which contains an Accept
   (O_ACCEPT) or Decline (O_DECLINE) option and does not need a response
   from the negotiation peer.  Negotiation may also end in failure
   (equivalent to a decline) if a timeout is exceeded or a loop count is
   exceeded.  When the procedure ends for whatever reason, the transport
   connection SHOULD be closed.  A transport session failure is treated
   as a negotiation failure.

   A negotiation procedure concerns one objective and one counterpart.
   Both the initiator and the counterpart may take part in simultaneous
   negotiations with various other ASAs or in simultaneous negotiations
   about different objectives.  Thus, GRASP is expected to be used in a
   multithreaded mode or its logical equivalent.  Certain negotiation
   objectives may have restrictions on multithreading, for example to
   avoid over-allocating resources.

   Some configuration actions, for example, wavelength switching in
   optical networks, might take considerable time to execute.  The ASA
   concerned needs to allow for this by design, but GRASP does allow for
   a peer to insert latency in a negotiation process if necessary
   (Section 2.8.9, M_WAIT).  Rapid Mode (Discovery/Negotiation Linkage)

   A Discovery message MAY include a Negotiation Objective option.  In
   this case, it is as if the initiator sent the sequence M_DISCOVERY
   immediately followed by M_REQ_NEG.  This has implications for the
   construction of the GRASP core, as it must carefully pass the
   contents of the Negotiation Objective option to the ASA so that it
   may evaluate the objective directly.  When a Negotiation Objective
   option is present, the ASA replies with an M_NEGOTIATE message (or
   M_END with O_ACCEPT if it is immediately satisfied with the proposal)
   rather than with an M_RESPONSE.  However, if the recipient node does
   not support rapid mode, discovery will continue normally.

   It is possible that a Discovery Response will arrive from a responder
   that does not support rapid mode before such a Negotiation message
   arrives.  In this case, rapid mode will not occur.

   This rapid mode could reduce the interactions between nodes so that a
   higher efficiency could be achieved.  However, a network in which
   some nodes support rapid mode and others do not will have complex
   timing-dependent behaviors.  Therefore, the rapid negotiation
   function SHOULD be disabled by default.

2.5.6.  Synchronization and Flooding Procedures  Unicast Synchronization

   A synchronization initiator opens a transport connection to a
   counterpart ASA using the address, protocol, and port obtained during
   discovery.  It then sends a Request Synchronization message
   (M_REQ_SYN, Section 2.8.6) to the counterpart, including a specific
   synchronization objective.  The counterpart responds with a
   Synchronization message (M_SYNCH, Section 2.8.10) containing the
   current value of the requested synchronization objective.  No further
   messages are needed, and the transport connection SHOULD be closed.
   A transport session failure is treated as a synchronization failure.

   If no reply message of any kind is received within a given timeout
   (default GRASP_DEF_TIMEOUT milliseconds, Section 2.6), the
   synchronization request MAY be repeated with a newly generated
   Session ID (Section 2.7).  An exponential backoff SHOULD be used for
   subsequent repetitions.  The details of the backoff algorithm will
   depend on the use case for the objective concerned.  Flooding

   In the case just described, the message exchange is unicast and
   concerns only one synchronization objective.  For large groups of
   nodes requiring the same data, synchronization flooding is available.
   For this, a flooding initiator MAY send an unsolicited Flood
   Synchronization message (Section 2.8.11) containing one or more
   Synchronization Objective option(s), if and only if the specification
   of those objectives permits it.  This is sent as a multicast message
   to the ALL_GRASP_NEIGHBORS multicast address (Section 2.6).

   Receiving flood multicasts is a function of the GRASP core, as in the
   case of discovery multicasts (Section

   To ensure that flooding does not result in a loop, the originator of
   the Flood Synchronization message MUST set the loop count in the
   objectives to a suitable value (the default is GRASP_DEF_LOOPCT).
   Also, a suitable mechanism is needed to avoid excessive multicast
   traffic.  This mechanism MUST be defined as part of the specification
   of the synchronization objective(s) concerned.  It might be a simple
   rate limit or a more complex mechanism such as the Trickle algorithm

   A GRASP device with multiple link-layer interfaces (typically a
   router) MUST support synchronization flooding on all GRASP
   interfaces.  If it receives a multicast Flood Synchronization message
   on a given interface, it MUST relay it by reissuing a Flood
   Synchronization message as a link-local multicast on its other GRASP
   interfaces.  The relayed message MUST have the same Session ID as the
   incoming message and MUST be tagged with the IP address of its
   original initiator.

   Link-layer flooding is supported by GRASP by setting the loop count
   to 1 and sending with a link-local source address.  Floods with link-
   local source addresses and a loop count other than 1 are invalid, and
   such messages MUST be discarded.

   The relaying device MUST decrement the loop count within the first
   objective and MUST NOT relay the Flood Synchronization message if the
   result is zero.  Also, it MUST limit the total rate at which it
   relays Flood Synchronization messages to a reasonable value, in order
   to mitigate possible denial-of-service attacks.  For example, the
   rate limit could be set to a small multiple of the observed rate of
   flood messages during normal operation.  The relaying device MUST
   cache the Session ID value and initiator address of each relayed
   Flood Synchronization message for a time not less than twice
   GRASP_DEF_TIMEOUT milliseconds.  To prevent loops, it MUST NOT relay
   a Flood Synchronization message that carries a given cached Session
   ID and initiator address more than once.  These precautions avoid
   synchronization loops and mitigate potential overload.

   Note that this mechanism is unreliable in the case of sleeping nodes,
   or new nodes that join the network, or nodes that rejoin the network
   after a fault.  An ASA that initiates a flood SHOULD repeat the flood
   at a suitable frequency, which MUST be consistent with the
   recommendations in [RFC8085] for low data-volume multicast.  The ASA
   SHOULD also act as a synchronization responder for the objective(s)
   concerned.  Thus nodes that require an objective subject to flooding
   can either wait for the next flood or request unicast synchronization
   for that objective.

   The multicast messages for synchronization flooding are subject to
   the security rules in Section 2.5.1.  In practice, this means that
   they MUST NOT be transmitted and MUST be ignored on receipt unless
   there is an operational ACP or equivalent strong security in place.
   However, because of the security weakness of link-local multicast
   (Section 3), synchronization objectives that are flooded SHOULD NOT
   contain unencrypted private information and SHOULD be validated by
   the recipient ASA.  Rapid Mode (Discovery/Synchronization Linkage)

   A Discovery message MAY include a Synchronization Objective option.
   In this case, the Discovery message also acts as a Request
   Synchronization message to indicate to the discovery responder that
   it could directly reply to the discovery initiator with a
   Synchronization message (Section 2.8.10) with synchronization data
   for rapid processing, if the discovery target supports the
   corresponding synchronization objective.  The design implications are
   similar to those discussed in Section

   It is possible that a Discovery Response will arrive from a responder
   that does not support rapid mode before such a Synchronization
   message arrives.  In this case, rapid mode will not occur.

   This rapid mode could reduce the interactions between nodes so that a
   higher efficiency could be achieved.  However, a network in which
   some nodes support rapid mode and others do not will have complex
   timing-dependent behaviors.  Therefore, the rapid synchronization
   function SHOULD be configured off by default and MAY be configured on
   or off by Intent.

2.6.  GRASP Constants

      A link-local scope multicast address used by a GRASP-enabled
      device to discover GRASP-enabled neighbor (i.e., on-link) devices.
      All devices that support GRASP are members of this multicast

      *  IPv6 multicast address: ff02::13

      *  IPv4 multicast address:

      A well-known UDP user port that every GRASP-enabled network device
      MUST listen to for link-local multicasts when UDP is used for
      M_DISCOVERY or M_FLOOD messages in the GRASP instance.  This user
      port MAY also be used to listen for TCP or UDP unicast messages in
      a simple implementation of GRASP (Section 2.5.3).

   GRASP_DEF_TIMEOUT (60000 milliseconds)
      The default timeout used to determine that an operation has failed
      to complete.

      The default loop count used to determine that a negotiation has
      failed to complete and to avoid looping messages.

      The default maximum message size in bytes.

2.7.  Session Identifier (Session ID)

   This is an up to 32-bit opaque value used to distinguish multiple
   sessions between the same two devices.  A new Session ID MUST be
   generated by the initiator for every new Discovery, Flood
   Synchronization, or Request message.  All responses and follow-up
   messages in the same discovery, synchronization, or negotiation
   procedure MUST carry the same Session ID.

   The Session ID SHOULD have a very low collision rate locally.  It
   MUST be generated by a pseudorandom number generator (PRNG) using a
   locally generated seed that is unlikely to be used by any other
   device in the same network.  The PRNG SHOULD be cryptographically
   strong [RFC4086].  When allocating a new Session ID, GRASP MUST check
   that the value is not already in use and SHOULD check that it has not
   been used recently by consulting a cache of current and recent
   sessions.  In the unlikely event of a clash, GRASP MUST generate a
   new value.

   However, there is a finite probability that two nodes might generate
   the same Session ID value.  For that reason, when a Session ID is
   communicated via GRASP, the receiving node MUST tag it with the
   initiator's IP address to allow disambiguation.  In the highly
   unlikely event of two peers opening sessions with the same Session ID
   value, this tag will allow the two sessions to be distinguished.
   Multicast GRASP messages and their responses, which may be relayed
   between links, therefore include a field that carries the initiator's
   global IP address.

   There is a highly unlikely race condition in which two peers start
   simultaneous negotiation sessions with each other using the same
   Session ID value.  Depending on various implementation choices, this
   might lead to the two sessions being confused.  See Section 2.8.6 for
   details of how to avoid this.

2.8.  GRASP Messages

2.8.1.  Message Overview

   This section defines the GRASP message format and message types.
   Message types not listed here are reserved for future use.

   The messages currently defined are:

      Discovery and Discovery Response (M_DISCOVERY, M_RESPONSE).

