RFC 8991 INFORMATIONAL PROTOCOL · API · AUTONOMIC

GeneRic Autonomic Signaling Protocol Application Program Interface (GRASP API)

The programming interface an autonomic service agent uses to exchange messages with other agents.

On rfc-editor.org ↗
PUBLISHED
May 2021
AUTHORS
B. Carpenter, B. Liu, W. Wang, X. Gong
STREAM
IETF
INTEGER WIDTH IN THE API
32-bit unsigned integers unless otherwise indicated
ASA_HANDLE SIZE
Up to 32 bits
TIMEOUT UNIT
Milliseconds
OBJECTIVE LOOP COUNT FIELD
uint8_t, default GRASP_DEF_LOOPCT
IANA ACTIONS
None
DOI
10.17487/RFC8991

RFC text

Contents
Original text · RFC Editor View on rfc-editor.org ↗
Internet Engineering Task Force (IETF)                      B. Carpenter
Request for Comments: 8991                             Univ. of Auckland
Category: Informational                                      B. Liu, Ed.
ISSN: 2070-1721                                      Huawei Technologies
                                                                 W. Wang
                                                                 X. Gong
                                                         BUPT University
                                                                May 2021


   GeneRic Autonomic Signaling Protocol Application Program Interface
                              (GRASP API)

Abstract

   This document is a conceptual outline of an Application Programming
   Interface (API) for the GeneRic Autonomic Signaling Protocol (GRASP).
   Such an API is needed for Autonomic Service Agents (ASAs) calling the
   GRASP protocol module to exchange Autonomic Network messages with
   other ASAs.  Since GRASP is designed to support asynchronous
   operations, the API will need to be adapted according to the support
   for asynchronicity in various programming languages and operating
   systems.

Status of This Memo

   This document is not an Internet Standards Track specification; it is
   published for informational purposes.

   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).  Not all documents
   approved by the IESG are candidates for any level of Internet
   Standard; see 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
   https://www.rfc-editor.org/info/rfc8991.

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.  GRASP API for ASA
     2.1.  Design Assumptions
     2.2.  Asynchronous Operations
       2.2.1.  Alternative Asynchronous Mechanisms
       2.2.2.  Multiple Negotiation Scenario
       2.2.3.  Overlapping Sessions and Operations
       2.2.4.  Session Termination
     2.3.  API Definition
       2.3.1.  Overview of Functions
       2.3.2.  Parameters and Data Structures
       2.3.3.  Registration
       2.3.4.  Discovery
       2.3.5.  Negotiation
       2.3.6.  Synchronization and Flooding
       2.3.7.  Invalid Message Function
   3.  Security Considerations
   4.  IANA Considerations
   5.  References
     5.1.  Normative References
     5.2.  Informative References
   Appendix A.  Error Codes
     Acknowledgements
   Authors' Addresses

1.  Introduction

   As defined in [RFC8993], the Autonomic Service Agent (ASA) is the
   atomic entity of an autonomic function, and it is instantiated on
   autonomic nodes.  These nodes are members of a secure Autonomic
   Control Plane (ACP) such as defined by [RFC8994].

   When ASAs communicate with each other, they should use the GeneRic
   Autonomic Signaling Protocol (GRASP) [RFC8990].  GRASP relies on the
   message confidentiality and integrity provided by the ACP; a
   consequence of this is that all nodes in a given Autonomic Network
   share the same trust boundary, i.e., the boundary of the ACP.  Nodes
   that have not successfully joined the ACP cannot send, receive, or
   intercept GRASP messages via the ACP and cannot usurp ACP addresses.
   An ASA runs in an ACP node and therefore benefits from the node's
   security properties when transmitting over the ACP, i.e., message
   integrity, message confidentiality, and the fact that unauthorized
   nodes cannot join the ACP.  All ASAs within a given Autonomic Network
   therefore trust each other's messages.  For these reasons, the API
   defined in this document has no explicit security features.