      Request Negotiation, Negotiation, Confirm Waiting, and Negotiation

      Request Synchronization, Synchronization, and Flood
      Synchronization (M_REQ_SYN, M_SYNCH, M_FLOOD).

      No Operation and Invalid (M_NOOP, M_INVALID).

2.8.2.  GRASP Message Format

   GRASP messages share an identical header format and a variable format
   area for options.  GRASP message headers and options are transmitted
   in Concise Binary Object Representation (CBOR) [RFC8949].  In this
   specification, they are described using Concise Data Definition
   Language (CDDL) [RFC8610].  Fragmentary CDDL is used to describe each
   item in this section.  A complete and normative CDDL specification of
   GRASP is given in Section 4, including constants such as message

   Every GRASP message, except the No Operation message, carries a
   Session ID (Section 2.7).  Options are then presented serially.

   In fragmentary CDDL, every GRASP message follows the pattern:

     grasp-message = (message .within message-structure) / noop-message

     message-structure = [MESSAGE_TYPE, session-id, ?initiator,

     MESSAGE_TYPE = 0..255
     session-id = 0..4294967295 ; up to 32 bits
     grasp-option = any

   The MESSAGE_TYPE indicates the type of the message and thus defines
   the expected options.  Any options received that are not consistent
   with the MESSAGE_TYPE SHOULD be silently discarded.

   The No Operation (noop) message is described in Section 2.8.13.

   The various MESSAGE_TYPE values are defined in Section 4.

   All other message elements are described below and formally defined
   in Section 4.

   If an unrecognized MESSAGE_TYPE is received in a unicast message, an
   Invalid message (Section 2.8.12) MAY be returned.  Otherwise, the
   message MAY be logged and MUST be discarded.  If an unrecognized
   MESSAGE_TYPE is received in a multicast message, it MAY be logged and
   MUST be silently discarded.

2.8.3.  Message Size

   GRASP nodes MUST be able to receive unicast messages of at least
   GRASP_DEF_MAX_SIZE bytes.  GRASP nodes MUST NOT send unicast messages
   longer than GRASP_DEF_MAX_SIZE bytes unless a longer size is
   explicitly allowed for the objective concerned.  For example, GRASP
   negotiation itself could be used to agree on a longer message size.

   The message parser used by GRASP should be configured to know about
   the GRASP_DEF_MAX_SIZE, or any larger negotiated message size, so
   that it may defend against overly long messages.

   The maximum size of multicast messages (M_DISCOVERY and M_FLOOD)
   depends on the link-layer technology or the link-adaptation layer in

2.8.4.  Discovery Message

   In fragmentary CDDL, a Discovery message follows the pattern:

     discovery-message = [M_DISCOVERY, session-id, initiator, objective]

   A discovery initiator sends a Discovery message to initiate a
   discovery process for a particular objective option.

   The discovery initiator sends all Discovery messages via UDP to port
   GRASP_LISTEN_PORT at the link-local ALL_GRASP_NEIGHBORS multicast
   address on each link-layer interface in use by GRASP.  It then
   listens for unicast TCP responses on a given port and stores the
   discovery results, including responding discovery objectives and
   corresponding unicast locators.

   The listening port used for TCP MUST be the same port as used for
   sending the Discovery UDP multicast, on a given interface.  In an
   implementation with a single GRASP instance in a node, this MAY be
   GRASP_LISTEN_PORT.  To support multiple instances in the same node,
   the GRASP discovery mechanism in each instance needs to find, for
   each interface, a dynamic port that it can bind to for both sending
   UDP link-local multicast and listening for TCP before initiating any

   The 'initiator' field in the message is a globally unique IP address
   of the initiator for the sole purpose of disambiguating the Session
   ID in other nodes.  If for some reason the initiator does not have a
   globally unique IP address, it MUST use a link-local address that is
   highly likely to be unique for this purpose, for example, using
   [RFC7217].  Determination of a node's globally unique IP address is
   implementation dependent.

   A Discovery message MUST include exactly one of the following:

   *  A Discovery Objective option (Section 2.10.1).  Its loop count
      MUST be set to a suitable value to prevent discovery loops
      (default value is GRASP_DEF_LOOPCT).  If the discovery initiator
      requires only on-link responses, the loop count MUST be set to 1.

   *  A Negotiation Objective option (Section 2.10.1).  This is used
      both for the purpose of discovery and to indicate to the discovery
      target that it MAY directly reply to the discovery initiator with
      a Negotiation message for rapid processing, if it could act as the
      corresponding negotiation counterpart.  The sender of such a
      Discovery message MUST initialize a negotiation timer and loop
      count in the same way as a Request Negotiation message
      (Section 2.8.6).

   *  A Synchronization Objective option (Section 2.10.1).  This is used
      both for the purpose of discovery and to indicate to the discovery
      target that it MAY directly reply to the discovery initiator with
      a Synchronization message for rapid processing, if it could act as
      the corresponding synchronization counterpart.  Its loop count
      MUST be set to a suitable value to prevent discovery loops
      (default value is GRASP_DEF_LOOPCT).

   As mentioned in Section, a Discovery message MAY be sent
   unicast to a peer node, which SHOULD then proceed exactly as if the
   message had been multicast.

2.8.5.  Discovery Response Message

   In fragmentary CDDL, a Discovery Response message follows the

     response-message = [M_RESPONSE, session-id, initiator, ttl,
                         (+locator-option // divert-option), ?objective]

     ttl = 0..4294967295 ; in milliseconds

   A node that receives a Discovery message SHOULD send a Discovery
   Response message if and only if it can respond to the discovery.

      It MUST contain the same Session ID and initiator as the Discovery

      It MUST contain a time-to-live (ttl) for the validity of the
      response, given as a positive integer value in milliseconds.  Zero
      implies a value significantly greater than GRASP_DEF_TIMEOUT
      milliseconds (Section 2.6).  A suggested value is ten times that

      It MAY include a copy of the discovery objective from the
      Discovery message.

   It is sent to the sender of the Discovery message via TCP at the port
   used to send the Discovery message (as explained in Section 2.8.4).
   In the case of a relayed Discovery message, the Discovery Response is
   thus sent to the relay, not the original initiator.

   In all cases, the transport session SHOULD be closed after sending
   the Discovery Response.  A transport session failure is treated as no

   If the responding node supports the discovery objective of the
   discovery, it MUST include at least one kind of locator option
   (Section 2.9.5) to indicate its own location.  A sequence of multiple
   kinds of locator options (e.g., IP address option and FQDN option) is
   also valid.

   If the responding node itself does not support the discovery
   objective, but it knows the locator of the discovery objective, then
   it SHOULD respond to the Discovery message with a Divert option
   (Section 2.9.2) embedding a locator option or a combination of
   multiple kinds of locator options that indicate the locator(s) of the
   discovery objective.

   More details on the processing of Discovery Responses are given in
   Section 2.5.4.

2.8.6.  Request Messages

   In fragmentary CDDL, Request Negotiation and Request Synchronization
   messages follow the patterns:

   request-negotiation-message = [M_REQ_NEG, session-id, objective]

   request-synchronization-message = [M_REQ_SYN, session-id, objective]

   A negotiation or synchronization requesting node sends the
   appropriate Request message to the unicast address of the negotiation
   or synchronization counterpart, using the appropriate protocol and
   port numbers (selected from the discovery result).  If the discovery
   result is an FQDN, it will be resolved first.

   A Request message MUST include the relevant objective option.  In the
   case of Request Negotiation, the objective option MUST include the
   requested value.

   When an initiator sends a Request Negotiation message, it MUST
   initialize a negotiation timer for the new negotiation thread.  The
   default is GRASP_DEF_TIMEOUT milliseconds.  Unless this timeout is
   modified by a Confirm Waiting message (Section 2.8.9), the initiator
   will consider that the negotiation has failed when the timer expires.

   Similarly, when an initiator sends a Request Synchronization, it
   SHOULD initialize a synchronization timer.  The default is
   GRASP_DEF_TIMEOUT milliseconds.  The initiator will consider that
   synchronization has failed if there is no response before the timer

   When an initiator sends a Request message, it MUST initialize the
   loop count of the objective option with a value defined in the
   specification of the option or, if no such value is specified, with

   If a node receives a Request message for an objective for which no
   ASA is currently listening, it MUST immediately close the relevant
   socket to indicate this to the initiator.  This is to avoid
   unnecessary timeouts if, for example, an ASA exits prematurely but
   the GRASP core is listening on its behalf.

   To avoid the highly unlikely race condition in which two nodes
   simultaneously request sessions with each other using the same
   Session ID (Section 2.7), a node MUST verify that the received
   Session ID is not already locally active when it receives a Request
   message.  In case of a clash, it MUST discard the Request message, in
   which case the initiator will detect a timeout.

2.8.7.  Negotiation Message

   In fragmentary CDDL, a Negotiation message follows the pattern:

     negotiation-message = [M_NEGOTIATE, session-id, objective]

   A negotiation counterpart sends a Negotiation message in response to
   a Request Negotiation message, a Negotiation message, or a Discovery
   message in rapid mode.  A negotiation process MAY include multiple

   The Negotiation message MUST include the relevant Negotiation
   Objective option, with its value updated according to progress in the
   negotiation.  The sender MUST decrement the loop count by 1.  If the
   loop count becomes zero, the message MUST NOT be sent.  In this case,
   the negotiation session has failed and will time out.

2.8.8.  Negotiation End Message

   In fragmentary CDDL, a Negotiation End message follows the pattern:

     end-message = [M_END, session-id, accept-option / decline-option]

   A negotiation counterpart sends a Negotiation End message to close
   the negotiation.  It MUST contain either an Accept option or a
   Decline option, defined in Section 2.9.3 and Section 2.9.4.  It could
   be sent either by the requesting node or the responding node.