   An important feature of GRASP is the concept of a GRASP objective.
   This is a data structure encoded, like all GRASP messages, in Concise
   Binary Object Representation (CBOR) [RFC8949].  Its main contents are
   a name and a value, explained at more length in the Terminology
   section of [RFC8990].  When an objective is passed from one ASA to
   another using GRASP, its value is either conveyed in one direction
   (by a process of synchronization or flooding) or negotiated
   bilaterally.  The semantics of the value are opaque to GRASP and
   therefore to the API.  Each objective must be accurately specified in
   a dedicated specification, as discussed in "Objective Options"
   (Section 2.10 of [RFC8990]).  In particular, the specification will
   define the syntax and semantics of the value of the objective,
   whether and how it supports a negotiation process, whether it
   supports a dry-run mode, and any other details needed for
   interoperability.  The use of CBOR, with Concise Data Definition
   Language (CDDL) [RFC8610] as the data definition language, allows the
   value to be passed between ASAs regardless of the programming
   languages in use.  Data storage and consistency during negotiation
   are the responsibility of the ASAs involved.  Additionally, GRASP
   needs to cache the latest values of objectives that are received by
   flooding.

   As Figure 1 shows, a GRASP implementation could contain several sub-
   layers.  The bottom layer is the GRASP base protocol module, which is
   only responsible for sending and receiving GRASP messages and
   maintaining shared data structures.  Above that is the basic API
   described in this document.  The upper layer contains some extended
   API functions based upon the GRASP basic protocol.  For example,
   [GRASP-DISTRIB] describes a possible extended function.

                +--------------+          +--------------+
                |     ASAs     |          |     ASAs     |
                +--------------+          +--------------+
                  |          |                    |
                  | +------------------+          |
                  | | GRASP Extended   |          |
                  | | Function API     |          |
                  | +------------------+          |
                  |          |                    |
               +------------------------------------------+
               |         Basic GRASP API Library          |
               +------------------------------------------+
                                   |
                           IPC or system call
                                   |
               +------------------------------------------+
               |  GRASP Core                              |
               |  (functions, data structures, daemon(s)) |
               +------------------------------------------+

                         Figure 1: Software Layout

   Multiple ASAs in a single node will share the same instance of GRASP,
   much as multiple applications share a single TCP/IP stack.  This
   aspect is hidden from individual ASAs by the API and is not further
   discussed here.

   It is desirable that ASAs be designed as portable user-space programs
   using a system-independent API.  In many implementations, the GRASP
   code will therefore be split between user space and kernel space.  In
   user space, library functions provide the API and communicate
   directly with ASAs.  In kernel space, a daemon, or a set of sub-
   services, provides GRASP core functions that are independent of
   specific ASAs, such as multicast handling and relaying, and common
   data structures, such as the discovery cache.  The GRASP API library
   would need to communicate with the GRASP core via an interprocess
   communication (IPC) or a system call mechanism.  The details of this
   are system-dependent.

   Both the GRASP library and the extended function modules should be
   available to the ASAs.  However, since the extended functions are
   expected to be added in an incremental manner, they will be the
   subject of future documents.  This document only describes the basic
   GRASP API.

   The functions provided by the API do not map one-to-one onto GRASP
   messages.  Rather, they are intended to offer convenient support for
   message sequences (such as a discovery request followed by responses
   from several peers or a negotiation request followed by various
   possible responses).  This choice was made to assist ASA programmers
   in writing code based on their application requirements rather than
   needing to understand protocol details.

   In addition to containing the autonomic infrastructure components
   described in [RFC8994] and [RFC8995], a simple autonomic node might
   contain very few ASAs.  Such a node might directly integrate a GRASP
   protocol stack in its code and therefore not require this API to be
   installed.  However, the programmer would need a deeper understanding
   of the GRASP protocol than what is needed to use the API.

   This document gives a conceptual outline of the API.  It is not a
   formal specification for any particular programming language or
   operating system, and it is expected that details will be clarified
   in individual implementations.

2.  GRASP API for ASA

2.1.  Design Assumptions

   The design assumes that an ASA needs to call a separate GRASP
   implementation.  The latter handles protocol details (security,
   sending and listening for GRASP messages, waiting, caching discovery
   results, negotiation looping, sending and receiving synchronization
   data, etc.) but understands nothing about individual GRASP objectives
   (see Section 2.10 of [RFC8990]).  The semantics of objectives are
   unknown to the GRASP protocol and are handled only by the ASAs.
   Thus, this is an abstract API for use by ASAs.  Individual language
   bindings should be defined in separate documents.

   Different ASAs may utilize GRASP features differently, by using GRASP
   for:

   *  discovery purposes only.

   *  negotiation but only as an initiator (client).

   *  negotiation but only as a responder.

   *  negotiation as an initiator or responder.

   *  synchronization but only as an initiator (recipient).

   *  synchronization but only as a responder and/or flooder.

   *  synchronization as an initiator, responder, and/or flooder.

   The API also assumes that one ASA may support multiple objectives.
   Nothing prevents an ASA from supporting some objectives for
   synchronization and others for negotiation.

   The API design assumes that the operating system and programming
   language provide a mechanism for simultaneous asynchronous
   operations.  This is discussed in detail in Section 2.2.

   A few items are out of scope in this version, since practical
   experience is required before including them:

   *  Authorization of ASAs is not defined as part of GRASP and is a
      subject for future study.

   *  User-supplied explicit locators for an objective are not
      supported.  The GRASP core will supply the locator, using the IP
      address of the node concerned.

   *  The rapid mode of GRASP (Section 2.5.4 of [RFC8990]) is not
      supported.

2.2.  Asynchronous Operations

   GRASP depends on asynchronous operations and wait states, and some of
   its messages are not idempotent, meaning that repeating a message may
   cause repeated changes of state in the recipient ASA.  Many ASAs will
   need to support several concurrent operations; for example, an ASA
   might need to negotiate one objective with a peer while discovering
   and synchronizing a different objective with a different peer.
   Alternatively, an ASA that acts as a resource manager might need to
   run simultaneous negotiations for a given objective with multiple
   different peers.  Such an ASA will probably need to support
   uninterruptible atomic changes to its internal data structures, using
   a mechanism provided by the operating system and programming language
   in use.

2.2.1.  Alternative Asynchronous Mechanisms

   Some ASAs need to support asynchronous operations; therefore, the
   GRASP core must do so.  Depending on both the operating system and
   the programming language in use, there are various techniques for
   such parallel operations, three of which we consider here:
   multithreading, an event loop structure using polling, and an event
   loop structure using callback functions.

   1.  In multithreading, the operating system and language will provide
       the necessary support for asynchronous operations, including
       creation of new threads, context switching between threads,
       queues, locks, and implicit wait states.  In this case, API calls
       can be treated as simple synchronous function calls within their
       own thread, even if the function includes wait states, blocking,
       and queueing.  Concurrent operations will each run in their own
       threads.  For example, the discover() call may not return until
       discovery results have arrived or a timeout has occurred.  If the
       ASA has other work to do, the discover() call must be in a thread
       of its own.

   2.  In an event loop implementation with polling, blocking calls are
       not acceptable.  Therefore, all calls must be non-blocking, and
       the main loop could support multiple GRASP sessions in parallel
       by repeatedly polling each one for a change of state.  To
       facilitate this, the API implementation would provide non-
       blocking versions of all the functions that otherwise involve
       blocking and queueing.  In these calls, a 'noReply' code will be
       returned by each call instead of blocking, until such time as the
       event for which it is waiting (or a failure) has occurred.  Thus,
       for example, discover() would return 'noReply' instead of waiting
       until discovery has succeeded or timed out.  The discover() call
       would be repeated in every cycle of the main loop until it
       completes.  Effectively, it becomes a polling call.

   3.  It was noted earlier that some GRASP messages are not idempotent;
       in particular, this applies to each step in a negotiation session
       -- sending the same message twice might produce unintended side
       effects.  This is not affected by event loop polling: repeating a
       call after a 'noReply' does not repeat a message; it simply
       checks whether a reply has been received.

   4.  In an event loop implementation with callbacks, the ASA
       programmer would provide a callback function for each
       asynchronous operation.  This would be called asynchronously when
       a reply is received or a failure such as a timeout occurs.