2.8.9.  Confirm Waiting Message

   In fragmentary CDDL, a Confirm Waiting message follows the pattern:

     wait-message = [M_WAIT, session-id, waiting-time]
     waiting-time = 0..4294967295 ; in milliseconds

   A responding node sends a Confirm Waiting message to ask the
   requesting node to wait for a further negotiation response.  It might
   be that the local process needs more time or that the negotiation
   depends on another triggered negotiation.  This message MUST NOT
   include any other options.  When received, the waiting time value
   overwrites and restarts the current negotiation timer
   (Section 2.8.6).

   The responding node SHOULD send a Negotiation, Negotiation End, or
   another Confirm Waiting message before the negotiation timer expires.
   If not, when the initiator's timer expires, the initiator MUST treat
   the negotiation procedure as failed.

2.8.10.  Synchronization Message

   In fragmentary CDDL, a Synchronization message follows the pattern:

     synch-message = [M_SYNCH, session-id, objective]

   A node that receives a Request Synchronization, or a Discovery
   message in rapid mode, sends back a unicast Synchronization message
   with the synchronization data, in the form of a GRASP option for the
   specific synchronization objective present in the Request

2.8.11.  Flood Synchronization Message

   In fragmentary CDDL, a Flood Synchronization message follows the

     flood-message = [M_FLOOD, session-id, initiator, ttl,
                      +[objective, (locator-option / [])]]

     ttl = 0..4294967295 ; in milliseconds

   A node MAY initiate flooding by sending an unsolicited Flood
   Synchronization message with synchronization data.  This MAY be sent
   to port GRASP_LISTEN_PORT at the link-local ALL_GRASP_NEIGHBORS
   multicast address, in accordance with the rules in Section 2.5.6.

      The initiator address is provided, as described for Discovery
      messages (Section 2.8.4), only to disambiguate the Session ID.

      The message MUST contain a time-to-live (ttl) for the validity of
      the contents, given as a positive integer value in milliseconds.
      There is no default; zero indicates an indefinite lifetime.

      The synchronization data are in the form of GRASP option(s) for
      specific synchronization objective(s).  The loop count(s) MUST be
      set to a suitable value to prevent flood loops (default value is

      Each objective option MAY be followed by a locator option
      (Section 2.9.5) associated with the flooded objective.  In its
      absence, an empty option MUST be included to indicate a null

   A node that receives a Flood Synchronization message MUST cache the
   received objectives for use by local ASAs.  Each cached objective
   MUST be tagged with the locator option sent with it, or with a null
   tag if an empty locator option was sent.  If a subsequent Flood
   Synchronization message carries an objective with the same name and
   the same tag, the corresponding cached copy of the objective MUST be
   overwritten.  If a subsequent Flood Synchronization message carrying
   an objective with same name arrives with a different tag, a new
   cached entry MUST be created.

   Note: the purpose of this mechanism is to allow the recipient of
   flooded values to distinguish between different senders of the same
   objective, and if necessary communicate with them using the locator,
   protocol, and port included in the locator option.  Many objectives
   will not need this mechanism, so they will be flooded with a null

   Cached entries MUST be ignored or deleted after their lifetime

2.8.12.  Invalid Message

   In fragmentary CDDL, an Invalid message follows the pattern:

     invalid-message = [M_INVALID, session-id, ?any]

   This message MAY be sent by an implementation in response to an
   incoming unicast message that it considers invalid.  The Session ID
   value MUST be copied from the incoming message.  The content SHOULD
   be diagnostic information such as a partial copy of the invalid
   message up to the maximum message size.  An M_INVALID message MAY be
   silently ignored by a recipient.  However, it could be used in
   support of extensibility, since it indicates that the remote node
   does not support a new or obsolete message or option.

   An M_INVALID message MUST NOT be sent in response to an M_INVALID

2.8.13.  No Operation Message

   In fragmentary CDDL, a No Operation message follows the pattern:

     noop-message = [M_NOOP]

   This message MAY be sent by an implementation that for practical
   reasons needs to initialize a socket.  It MUST be silently ignored by
   a recipient.

2.9.  GRASP Options

   This section defines the GRASP options for the negotiation and
   synchronization protocol signaling.  Additional options may be
   defined in the future.

2.9.1.  Format of GRASP Options

   GRASP options SHOULD be CBOR arrays that MUST start with an unsigned
   integer identifying the specific option type carried in this option.
   These option types are formally defined in Section 4.

   GRASP options may be defined to include encapsulated GRASP options.

2.9.2.  Divert Option

   The Divert option is used to redirect a GRASP request to another
   node, which may be more appropriate for the intended negotiation or
   synchronization.  It may redirect to an entity that is known as a
   specific negotiation or synchronization counterpart (on-link or off-
   link) or a default gateway.  The Divert option MUST only be
   encapsulated in Discovery Response messages.  If found elsewhere, it
   SHOULD be silently ignored.

   A discovery initiator MAY ignore a Divert option if it only requires
   direct Discovery Responses.

   In fragmentary CDDL, the Divert option follows the pattern:

     divert-option = [O_DIVERT, +locator-option]

   The embedded locator option(s) (Section 2.9.5) point to diverted
   destination target(s) in response to a Discovery message.

2.9.3.  Accept Option

   The Accept option is used to indicate to the negotiation counterpart
   that the proposed negotiation content is accepted.

   The Accept option MUST only be encapsulated in Negotiation End
   messages.  If found elsewhere, it SHOULD be silently ignored.

   In fragmentary CDDL, the Accept option follows the pattern:

     accept-option = [O_ACCEPT]

2.9.4.  Decline Option

   The Decline option is used to indicate to the negotiation counterpart
   the proposed negotiation content is declined and to end the
   negotiation process.

   The Decline option MUST only be encapsulated in Negotiation End
   messages.  If found elsewhere, it SHOULD be silently ignored.

   In fragmentary CDDL, the Decline option follows the pattern:

     decline-option = [O_DECLINE, ?reason]
     reason = text  ; optional UTF-8 error message

   Note: there might be scenarios where an ASA wants to decline the
   proposed value and restart the negotiation process.  In this case, it
   is an implementation choice whether to send a Decline option or to
   continue with a Negotiation message, with an objective option that
   contains a null value or one that contains a new value that might
   achieve convergence.

2.9.5.  Locator Options

   These locator options are used to present reachability information
   for an ASA, a device, or an interface.  They are Locator IPv6 Address
   option, Locator IPv4 Address option, Locator FQDN option, and Locator
   URI option.

   Since ASAs will normally run as independent user programs, locator
   options need to indicate the network-layer locator plus the transport
   protocol and port number for reaching the target.  For this reason,
   the locator options for IP addresses and FQDNs include this
   information explicitly.  In the case of the Locator URI option, this
   information can be encoded in the URI itself.

   Note: It is assumed that all locators used in locator options are in
   scope throughout the GRASP domain.  As stated in Section 2.2, GRASP
   is not intended to work across disjoint addressing or naming realms.  Locator IPv6 Address Option

   In fragmentary CDDL, the Locator IPv6 Address option follows the

     ipv6-locator-option = [O_IPv6_LOCATOR, ipv6-address,
                            transport-proto, port-number]
     ipv6-address = bytes .size 16

     transport-proto = IPPROTO_TCP / IPPROTO_UDP
     IPPROTO_TCP = 6
     IPPROTO_UDP = 17
     port-number = 0..65535

   The content of this option is a binary IPv6 address followed by the
   protocol number and port number to be used.

   Note 1: The IPv6 address MUST normally have global scope.  However,
   during initialization, a link-local address MAY be used for specific
   objectives only (Section 2.5.2).  In this case, the corresponding
   Discovery Response message MUST be sent via the interface to which
   the link-local address applies.

   Note 2: A link-local IPv6 address MUST NOT be used when this option
   is included in a Divert option.

   Note 3: The IPPROTO values are taken from the existing IANA Protocol
   Numbers registry in order to specify TCP or UDP.  If GRASP requires
   future values that are not in that registry, a new registry for
   values outside the range 0..255 will be needed.  Locator IPv4 Address Option

   In fragmentary CDDL, the Locator IPv4 Address option follows the

     ipv4-locator-option = [O_IPv4_LOCATOR, ipv4-address,
                            transport-proto, port-number]
     ipv4-address = bytes .size 4

   The content of this option is a binary IPv4 address followed by the
   protocol number and port number to be used.

   Note: If an operator has internal network address translation for
   IPv4, this option MUST NOT be used within the Divert option.  Locator FQDN Option

   In fragmentary CDDL, the Locator FQDN option follows the pattern:

     fqdn-locator-option = [O_FQDN_LOCATOR, text,
                            transport-proto, port-number]

   The content of this option is the FQDN of the target followed by the
   protocol number and port number to be used.

   Note 1: Any FQDN that might not be valid throughout the network in
   question, such as a Multicast DNS name [RFC6762], MUST NOT be used
   when this option is used within the Divert option.

   Note 2: Normal GRASP operations are not expected to use this option.
   It is intended for special purposes such as discovering external
   services.  Locator URI Option

   In fragmentary CDDL, the Locator URI option follows the pattern:

     uri-locator-option = [O_URI_LOCATOR, text,
                           transport-proto / null, port-number / null]

   The content of this option is the URI of the target followed by the
   protocol number and port number to be used (or by null values if not
   required) [RFC3986].

   Note 1: Any URI which might not be valid throughout the network in
   question, such as one based on a Multicast DNS name [RFC6762], MUST
be used when this option is used within the Divert option.

   Note 2: Normal GRASP operations are not expected to use this option.
   It is intended for special purposes such as discovering external
   services.  Therefore, its use is not further described in this

2.10.  Objective Options

2.10.1.  Format of Objective Options

   An objective option is used to identify objectives for the purposes
   of discovery, negotiation, or synchronization.  All objectives MUST
   be in the following format, described in fragmentary CDDL:

   objective = [objective-name, objective-flags,
                loop-count, ?objective-value]

   objective-name = text
   objective-value = any
   loop-count = 0..255

   All objectives are identified by a unique name that is a UTF-8 string
   [RFC3629], to be compared byte by byte.

   The names of generic objectives MUST NOT include a colon (":") and
   MUST be registered with IANA (Section 5).

   The names of privately defined objectives MUST include at least one
   colon (":").  The string preceding the last colon in the name MUST be
   globally unique and in some way identify the entity or person
   defining the objective.  The following three methods MAY be used to
   create such a globally unique string:

   1.  The unique string is a decimal number representing a registered
       32-bit Private Enterprise Number (PEN) [RFC5612] that uniquely
       identifies the enterprise defining the objective.

   2.  The unique string is a FQDN that uniquely identifies the entity
       or person defining the objective.

   3.  The unique string is an email address that uniquely identifies
       the entity or person defining the objective.

   GRASP treats the objective name as an opaque string.  For example,
   "EX1", "32473:EX1", "example.com:EX1", "example.org:EX1", and
   "user@example.org:EX1" are five different objectives.

   The 'objective-flags' field is described in Section 2.10.2.

   The 'loop-count' field is used for terminating negotiation as
   described in Section 2.8.7.  It is also used for terminating
   discovery as described in Section 2.5.4 and for terminating flooding
   as described in Section  It is placed in the objective
   rather than in the GRASP message format because, as far as the ASA is
   concerned, it is a property of the objective itself.

   The 'objective-value' field expresses the actual value of a
   negotiation or synchronization objective.  Its format is defined in
   the specification of the objective and may be a simple value or a
   data structure of any kind, as long as it can be represented in CBOR.
   It is optional only in a Discovery or Discovery Response message.

2.10.2.  Objective Flags

   An objective may be relevant for discovery only, for discovery and
   negotiation, or for discovery and synchronization.  This is expressed
   in the objective by logical flag bits:

     objective-flags = uint .bits objective-flag
     objective-flag = &(
       F_DISC: 0    ; valid for discovery
       F_NEG: 1     ; valid for negotiation
       F_SYNCH: 2   ; valid for synchronization
       F_NEG_DRY: 3 ; negotiation is a dry run

   These bits are independent and may be combined appropriately, e.g.,
   (F_DISC and F_SYNCH) or (F_DISC and F_NEG) or (F_DISC and F_NEG and

   Note that for a given negotiation session, an objective must be used
   either for negotiation or for dry-run negotiation.  Mixing the two
   modes in a single negotiation is not possible.

2.10.3.  General Considerations for Objective Options

   As mentioned above, objective options MUST be assigned a unique name.
   As long as privately defined objective options obey the rules above,
   this document does not restrict their choice of name, but the entity
   or person concerned SHOULD publish the names in use.

   Names are expressed as UTF-8 strings for convenience in designing
   objective options for localized use.  For generic usage, names
   expressed in the ASCII subset of UTF-8 are RECOMMENDED.  Designers
   planning to use non-ASCII names are strongly advised to consult
   [RFC8264] or its successor to understand the complexities involved.
   Since GRASP compares names byte by byte, all issues of Unicode
   profiling and canonicalization MUST be specified in the design of the
   objective option.

   All objective options MUST respect the CBOR patterns defined above as
   "objective" and MUST replace the 'any' field with a valid CBOR data
   definition for the relevant use case and application.

   An objective option that contains no additional fields beyond its
   'loop-count' can only be a discovery objective and MUST only be used
   in Discovery and Discovery Response messages.

   The Negotiation Objective options contain negotiation objectives,
   which vary according to different functions and/or services.  They
   MUST be carried by Discovery, Request Negotiation, or Negotiation
   messages only.  The negotiation initiator MUST set the initial 'loop-
   count' to a value specified in the specification of the objective or,
   if no such value is specified, to GRASP_DEF_LOOPCT.

   For most scenarios, there should be initial values in the negotiation
   requests.  Consequently, the Negotiation Objective options MUST
   always be completely presented in a Request Negotiation message, or
   in a Discovery message in rapid mode.  If there is no initial value,
   the 'value' field SHOULD be set to the 'null' value defined by CBOR.

   Synchronization Objective options are similar, but MUST be carried by
   Discovery, Discovery Response, Request Synchronization, or Flood
   Synchronization messages only.  They include 'value' fields only in
   Synchronization or Flood Synchronization messages.

   The design of an objective interacts in various ways with the design
   of the ASAs that will use it.  ASA design considerations are
   discussed in [ASA-GUIDELINES].

2.10.4.  Organizing of Objective Options

   Generic objective options MUST be specified in documents available to
   the public and SHOULD be designed to use either the negotiation or
   the synchronization mechanism described above.

   As noted earlier, one negotiation objective is handled by each GRASP
   negotiation thread.  Therefore, a negotiation objective, which is
   based on a specific function or action, SHOULD be organized as a
   single GRASP option.  It is NOT RECOMMENDED to organize multiple
   negotiation objectives into a single option nor to split a single
   function or action into multiple negotiation objectives.

   It is important to understand that GRASP negotiation does not support
   transactional integrity.  If transactional integrity is needed for a
   specific objective, this must be ensured by the ASA.  For example, an
   ASA might need to ensure that it only participates in one negotiation
   thread at the same time.  Such an ASA would need to stop listening
   for incoming negotiation requests before generating an outgoing
   negotiation request.

   A synchronization objective SHOULD be organized as a single GRASP

   Some objectives will support more than one operational mode.  An
   example is a negotiation objective with both a dry-run mode (where
   the negotiation is to determine whether the other end can, in fact,
   make the requested change without problems) and a live mode, as
   explained in Section 2.5.5.  The semantics of such modes will be
   defined in the specification of the objectives.  These objectives
   SHOULD include flags indicating the applicable mode(s).

   An issue requiring particular attention is that GRASP itself is not a
   transactionally safe protocol.  Any state associated with a dry-run
   operation, such as temporarily reserving a resource for subsequent
   use in a live run, is entirely a matter for the designer of the ASA

   As indicated in Section 2.1, an objective's value may include
   multiple parameters.  Parameters might be categorized into two
   classes: the obligatory ones presented as fixed fields and the
   optional ones presented in some other form of data structure embedded
   in CBOR.  The format might be inherited from an existing management
   or configuration protocol, with the objective option acting as a
   carrier for that format.  The data structure might be defined in a
   formal language, but that is a matter for the specifications of
   individual objectives.  There are many candidates, according to the
   context, such as ABNF, RBNF, XML Schema, YANG, etc.  GRASP itself is
   agnostic on these questions.  The only restriction is that the format
   can be mapped into CBOR.

   It is NOT RECOMMENDED to mix parameters that have significantly
   different response-time characteristics in a single objective.
   Separate objectives are more suitable for such a scenario.

   All objectives MUST support GRASP discovery.  However, as mentioned
   in Section 2.3, it is acceptable for an ASA to use an alternative
   method of discovery.

   Normally, a GRASP objective will refer to specific technical
   parameters as explained in Section 2.1.  However, it is acceptable to
   define an abstract objective for the purpose of managing or
   coordinating ASAs.  It is also acceptable to define a special-purpose
   objective for purposes such as trust bootstrapping or formation of
   the ACP.

   To guarantee convergence, a limited number of rounds or a timeout is
   needed for each negotiation objective.  Therefore, the definition of
   each negotiation objective SHOULD clearly specify this, for example,
   a default loop count and timeout, so that the negotiation can always
   be terminated properly.  If not, the GRASP defaults will apply.

   There must be a well-defined procedure for concluding that a
   negotiation cannot succeed, and if so, deciding what happens next
   (e.g., deadlock resolution, tie-breaking, or reversion to best-effort
   service).  This MUST be specified for individual negotiation

2.10.5.  Experimental and Example Objective Options

   The names "EX0" through "EX9" have been reserved for experimental
   options.  Multiple names have been assigned because a single
   experiment may use multiple options simultaneously.  These
   experimental options are highly likely to have different meanings
   when used for different experiments.  Therefore, they SHOULD NOT be
   used without an explicit human decision and MUST NOT be used in
   unmanaged networks such as home networks.

   These names are also RECOMMENDED for use in documentation examples.

3.  Security Considerations

   A successful attack on negotiation-enabled nodes would be extremely
   harmful, as such nodes might end up with a completely undesirable
   configuration that would also adversely affect their peers.  GRASP
   nodes and messages therefore require full protection.  As explained
   in Section 2.5.1, GRASP MUST run within a secure environment such as
   the ACP [RFC8994], except for the constrained instances described in
   Section 2.5.2.

      A cryptographically authenticated identity for each device is
      needed in an Autonomic Network.  It is not safe to assume that a
      large network is physically secured against interference or that
      all personnel are trustworthy.  Each autonomic node MUST be
      capable of proving its identity and authenticating its messages.
      GRASP relies on a separate, external certificate-based security
      mechanism to support authentication, data integrity protection,
      and anti-replay protection.

      Since GRASP must be deployed in an existing secure environment,
      the protocol itself specifies nothing concerning the trust anchor
      and certification authority.  For example, in the ACP [RFC8994],
      all nodes can trust each other and the ASAs installed in them.

      If GRASP is used temporarily without an external security
      mechanism, for example, during system bootstrap (Section 2.5.1),
      the Session ID (Section 2.7) will act as a nonce to provide
      limited protection against the injecting of responses by third
      parties.  A full analysis of the secure bootstrap process is in

   Authorization and roles
      GRASP is agnostic about the roles and capabilities of individual
      ASAs and about which objectives a particular ASA is authorized to
      support.  An implementation might support precautions such as
      allowing only one ASA in a given node to modify a given objective,
      but this may not be appropriate in all cases.  For example, it
      might be operationally useful to allow an old and a new version of
      the same ASA to run simultaneously during an overlap period.
      These questions are out of scope for the present specification.