2.2.2.  Multiple Negotiation Scenario

   The design of GRASP allows the following scenario.  Consider an ASA
   "A" that acts as a resource allocator for some objective.  An ASA "B"
   launches a negotiation with "A" to obtain or release a quantity of
   the resource.  While this negotiation is under way, "B" chooses to
   launch a second simultaneous negotiation with "A" for a different
   quantity of the same resource.  "A" must therefore conduct two
   separate negotiation sessions at the same time with the same peer and
   must not mix them up.

   Note that ASAs could be designed to avoid such a scenario, i.e.,
   restricted to exactly one negotiation session at a time for a given
   objective, but this would be a voluntary restriction not required by
   the GRASP protocol.  In fact, GRASP assumes that any ASA managing a
   resource may need to conduct multiple parallel negotiations, possibly
   with the same peer.  Communication patterns could be very complex,
   with a group of ASAs overlapping negotiations among themselves, as
   described in [ANIMA-COORD].  Therefore, the API design allows for
   such scenarios.

   In the callback model, for the scenario just described, the ASAs "A"
   and "B" will each provide two instances of the callback function, one
   for each session.  For this reason, each ASA must be able to
   distinguish the two sessions, and the peer's IP address is not
   sufficient for this.  It is also not safe to rely on transport port
   numbers for this, since future variants of GRASP might use shared
   ports rather than a separate port per session.  Hence, the GRASP
   design includes a Session ID.  Thus, when necessary, a session handle
   (see the next section) is used in the API to distinguish simultaneous
   GRASP sessions from each other, so that any number of sessions may
   proceed asynchronously in parallel.

2.2.3.  Overlapping Sessions and Operations

   A GRASP session consists of a finite sequence of messages (for
   discovery, synchronization, or negotiation) between two ASAs.  It is
   uniquely identified on the wire by a pseudorandom Session ID plus the
   IP address of the initiator of the session.  Further details are
   given in "Session Identifier (Session ID)" (Section 2.7 of
   [RFC8990]).

   On the first call in a new GRASP session, the API returns a
   'session_handle' handle that uniquely identifies the session within
   the API, so that multiple overlapping sessions can be distinguished.
   A likely implementation is to form the handle from the underlying
   GRASP Session ID and IP address.  This handle must be used in all
   subsequent calls for the same session.  Also see Section 2.3.2.8.

   An additional mechanism that might increase efficiency for polling
   implementations is to add a general call, say notify(), which would
   check the status of all outstanding operations for the calling ASA
   and return the session_handle values for all sessions that have
   changed state.  This would eliminate the need for repeated calls to
   the individual functions returning a 'noReply'.  This call is not
   described below as the details are likely to be implementation
   specific.

   An implication of the above for all GRASP implementations is that the
   GRASP core must keep state for each GRASP operation in progress, most
   likely keyed by the GRASP Session ID and the GRASP source address of
   the session initiator.  Even in a threaded implementation, the GRASP
   core will need such state internally.  The session_handle parameter
   exposes this aspect of the implementation.

2.2.4.  Session Termination

   GRASP sessions may terminate for numerous reasons.  A session ends
   when discovery succeeds or times out, negotiation succeeds or fails,
   a synchronization result is delivered, the other end fails to respond
   before a timeout expires, a loop count expires, or a network socket
   error occurs.  Note that a timeout at one end of a session might
   result in a timeout or a socket error at the other end, since GRASP
   does not send error messages in this case.  In all cases, the API
   will return an appropriate code to the caller, which should then
   release any reserved resources.  After failure cases, the GRASP
   specification recommends an exponential backoff before retrying.

2.3.  API Definition

2.3.1.  Overview of Functions

   The functions provided by the API fall into several groups:

   Registration:  These functions allow an ASA to register itself with
      the GRASP core and allow a registered ASA to register the GRASP
      objectives that it will manipulate.

   Discovery:  This function allows an ASA that needs to initiate
      negotiation or synchronization of a particular objective to
      discover a peer willing to respond.

   Negotiation:  These functions allow an ASA to act as an initiator
      (requester) or responder (listener) for a GRASP negotiation
      session.  After initiation, negotiation is a symmetric process, so
      most of the functions can be used by either party.