   Privacy and confidentiality
      GRASP is intended for network-management purposes involving
      network elements, not end hosts.  Therefore, no personal
      information is expected to be involved in the signaling protocol,
      so there should be no direct impact on personal privacy.
      Nevertheless, applications that do convey personal information
      cannot be excluded.  Also, traffic flow paths, VPNs, etc., could
      be negotiated, which could be of interest for traffic analysis.
      Operators generally want to conceal details of their network
      topology and traffic density from outsiders.  Therefore, since
      insider attacks cannot be excluded in a large network, the
      security mechanism for the protocol MUST provide message
      confidentiality.  This is why Section 2.5.1 requires either an ACP
      or an alternative security mechanism.

   Link-local multicast security
      GRASP has no reasonable alternative to using link-local multicast
      for Discovery or Flood Synchronization messages, and these
      messages are sent in the clear and with no authentication.  They
      are only sent on interfaces within the Autonomic Network (see
      Section 2.1 and Section 2.5.1).  They are, however, available to
      on-link eavesdroppers and could be forged by on-link attackers.
      In the case of discovery, the Discovery Responses are unicast and
      will therefore be protected (Section 2.5.1), and an untrusted
      forger will not be able to receive responses.  In the case of
      flood synchronization, an on-link eavesdropper will be able to
      receive the flooded objectives, but there is no response message
      to consider.  Some precautions for Flood Synchronization messages
      are suggested in Section

   DoS attack protection
      GRASP discovery partly relies on insecure link-local multicast.
      Since routers participating in GRASP sometimes relay Discovery
      messages from one link to another, this could be a vector for
      denial-of-service attacks.  Some mitigations are specified in
      Section 2.5.4.  However, malicious code installed inside the ACP
      could always launch DoS attacks consisting of either spurious
      Discovery messages or spurious Discovery Responses.  It is
      important that firewalls prevent any GRASP messages from entering
      the domain from an unknown source.

   Security during bootstrap and discovery
      A node cannot trust GRASP traffic from other nodes until the
      security environment (such as the ACP) has identified the trust
      anchor and can authenticate traffic by validating certificates for
      other nodes.  Also, until it has successfully enrolled [RFC8995],
      a node cannot assume that other nodes are able to authenticate its
      own traffic.  Therefore, GRASP discovery during the bootstrap
      phase for a new device will inevitably be insecure.  Secure
      synchronization and negotiation will be impossible until
      enrollment is complete.  Further details are given in
      Section 2.5.2.

   Security of discovered locators
      When GRASP discovery returns an IP address, it MUST be that of a
      node within the secure environment (Section 2.5.1).  If it returns
      an FQDN or a URI, the ASA that receives it MUST NOT assume that
      the target of the locator is within the secure environment.

4.  CDDL Specification of GRASP

   <CODE BEGINS> file "grasp.cddl"
   grasp-message = (message .within message-structure) / noop-message

   message-structure = [MESSAGE_TYPE, session-id, ?initiator,

   MESSAGE_TYPE = 0..255
   session-id = 0..4294967295 ; up to 32 bits
   grasp-option = any

   message /= discovery-message
   discovery-message = [M_DISCOVERY, session-id, initiator, objective]

   message /= response-message ; response to Discovery
   response-message = [M_RESPONSE, session-id, initiator, ttl,
                       (+locator-option // divert-option), ?objective]

   message /= synch-message ; response to Synchronization request
   synch-message = [M_SYNCH, session-id, objective]

   message /= flood-message
   flood-message = [M_FLOOD, session-id, initiator, ttl,
                    +[objective, (locator-option / [])]]

   message /= request-negotiation-message
   request-negotiation-message = [M_REQ_NEG, session-id, objective]

   message /= request-synchronization-message
   request-synchronization-message = [M_REQ_SYN, session-id, objective]

   message /= negotiation-message
   negotiation-message = [M_NEGOTIATE, session-id, objective]

   message /= end-message
   end-message = [M_END, session-id, accept-option / decline-option]

   message /= wait-message
   wait-message = [M_WAIT, session-id, waiting-time]

   message /= invalid-message
   invalid-message = [M_INVALID, session-id, ?any]

   noop-message = [M_NOOP]

   divert-option = [O_DIVERT, +locator-option]

   accept-option = [O_ACCEPT]

   decline-option = [O_DECLINE, ?reason]
   reason = text  ; optional UTF-8 error message

   waiting-time = 0..4294967295 ; in milliseconds
   ttl = 0..4294967295 ; in milliseconds

   locator-option /= [O_IPv4_LOCATOR, ipv4-address,
                      transport-proto, port-number]
   ipv4-address = bytes .size 4

   locator-option /= [O_IPv6_LOCATOR, ipv6-address,
                      transport-proto, port-number]
   ipv6-address = bytes .size 16

   locator-option /= [O_FQDN_LOCATOR, text, transport-proto,

   locator-option /= [O_URI_LOCATOR, text,
                      transport-proto / null, port-number / null]

   transport-proto = IPPROTO_TCP / IPPROTO_UDP
   port-number = 0..65535

   initiator = ipv4-address / ipv6-address

   objective-flags = uint .bits objective-flag

   objective-flag = &(
     F_DISC: 0    ; valid for discovery
     F_NEG: 1     ; valid for negotiation
     F_SYNCH: 2   ; valid for synchronization
     F_NEG_DRY: 3 ; negotiation is a dry run

   objective = [objective-name, objective-flags,
                loop-count, ?objective-value]

   objective-name = text ; see section "Format of Objective Options"

   objective-value = any

   loop-count = 0..255

   ; Constants for message types and option types

   M_NOOP = 0
   M_REQ_NEG = 3
   M_REQ_SYN = 4
   M_END = 6
   M_WAIT = 7
   M_SYNCH = 8
   M_FLOOD = 9
   M_INVALID = 99

   O_DIVERT = 100
   O_ACCEPT = 101
   O_DECLINE = 102
   O_IPv6_LOCATOR = 103
   O_IPv4_LOCATOR = 104
   O_URI_LOCATOR = 106

5.  IANA Considerations

   This document defines the GeneRic Autonomic Signaling Protocol

   Section 2.6 explains the following link-local multicast addresses
   that IANA has assigned for use by GRASP.

   Assigned in the "Link-Local Scope Multicast Addresses" subregistry of
   the "IPv6 Multicast Address Space Registry":

   Address(es):  ff02::13
   Description:  ALL_GRASP_NEIGHBORS
   Reference:  RFC 8990

   Assigned in the "Local Network Control Block ( -
   (224.0.0/24))" subregistry of the "IPv4 Multicast Address Space

   Description:  ALL_GRASP_NEIGHBORS
   Reference:  RFC 8990

   Section 2.6 explains the following User Port (GRASP_LISTEN_PORT),
   which IANA has assigned for use by GRASP for both UDP and TCP:

   Service Name:  grasp
   Port Number:  7017
   Transport Protocol:  udp, tcp
   Description  GeneRic Autonomic Signaling Protocol
   Assignee:  IESG <iesg@ietf.org>
   Contact:  IETF Chair <chair@ietf.org>
   Reference:  RFC 8990

   The IANA has created the "GeneRic Autonomic Signaling Protocol
   (GRASP) Parameters" registry, which includes two subregistries:
   "GRASP Messages and Options" and "GRASP Objective Names".

   The values in the "GRASP Messages and Options" subregistry are names
   paired with decimal integers.  Future values MUST be assigned using
   the Standards Action policy defined by [RFC8126].  The following
   initial values are assigned by this document:

                        | Value | Message/Option |
                        | 0     | M_NOOP         |
                        | 1     | M_DISCOVERY    |
                        | 2     | M_RESPONSE     |
                        | 3     | M_REQ_NEG      |
                        | 4     | M_REQ_SYN      |
                        | 5     | M_NEGOTIATE    |
                        | 6     | M_END          |
                        | 7     | M_WAIT         |
                        | 8     | M_SYNCH        |
                        | 9     | M_FLOOD        |
                        | 99    | M_INVALID      |
                        | 100   | O_DIVERT       |
                        | 101   | O_ACCEPT       |
                        | 102   | O_DECLINE      |
                        | 103   | O_IPv6_LOCATOR |
                        | 104   | O_IPv4_LOCATOR |
                        | 105   | O_FQDN_LOCATOR |
                        | 106   | O_URI_LOCATOR  |

                             Table 1: Initial
                           Values of the "GRASP
                          Messages and Options"

   The values in the "GRASP Objective Names" subregistry are UTF-8
   strings that MUST NOT include a colon (":"), according to
   Section 2.10.1.  Future values MUST be assigned using the
   Specification Required policy defined by [RFC8126].

   To assist expert review of a new objective, the specification should
   include a precise description of the format of the new objective,
   with sufficient explanation of its semantics to allow independent
   implementations.  See Section 2.10.3 for more details.  If the new
   objective is similar in name or purpose to a previously registered
   objective, the specification should explain why a new objective is

   The following initial values are assigned by this document:

                      | Objective Name | Reference |
                      | EX0            | RFC 8990  |
                      | EX1            | RFC 8990  |
                      | EX2            | RFC 8990  |
                      | EX3            | RFC 8990  |
                      | EX4            | RFC 8990  |
                      | EX5            | RFC 8990  |
                      | EX6            | RFC 8990  |
                      | EX7            | RFC 8990  |
                      | EX8            | RFC 8990  |
                      | EX9            | RFC 8990  |

                        Table 2: Initial Values of
                           the "GRASP Objective
                            Names" Subregistry

6.  References

6.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,

   [RFC3629]  Yergeau, F., "UTF-8, a transformation format of ISO
              10646", STD 63, RFC 3629, DOI 10.17487/RFC3629, November
              2003, <https://www.rfc-editor.org/info/rfc3629>.

   [RFC3986]  Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform
              Resource Identifier (URI): Generic Syntax", STD 66,
              RFC 3986, DOI 10.17487/RFC3986, January 2005,

   [RFC4086]  Eastlake 3rd, D., Schiller, J., and S. Crocker,
              "Randomness Requirements for Security", BCP 106, RFC 4086,
              DOI 10.17487/RFC4086, June 2005,

   [RFC7217]  Gont, F., "A Method for Generating Semantically Opaque
              Interface Identifiers with IPv6 Stateless Address
              Autoconfiguration (SLAAC)", RFC 7217,
              DOI 10.17487/RFC7217, April 2014,

   [RFC8085]  Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage
              Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085,
              March 2017, <https://www.rfc-editor.org/info/rfc8085>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <https://www.rfc-editor.org/info/rfc8174>.