   Synchronization:  These functions allow an ASA to act as an initiator
      (requester) or responder (listener and data source) for a GRASP
      synchronization session.

   Flooding:  These functions allow an ASA to send and receive an
      objective that is flooded to all nodes of the ACP.

   Some example logic flows for a resource management ASA are given in
   [ASA-GUIDE], which may be of help in understanding the following
   descriptions.  The next section describes parameters and data
   structures used in multiple API calls.  The following sections
   describe various groups of function APIs.  Those APIs that do not
   list asynchronous mechanisms are implicitly synchronous in their
   behavior.

2.3.2.  Parameters and Data Structures

2.3.2.1.  Integers

   In this API, integers are assumed to be 32-bit unsigned integers
   (uint32_t) unless otherwise indicated.

2.3.2.2.  Errorcode

   All functions in the API have an unsigned 'errorcode' integer as
   their return value (the first return value in languages that allow
   multiple return values).  An errorcode of zero indicates success.
   Any other value indicates failure of some kind.  The first three
   errorcodes have special importance:

   1 - Declined:  used to indicate that the other end has sent a GRASP
      Negotiation End message (M_END) with a Decline option (O_DECLINE).

   2 - No reply:  used in non-blocking calls to indicate that the other
      end has sent no reply so far (see Section 2.2).

   3 - Unspecified error:  used when no more specific error codes apply.

   Appendix A gives a full list of currently suggested error codes,
   based on implementation experience.  While there is no absolute
   requirement for all implementations to use the same error codes, this
   is highly recommended for portability of applications.

2.3.2.3.  Timeout

   Whenever a 'timeout' parameter appears, it is an unsigned integer
   expressed in milliseconds.  If it is zero, the GRASP default timeout
   (GRASP_DEF_TIMEOUT; see [RFC8990]) will apply.  An exception is the
   discover() function, which has a different interpretation of a zero
   timeout.  If no response is received before the timeout expires, the
   call will fail unless otherwise noted.

2.3.2.4.  Objective

   An 'objective' parameter is a data structure with the following
   components:

   name (UTF-8 string):  The objective's name

   neg (Boolean flag):  True if objective supports negotiation (default
      False)

   synch (Boolean flag):  True if objective supports synchronization
      (default False)

   dry (Boolean flag):  True if objective supports dry-run negotiation
      (default False)

      Note 1:  Only one of 'synch' or 'neg' may be True.
      Note 2:  'dry' must not be True unless 'neg' is also True.
      Note 3:  In some programming languages, the preferred
         implementation may be to represent the Boolean flags as bits in
         a single byte, which is how they are encoded in GRASP messages.
         In other languages, an enumeration might be preferable.

   loop_count (unsigned integer, uint8_t):  Limit on negotiation steps,
      etc. (default GRASP_DEF_LOOPCT; see [RFC8990]).  The 'loop_count'
      is set to a suitable value by the initiator of a negotiation, to
      prevent indefinite loops.  It is also used to limit the
      propagation of discovery and flood messages.

   value:  A specific data structure expressing the value of the
      objective.  The format is language dependent, with the constraint
      that it can be validly represented in CBOR [RFC8949].

         An important advantage of CBOR is that the value of an
         objective can be completely opaque to the GRASP core yet pass
         transparently through it to and from the ASA.  Although the
         GRASP core must validate the format and syntax of GRASP
         messages, it cannot validate the value of an objective; all it
         can do is detect malformed CBOR.  The handling of decoding
         errors depends on the CBOR library in use, but a corresponding
         error code ('CBORfail') is defined in the API and will be
         returned to the ASA if a faulty message can be assigned to a
         current GRASP session.  However, it is the responsibility of
         each ASA to validate the value of a received objective, as
         discussed in Section 5.3 of [RFC8949].  If the programming
         language in use is suitably object-oriented, the GRASP API may
         deserialize the value and present it to the ASA as an object.
         If not, it will be presented as a CBOR data item.  In all
         cases, the syntax and semantics of the objective value are the
         responsibility of the ASA.

         A requirement for all language mappings and all API
         implementations is that, regardless of what other options exist
         for a language-specific representation of the value, there is
         always an option to use a raw CBOR data item as the value.  The
         API will then wrap this with CBOR Tag 24 as an encoded CBOR
         data item for transmission via GRASP, and unwrap it after
         reception.  By this means, ASAs will be able to communicate
         regardless of programming language.

   The 'name' and 'value' fields are of variable length.  GRASP does not
   set a maximum length for these fields, but only for the total length
   of a GRASP message.  Implementations might impose length limits.

   An example data structure definition for an objective in the C
   language, using at least the C99 version, and assuming the use of a
   particular CBOR library [libcbor], is:

    typedef struct {
       unsigned char *name;
       uint8_t flags;            // flag bits as defined by GRASP
       uint8_t loop_count;
       uint32_t value_size;      // size of value in bytes
       cbor_mutable_data cbor_value;
                                // CBOR bytestring (libcbor/cbor/data.h)
                    } objective;

   An example data structure definition for an objective in the Python
   language (version 3.4 or later) is:

    class objective:
       """A GRASP objective"""
       def __init__(self, name):
           self.name = name        #Unique name (string)
           self.negotiate = False  #True if negotiation supported
           self.dryrun = False     #True if dry-run supported
           self.synch = False      #True if synchronization supported
           self.loop_count = GRASP_DEF_LOOPCT  # Default starting value
           self.value = None       #Place holder; any Python object

2.3.2.5.  asa_locator

   An 'asa_locator' parameter is a data structure with the following
   contents:

   locator:  The actual locator, either an IP address or an ASCII
      string.

   ifi (unsigned integer):  The interface identifier index via which
      this was discovered (of limited use to most ASAs).

   expire (system dependent type):  The time on the local system clock
      when this locator will expire from the cache.

   The following covers all locator types currently supported by
   GRASP:
      *  is_ipaddress (Boolean) - True if the locator is an IP address.

      *  is_fqdn (Boolean) - True if the locator is a Fully Qualified
         Domain Name (FQDN).

      *  is_uri (Boolean) - True if the locator is a URI.

      These options are mutually exclusive.  Depending on the
      programming language, they could be represented as a bit pattern
      or an enumeration.

   diverted (Boolean):  True if the locator was discovered via a Divert
      option.

   protocol (unsigned integer):  Applicable transport protocol
      (IPPROTO_TCP or IPPROTO_UDP).  These constants are defined in the
      CDDL specification of GRASP [RFC8990].

   port (unsigned integer):  Applicable port number.

   The 'locator' field is of variable length in the case of an FQDN or a
   URI.  GRASP does not set a maximum length for this field, but only
   for the total length of a GRASP message.  Implementations might
   impose length limits.

   It should be noted that when one ASA discovers the asa_locator of
   another, there is no explicit authentication mechanism.  In
   accordance with the trust model provided by the secure ACP, ASAs are
   presumed to provide correct locators in response to discovery.  See
   "Locator Options" (Section 2.9.5 of [RFC8990]) for further details.

2.3.2.6.  Tagged_objective

   A 'tagged_objective' parameter is a data structure with the following
   contents:

   objective:  An objective.

   locator:  The asa_locator associated with the objective, or a null
      value.

2.3.2.7.  asa_handle

   Although an authentication and authorization scheme for ASAs has not
   been defined, the API provides a very simple hook for such a scheme.
   When an ASA starts up, it registers itself with the GRASP core, which
   provides it with an opaque handle that, although not
   cryptographically protected, would be difficult for a third party to
   predict.  The ASA must present this handle in future calls.  This
   mechanism will prevent some elementary errors or trivial attacks such
   as an ASA manipulating an objective it has not registered to use.

   Thus, in most calls, an 'asa_handle' parameter is required.  It is
   generated when an ASA first registers with GRASP, and the ASA must
   then store the asa_handle and use it in every subsequent GRASP call.
   Any call in which an invalid handle is presented will fail.  It is an
   up to 32-bit opaque value (for example, represented as a uint32_t,
   depending on the language).  Since it is only used locally, and not
   in GRASP messages, it is only required to be unique within the local
   GRASP instance.  It is valid until the ASA terminates.  It should be
   unpredictable; a possible implementation is to use the same mechanism
   that GRASP uses to generate Session IDs (see Section 2.3.2.8).