   [RFC8610]  Birkholz, H., Vigano, C., and C. Bormann, "Concise Data
              Definition Language (CDDL): A Notational Convention to
              Express Concise Binary Object Representation (CBOR) and
              JSON Data Structures", RFC 8610, DOI 10.17487/RFC8610,
              June 2019, <https://www.rfc-editor.org/info/rfc8610>.

   [RFC8949]  Bormann, C. and P. Hoffman, "Concise Binary Object
              Representation (CBOR)", STD 94, RFC 8949,
              DOI 10.17487/RFC8949, December 2020,

   [RFC8994]  Eckert, T., Ed., Behringer, M., Ed., and S. Bjarnason, "An
              Autonomic Control Plane (ACP)", RFC 8994,
              DOI 10.17487/RFC8994, May 2021,

6.2.  Informative References

   [ADNCP]    Stenberg, M., "Autonomic Distributed Node Consensus
              Protocol", Work in Progress, Internet-Draft, draft-
              stenberg-anima-adncp-00, 5 March 2015,

              Carpenter, B., Ciavaglia, L., Jiang, S., and P. Peloso,
              "Guidelines for Autonomic Service Agents", Work in
              Progress, Internet-Draft, draft-ietf-anima-asa-guidelines-
              00, 14 November 2020, <https://tools.ietf.org/html/draft-

   [IGCP]     Behringer, M. H., Chaparadza, R., Xin, L., Mahkonen, H.,
              and R. Petre, "IP based Generic Control Protocol (IGCP)",
              Work in Progress, Internet-Draft, draft-chaparadza-
              intarea-igcp-00, 25 July 2011,

   [RFC2205]  Braden, R., Ed., Zhang, L., Berson, S., Herzog, S., and S.
              Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1
              Functional Specification", RFC 2205, DOI 10.17487/RFC2205,
              September 1997, <https://www.rfc-editor.org/info/rfc2205>.

   [RFC2334]  Luciani, J., Armitage, G., Halpern, J., and N. Doraswamy,
              "Server Cache Synchronization Protocol (SCSP)", RFC 2334,
              DOI 10.17487/RFC2334, April 1998,

   [RFC2608]  Guttman, E., Perkins, C., Veizades, J., and M. Day,
              "Service Location Protocol, Version 2", RFC 2608,
              DOI 10.17487/RFC2608, June 1999,

   [RFC2865]  Rigney, C., Willens, S., Rubens, A., and W. Simpson,
              "Remote Authentication Dial In User Service (RADIUS)",
              RFC 2865, DOI 10.17487/RFC2865, June 2000,

   [RFC3416]  Presuhn, R., Ed., "Version 2 of the Protocol Operations
              for the Simple Network Management Protocol (SNMP)",
              STD 62, RFC 3416, DOI 10.17487/RFC3416, December 2002,

   [RFC3493]  Gilligan, R., Thomson, S., Bound, J., McCann, J., and W.
              Stevens, "Basic Socket Interface Extensions for IPv6",
              RFC 3493, DOI 10.17487/RFC3493, February 2003,

   [RFC4861]  Narten, T., Nordmark, E., Simpson, W., and H. Soliman,
              "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861,
              DOI 10.17487/RFC4861, September 2007,

   [RFC5612]  Eronen, P. and D. Harrington, "Enterprise Number for
              Documentation Use", RFC 5612, DOI 10.17487/RFC5612, August
              2009, <https://www.rfc-editor.org/info/rfc5612>.

   [RFC5971]  Schulzrinne, H. and R. Hancock, "GIST: General Internet
              Signalling Transport", RFC 5971, DOI 10.17487/RFC5971,
              October 2010, <https://www.rfc-editor.org/info/rfc5971>.

   [RFC6206]  Levis, P., Clausen, T., Hui, J., Gnawali, O., and J. Ko,
              "The Trickle Algorithm", RFC 6206, DOI 10.17487/RFC6206,
              March 2011, <https://www.rfc-editor.org/info/rfc6206>.

   [RFC6241]  Enns, R., Ed., Bjorklund, M., Ed., Schoenwaelder, J., Ed.,
              and A. Bierman, Ed., "Network Configuration Protocol
              (NETCONF)", RFC 6241, DOI 10.17487/RFC6241, June 2011,

   [RFC6733]  Fajardo, V., Ed., Arkko, J., Loughney, J., and G. Zorn,
              Ed., "Diameter Base Protocol", RFC 6733,
              DOI 10.17487/RFC6733, October 2012,

   [RFC6762]  Cheshire, S. and M. Krochmal, "Multicast DNS", RFC 6762,
              DOI 10.17487/RFC6762, February 2013,

   [RFC6763]  Cheshire, S. and M. Krochmal, "DNS-Based Service
              Discovery", RFC 6763, DOI 10.17487/RFC6763, February 2013,

   [RFC6887]  Wing, D., Ed., Cheshire, S., Boucadair, M., Penno, R., and
              P. Selkirk, "Port Control Protocol (PCP)", RFC 6887,
              DOI 10.17487/RFC6887, April 2013,

   [RFC7558]  Lynn, K., Cheshire, S., Blanchet, M., and D. Migault,
              "Requirements for Scalable DNS-Based Service Discovery
              (DNS-SD) / Multicast DNS (mDNS) Extensions", RFC 7558,
              DOI 10.17487/RFC7558, July 2015,

   [RFC7575]  Behringer, M., Pritikin, M., Bjarnason, S., Clemm, A.,
              Carpenter, B., Jiang, S., and L. Ciavaglia, "Autonomic
              Networking: Definitions and Design Goals", RFC 7575,
              DOI 10.17487/RFC7575, June 2015,

   [RFC7576]  Jiang, S., Carpenter, B., and M. Behringer, "General Gap
              Analysis for Autonomic Networking", RFC 7576,
              DOI 10.17487/RFC7576, June 2015,

   [RFC7787]  Stenberg, M. and S. Barth, "Distributed Node Consensus
              Protocol", RFC 7787, DOI 10.17487/RFC7787, April 2016,

   [RFC7788]  Stenberg, M., Barth, S., and P. Pfister, "Home Networking
              Control Protocol", RFC 7788, DOI 10.17487/RFC7788, April
              2016, <https://www.rfc-editor.org/info/rfc7788>.

   [RFC8040]  Bierman, A., Bjorklund, M., and K. Watsen, "RESTCONF
              Protocol", RFC 8040, DOI 10.17487/RFC8040, January 2017,

   [RFC8126]  Cotton, M., Leiba, B., and T. Narten, "Guidelines for
              Writing an IANA Considerations Section in RFCs", BCP 26,
              RFC 8126, DOI 10.17487/RFC8126, June 2017,

   [RFC8264]  Saint-Andre, P. and M. Blanchet, "PRECIS Framework:
              Preparation, Enforcement, and Comparison of
              Internationalized Strings in Application Protocols",
              RFC 8264, DOI 10.17487/RFC8264, October 2017,

   [RFC8368]  Eckert, T., Ed. and M. Behringer, "Using an Autonomic
              Control Plane for Stable Connectivity of Network
              Operations, Administration, and Maintenance (OAM)",
              RFC 8368, DOI 10.17487/RFC8368, May 2018,

   [RFC8415]  Mrugalski, T., Siodelski, M., Volz, B., Yourtchenko, A.,
              Richardson, M., Jiang, S., Lemon, T., and T. Winters,
              "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)",
              RFC 8415, DOI 10.17487/RFC8415, November 2018,

   [RFC8991]  Carpenter, B., Liu, B., Ed., Wang, W., and X. Gong,
              "GeneRic Autonomic Signaling Protocol Application Program
              Interface (GRASP API)", RFC 8991, DOI 10.17487/RFC8991,
              May 2021, <https://www.rfc-editor.org/info/rfc8991>.

   [RFC8993]  Behringer, M., Ed., Carpenter, B., Eckert, T., Ciavaglia,
              L., and J. Nobre, "A Reference Model for Autonomic
              Networking", RFC 8993, DOI 10.17487/RFC8993, May 2021,

   [RFC8995]  Pritikin, M., Richardson, M., Eckert, T., Behringer, M.,
              and K. Watsen, "Bootstrapping Remote Secure Key
              Infrastructure (BRSKI)", RFC 8995, DOI 10.17487/RFC8995,
              May 2021, <https://www.rfc-editor.org/info/rfc8995>.

Appendix A.  Example Message Formats

   For readers unfamiliar with CBOR, this appendix shows a number of
   example GRASP messages conforming to the CDDL syntax given in
   Section 4.  Each message is shown three times in the following

   1.  CBOR diagnostic notation.

   2.  Similar, but showing the names of the constants.  (Details of the
       flag bit encoding are omitted.)

   3.  Hexadecimal version of the CBOR wire format.

   Long lines are split for display purposes only.

A.1.  Discovery Example

   The initiator (2001:db8:f000:baaa:28cc:dc4c:9703:6781) multicasts a
   Discovery message looking for objective EX1:

   [1, 13948744, h'20010db8f000baaa28ccdc4c97036781', ["EX1", 5, 2, 0]]
   [M_DISCOVERY, 13948744, h'20010db8f000baaa28ccdc4c97036781',
                 ["EX1", F_SYNCH_bits, 2, 0]]

   A peer (2001:0db8:f000:baaa:f000:baaa:f000:baaa) responds with a

   [2, 13948744, h'20010db8f000baaa28ccdc4c97036781', 60000,
                 [103, h'20010db8f000baaaf000baaaf000baaa', 6, 49443]]
   [M_RESPONSE, 13948744, h'20010db8f000baaa28ccdc4c97036781', 60000,
                 [O_IPv6_LOCATOR, h'20010db8f000baaaf000baaaf000baaa',
                  IPPROTO_TCP, 49443]]

A.2.  Flood Example

   The initiator multicasts a Flood Synchronization message.  The single
   objective has a null locator.  There is no response:

[9, 3504974, h'20010db8f000baaa28ccdc4c97036781', 10000,
             [["EX1", 5, 2, ["Example 1 value=", 100]],[] ] ]
[M_FLOOD, 3504974, h'20010db8f000baaa28ccdc4c97036781', 10000,
             [["EX1", F_SYNCH_bits, 2, ["Example 1 value=", 100]],[] ] ]

A.3.  Synchronization Example

   Following successful discovery of objective EX2, the initiator
   unicasts a Request Synchronization message:

   [4, 4038926, ["EX2", 5, 5, 0]]
   [M_REQ_SYN, 4038926, ["EX2", F_SYNCH_bits, 5, 0]]

   The peer responds with a value:

 [8, 4038926, ["EX2", 5, 5, ["Example 2 value=", 200]]]
 [M_SYNCH, 4038926, ["EX2", F_SYNCH_bits, 5, ["Example 2 value=", 200]]]

A.4.  Simple Negotiation Example

   Following successful discovery of objective EX3, the initiator
   unicasts a Request Negotiation message:

   [3, 802813, ["EX3", 3, 6, ["NZD", 47]]]
   [M_REQ_NEG, 802813, ["EX3", F_NEG_bits, 6, ["NZD", 47]]]

   The peer responds with immediate acceptance.  Note that no objective
   is needed because the initiator's request was accepted without

   [6, 802813, [101]]
   [M_END , 802813, [O_ACCEPT]]

A.5.  Complete Negotiation Example

   Again the initiator unicasts a Request Negotiation message:

   [3, 13767778, ["EX3", 3, 6, ["NZD", 410]]]
   [M_REQ_NEG, 13767778, ["EX3", F_NEG_bits, 6, ["NZD", 410]]]

   The responder starts to negotiate (making an offer):

   [5, 13767778, ["EX3", 3, 6, ["NZD", 80]]]
   [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 6, ["NZD", 80]]]

   The initiator continues to negotiate (reducing its request, and note
   that the loop count is decremented):

   [5, 13767778, ["EX3", 3, 5, ["NZD", 307]]]
   [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 5, ["NZD", 307]]]

   The responder asks for more time:

   [7, 13767778, 34965]
   [M_WAIT, 13767778, 34965]

   The responder continues to negotiate (increasing its offer):

   [5, 13767778, ["EX3", 3, 4, ["NZD", 120]]]
   [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 4, ["NZD", 120]]]

   The initiator continues to negotiate (reducing its request):

   [5, 13767778, ["EX3", 3, 3, ["NZD", 246]]]
   [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 3, ["NZD", 246]]]

   The responder refuses to negotiate further:

   [6, 13767778, [102, "Insufficient funds"]]
   [M_END , 13767778, [O_DECLINE, "Insufficient funds"]]

   This negotiation has failed.  If either side had sent [M_END,
   13767778, [O_ACCEPT]] it would have succeeded, converging on the
   objective value in the preceding M_NEGOTIATE.  Note that apart from
   the initial M_REQ_NEG, the process is symmetrical.

Appendix B.  Requirement Analysis of Discovery, Synchronization, and

   This section discusses the requirements for discovery, negotiation,
   and synchronization capabilities.  The primary user of the protocol
   is an Autonomic Service Agent (ASA), so the requirements are mainly
   expressed as the features needed by an ASA.  A single physical device
   might contain several ASAs, and a single ASA might manage several
   technical objectives.  If a technical objective is managed by several
   ASAs, any necessary coordination is outside the scope of GRASP.
   Furthermore, requirements for ASAs themselves, such as the processing
   of Intent [RFC7575], are out of scope for the present document.

B.1.  Requirements for Discovery

   D1.   ASAs may be designed to manage any type of configurable device
         or software, as required in Appendix B.2.  A basic requirement
         is therefore that the protocol can represent and discover any
         kind of technical objective (as defined in Section 2.1) among
         arbitrary subsets of participating nodes.

         In an Autonomic Network, we must assume that when a device
         starts up, it has no information about any peer devices, the
         network structure, or the specific role it must play.  The
         ASA(s) inside the device are in the same situation.  In some
         cases, when a new application session starts within a device,
         the device or ASA may again lack information about relevant
         peers.  For example, it might be necessary to set up resources
         on multiple other devices, coordinated and matched to each
         other so that there is no wasted resource.  Security settings
         might also need updating to allow for the new device or user.
         The relevant peers may be different for different technical
         objectives.  Therefore discovery needs to be repeated as often
         as necessary to find peers capable of acting as counterparts
         for each objective that a discovery initiator needs to handle.
         From this background we derive the next three requirements:

   D2.   When an ASA first starts up, it may have no knowledge of the
         specific network to which it is attached.  Therefore the
         discovery process must be able to support any network scenario,
         assuming only that the device concerned is bootstrapped from
         factory condition.

   D3.   When an ASA starts up, it must require no configured location
         information about any peers in order to discover them.

   D4.   If an ASA supports multiple technical objectives, relevant
         peers may be different for different discovery objectives, so
         discovery needs to be performed separately to find counterparts
         for each objective.  Thus, there must be a mechanism by which
         an ASA can separately discover peer ASAs for each of the
         technical objectives that it needs to manage, whenever

   D5.   Following discovery, an ASA will normally perform negotiation
         or synchronization for the corresponding objectives.  The
         design should allow for this by conveniently linking discovery
         to negotiation and synchronization.  It may provide an optional
         mechanism to combine discovery and negotiation/synchronization
         in a single protocol exchange.

   D6.   Some objectives may only be significant on the local link, but
         others may be significant across the routed network and require
         off-link operations.  Thus, the relevant peers might be
         immediate neighbors on the same layer 2 link, or they might be
         more distant and only accessible via layer 3.  The mechanism
         must therefore provide both on-link and off-link discovery of
         ASAs supporting specific technical objectives.

   D7.   The discovery process should be flexible enough to allow for
         special cases, such as the following:

         *  During initialization, a device must be able to establish
            mutual trust with autonomic nodes elsewhere in the network
            and participate in an authentication mechanism.  Although
            this will inevitably start with a discovery action, it is a
            special case precisely because trust is not yet established.
            This topic is the subject of [RFC8995].  We require that
            once trust has been established for a device, all ASAs
            within the device inherit the device's credentials and are
            also trusted.  This does not preclude the device having
            multiple credentials.

         *  Depending on the type of network involved, discovery of
            other central functions might be needed, such as the Network
            Operations Center (NOC) [RFC8368].  The protocol must be
            capable of supporting such discovery during initialization,
            as well as discovery during ongoing operation.

   D8.   The discovery process must not generate excessive traffic and
         must take account of sleeping nodes.

   D9.   There must be a mechanism for handling stale discovery results.

B.2.  Requirements for Synchronization and Negotiation Capability

   Autonomic Networks need to be able to manage many different types of
   parameters and consider many dimensions, such as latency, load,
   unused or limited resources, conflicting resource requests, security
   settings, power saving, load balancing, etc.  Status information and
   resource metrics need to be shared between nodes for dynamic
   adjustment of resources and for monitoring purposes.  While this
   might be achieved by existing protocols when they are available, the
   new protocol needs to be able to support parameter exchange,
   including mutual synchronization, even when no negotiation as such is
   required.  In general, these parameters do not apply to all
   participating nodes, but only to a subset.

   SN1.  A basic requirement for the protocol is therefore the ability
         to represent, discover, synchronize, and negotiate almost any
         kind of network parameter among selected subsets of
         participating nodes.

   SN2.  Negotiation is an iterative request/response process that must
         be guaranteed to terminate (with success or failure).  While
         tie-breaking rules must be defined specifically for each use
         case, the protocol should have some general mechanisms in
         support of loop and deadlock prevention, such as hop-count
         limits or timeouts.

   SN3.  Synchronization must be possible for groups of nodes ranging
         from small to very large.

   SN4.  To avoid "reinventing the wheel", the protocol should be able
         to encapsulate the data formats used by existing configuration
         protocols (such as Network Configuration Protocol (NETCONF) and
         YANG) in cases where that is convenient.

   SN5.  Human intervention in complex situations is costly and error
         prone.  Therefore, synchronization or negotiation of parameters
         without human intervention is desirable whenever the
         coordination of multiple devices can improve overall network
         performance.  It follows that the protocol's resource
         requirements must be small enough to fit in any device that
         would otherwise need human intervention.  The issue of running
         in constrained nodes is discussed in [RFC8993].

   SN6.  Human intervention in large networks is often replaced by use
         of a top-down network management system (NMS).  It therefore
         follows that the protocol, as part of the Autonomic Networking
         Infrastructure, should be capable of running in any device that
         would otherwise be managed by an NMS, and that it can coexist
         with an NMS and with protocols such as SNMP and NETCONF.

   SN7.  Specific autonomic features are expected to be implemented by
         individual ASAs, but the protocol must be general enough to
         allow them.  Some examples follow:

         *  Dependencies and conflicts: In order to decide upon a
            configuration for a given device, the device may need
            information from neighbors.  This can be established through
            the negotiation procedure, or through synchronization if
            that is sufficient.  However, a given item in a neighbor may
            depend on other information from its own neighbors, which
            may need another negotiation or synchronization procedure to
            obtain or decide.  Therefore, there are potential
            dependencies and conflicts among negotiation or
            synchronization procedures.  Resolving dependencies and
            conflicts is a matter for the individual ASAs involved.  To
            allow this, there need to be clear boundaries and
            convergence mechanisms for negotiations.  Also some
            mechanisms are needed to avoid loop dependencies or
            uncontrolled growth in a tree of dependencies.  It is the
            ASA designer's responsibility to avoid or detect looping
            dependencies or excessive growth of dependency trees.  The
            protocol's role is limited to bilateral signaling between
            ASAs and the avoidance of loops during bilateral signaling.

         *  Recovery from faults and identification of faulty devices
            should be as automatic as possible.  The protocol's role is
            limited to discovery, synchronization, and negotiation.
            These processes can occur at any time, and an ASA may need
            to repeat any of these steps when the ASA detects an event
            such as a negotiation counterpart failing.

         *  Since a major goal is to minimize human intervention, it is
            necessary that the network can in effect "think ahead"
            before changing its parameters.  One aspect of this is an
            ASA that relies on a knowledge base to predict network
            behavior.  This is out of scope for the signaling protocol.
            However, another aspect is forecasting the effect of a
            change by a "dry run" negotiation before actually installing
            the change.  Signaling a dry run is therefore a desirable
            feature of the protocol.

         Note that management logging, monitoring, alerts, and tools for
         intervention are required.  However, these can only be features
         of individual ASAs, not of the protocol itself.  Another
         document [RFC8368] discusses how such agents may be linked into
         conventional Operations, Administration, and Maintenance (OAM)
         systems via an Autonomic Control Plane [RFC8994].

   SN8.  The protocol will be able to deal with a wide variety of
         technical objectives, covering any type of network parameter.
         Therefore the protocol will need a flexible and easily
         extensible format for describing objectives.  At a later stage,
         it may be desirable to adopt an explicit information model.
         One consideration is whether to adopt an existing information
         model or to design a new one.

B.3.  Specific Technical Requirements

   T1.   It should be convenient for ASA designers to define new
         technical objectives and for programmers to express them,
         without excessive impact on runtime efficiency and footprint.
         In particular, it should be convenient for ASAs to be
         implemented independently of each other as user-space programs
         rather than as kernel code, where such a programming model is
         possible.  The classes of device in which the protocol might
         run is discussed in [RFC8993].

   T2.   The protocol should be easily extensible in case the initially
         defined discovery, synchronization, and negotiation mechanisms
         prove to be insufficient.

   T3.   To be a generic platform, the protocol payload format should be
         independent of the transport protocol or IP version.  In
         particular, it should be able to run over IPv6 or IPv4.
         However, some functions, such as multicasting on a link, might
         need to be IP version dependent.  By default, IPv6 should be

   T4.   The protocol must be able to access off-link counterparts via
         routable addresses, i.e., must not be restricted to link-local

   T5.   It must also be possible for an external discovery mechanism to
         be used, if appropriate for a given technical objective.  In
         other words, GRASP discovery must not be a prerequisite for
         GRASP negotiation or synchronization.

   T6.   The protocol must be capable of distinguishing multiple
         simultaneous operations with one or more peers, especially when
         wait states occur.

   T7.   Intent: Although the distribution of Intent is out of scope for
         this document, the protocol must not by design exclude its use
         for Intent distribution.

   T8.   Management monitoring, alerts, and intervention: Devices should
         be able to report to a monitoring system.  Some events must be
         able to generate operator alerts, and some provision for
         emergency intervention must be possible (e.g., to freeze
         synchronization or negotiation in a misbehaving device).  These
         features might not use the signaling protocol itself, but its
         design should not exclude such use.

   T9.   Because this protocol may directly cause changes to device
         configurations and have significant impacts on a running
         network, all protocol exchanges need to be fully secured
         against forged messages and man-in-the-middle attacks, and
         secured as much as reasonably possible against denial-of-
         service attacks.  There must also be an encryption mechanism to
         resist unwanted monitoring.  However, it is not required that
         the protocol itself provides these security features; it may
         depend on an existing secure environment.

Appendix C.  Capability Analysis of Current Protocols

   This appendix discusses various existing protocols with properties
   related to the requirements described in Appendix B.  The purpose is
   to evaluate whether any existing protocol, or a simple combination of
   existing protocols, can meet those requirements.

   Numerous protocols include some form of discovery, but these all
   appear to be very specific in their applicability.  Service Location
   Protocol (SLP) [RFC2608] provides service discovery for managed
   networks, but it requires configuration of its own servers.  DNS-
   Based Service Discovery (DNS-SD) [RFC6763] combined with Multicast
   DNS (mDNS) [RFC6762] provides service discovery for small networks
   with a single link layer.  [RFC7558] aims to extend this to larger
   autonomous networks, but this is not yet standardized.  However, both
   SLP and DNS-SD appear to target primarily application-layer services,
   not the layer 2 and 3 objectives relevant to basic network
   configuration.  Both SLP and DNS-SD are text-based protocols.

   Simple Network Management Protocol (SNMP) [RFC3416] uses a command/
   response model not well suited for peer negotiation.  NETCONF
   [RFC6241] uses an RPC model that does allow positive or negative
   responses from the target system, but this is still not adequate for

   There are various existing protocols that have elementary negotiation
   abilities, such as Dynamic Host Configuration Protocol for IPv6
   (DHCPv6) [RFC8415], Neighbor Discovery (ND) [RFC4861], Port Control
   Protocol (PCP) [RFC6887], Remote Authentication Dial-In User Service
   (RADIUS) [RFC2865], Diameter [RFC6733], etc.  Most of them are
   configuration or management protocols.  However, they either provide
   only a simple request/response model in a master/slave context or
   very limited negotiation abilities.

   There are some signaling protocols with an element of negotiation.
   For example, Resource ReSerVation Protocol (RSVP) [RFC2205] was
   designed for negotiating quality-of-service parameters along the path
   of a unicast or multicast flow.  RSVP is a very specialized protocol
   aimed at end-to-end flows.  A more generic design is General Internet
   Signalling Transport (GIST) [RFC5971]; however, it tries to solve
   many problems, making it complex, and is also aimed at per-flow
   signaling across many hops rather than at device-to-device signaling.
   However, we cannot completely exclude extended RSVP or GIST as a
   synchronization and negotiation protocol.  They do not appear to be
   directly usable for peer discovery.

   RESTCONF [RFC8040] is a protocol intended to convey NETCONF
   information expressed in the YANG language via HTTP, including the
   ability to transit HTML intermediaries.  While this is a powerful
   approach in the context of centralized configuration of a complex
   network, it is not well adapted to efficient interactive negotiation
   between peer devices, especially simple ones that might not include
   YANG processing already.

   The Distributed Node Consensus Protocol (DNCP) [RFC7787] is defined
   as a generic form of a state synchronization protocol, with a
   proposed usage profile being the Home Networking Control Protocol
   (HNCP) [RFC7788] for configuring Homenet routers.  A specific
   application of DNCP for Autonomic Networking was proposed in [ADNCP].
   According to [RFC7787]:

   |  DNCP is designed to provide a way for each participating node to
   |  publish a set of TLV (Type-Length-Value) tuples (at most 64 KB)
   |  and to provide a shared and common view about the data
   |  published...
   |  DNCP is most suitable for data that changes only infrequently...
   |  If constant rapid state changes are needed, the preferable choice
   |  is to use an additional point-to-point channel...

   Specific features of DNCP include:

   *  Every participating node has a unique node identifier.

   *  DNCP messages are encoded as a sequence of TLV objects and sent
      over unicast UDP or TCP, with or without (D)TLS security.

   *  Multicast is used only for discovery of DNCP neighbors when lower
      security is acceptable.

   *  Synchronization of state is maintained by a flooding process using
      the Trickle algorithm.  There is no bilateral synchronization or
      negotiation capability.

   *  The HNCP profile of DNCP is designed to operate between directly
      connected neighbors on a shared link using UDP and link-local IPv6

   DNCP does not meet the needs of a general negotiation protocol
   because it is designed specifically for flooding synchronization.
   Also, in its HNCP profile, it is limited to link-local messages and
   to IPv6.  However, at the minimum, it is a very interesting test case
   for this style of interaction between devices without needing a
   central authority, and it is a proven method of network-wide state
   synchronization by flooding.

   The Server Cache Synchronization Protocol (SCSP) [RFC2334] also
   describes a method for cache synchronization and cache replication
   among a group of nodes.

   A proposal was made some years ago for an IP based Generic Control
   Protocol (IGCP) [IGCP].  This was aimed at information exchange and
   negotiation but not directly at peer discovery.  However, it has many
   points in common with the present work.

   None of the above solutions appears to completely meet the needs of
   generic discovery, state synchronization, and negotiation in a single
   solution.  Many of the protocols assume that they are working in a
   traditional top-down or north-south scenario, rather than a fluid
   peer-to-peer scenario.  Most of them are specialized in one way or
   another.  As a result, we have not identified a combination of
   existing protocols that meets the requirements in Appendix B.  Also,
   we have not identified a path by which one of the existing protocols
   could be extended to meet the requirements.


   A major contribution to the original draft version of this document
   was made by Sheng Jiang, and significant contributions were made by
   Toerless Eckert.  Significant early review inputs were received from
   Joel Halpern, Barry Leiba, Charles E. Perkins, and Michael
   Richardson.  William Atwood provided important assistance in
   debugging a prototype implementation.

   Valuable comments were received from Michael Behringer, Jéferson
   Campos Nobre, Laurent Ciavaglia, Zongpeng Du, Yu Fu, Joel Jaeggli,
   Zhenbin Li, Dimitri Papadimitriou, Pierre Peloso, Reshad Rahman,
   Markus Stenberg, Martin Stiemerling, Rene Struik, Martin Thomson,
   Dacheng Zhang, and participants in the Network Management Research
   Group, the ANIMA Working Group, and the IESG.

Authors' Addresses

   Carsten Bormann
   Universität Bremen TZI
   Postfach 330440
   D-28359 Bremen

   Email: cabo@tzi.org

   Brian Carpenter (editor)
   School of Computer Science
   University of Auckland
   PB 92019
   Auckland 1142
   New Zealand

   Email: brian.e.carpenter@gmail.com

   Bing Liu (editor)
   Huawei Technologies Co., Ltd
   Q14, Huawei Campus
   Hai-Dian District
   No.156 Beiqing Road

   Email: leo.liubing@huawei.com