2.3.2.8.  Session_handle and Callbacks

   In some calls, a 'session_handle' parameter is required.  This is an
   opaque data structure as far as the ASA is concerned, used to
   identify calls to the API as belonging to a specific GRASP session
   (see Section 2.2.3).  It will be provided as a parameter in callback
   functions.  As well as distinguishing calls from different sessions,
   it also allows GRASP to detect and ignore calls from non-existent or
   timed-out sessions.

   In an event loop implementation, callback functions (Section 2.2.1)
   may be supported for all API functions that involve waiting for a
   remote operation:

      discover() whose callback would be discovery_received().

      request_negotiate() whose callback would be
      negotiate_step_received().

      negotiate_step() whose callback would be
      negotiate_step_received().

      listen_negotiate() whose callback would be
      negotiate_step_received().

      synchronize() whose callback would be synchronization_received().

   Further details of callbacks are implementation dependent.

2.3.3.  Registration

   These functions are used to register an ASA, and the objectives that
   it modifies, with the GRASP module.  In the absence of an
   authorization model, these functions are very simple, but they will
   avoid multiple ASAs choosing the same name and will prevent multiple
   ASAs manipulating the same objective.  If an authorization model is
   added to GRASP, these API calls would need to be modified
   accordingly.

   *  register_asa()

      All ASAs must use this call before issuing any other API calls.

      -  Input parameter:

            name of the ASA (UTF-8 string)

      -  Return value:

            errorcode (unsigned integer)

            asa_handle (unsigned integer)

      -  This initializes the state in the GRASP module for the calling
         entity (the ASA).  In the case of success, an 'asa_handle' is
         returned, which the ASA must present in all subsequent calls.
         In the case of failure, the ASA has not been authorized and
         cannot operate.  The 'asa_handle' value is undefined.

   *  deregister_asa()

      -  Input parameters:

            asa_handle (unsigned integer)

            name of the ASA (UTF-8 string)

      -  Return value:

            errorcode (unsigned integer)

      -  This removes all state in the GRASP module for the calling
         entity (the ASA) and deregisters any objectives it has
         registered.  Note that these actions must also happen
         automatically if an ASA exits.

      -  Note -- the ASA name is, strictly speaking, redundant in this
         call but is present to detect and reject erroneous
         deregistrations.

   *  register_objective()

      ASAs must use this call for any objective whose value they need to
      transmit by negotiation, synchronization, or flooding.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

            ttl (unsigned integer -- default GRASP_DEF_TIMEOUT)

            discoverable (Boolean -- default False)

            overlap (Boolean -- default False)

            local (Boolean -- default False)

      -  Return value:

            errorcode (unsigned integer)

      -  This registers an objective that this ASA may modify and
         transmit to other ASAs by flooding or negotiation.  It is not
         necessary to register an objective that is only received by
         GRASP synchronization or flooding.  The 'objective' becomes a
         candidate for discovery.  However, discovery responses should
         not be enabled until the ASA calls listen_negotiate() or
         listen_synchronize(), showing that it is able to act as a
         responder.  The ASA may negotiate the objective or send
         synchronization or flood data.  Registration is not needed for
         "read-only" operations, i.e., the ASA only wants to receive
         synchronization or flooded data for the objective concerned.

      -  The 'ttl' parameter is the valid lifetime (time to live) in
         milliseconds of any discovery response generated for this
         objective.  The default value should be the GRASP default
         timeout (GRASP_DEF_TIMEOUT; see [RFC8990]).

      -  If the parameter 'discoverable' is True, the objective is
         immediately discoverable.  This is intended for objectives that
         are only defined for GRASP discovery and that do not support
         negotiation or synchronization.

      -  If the parameter 'overlap' is True, more than one ASA may
         register this objective in the same GRASP instance.  This is of
         value for life cycle management of ASAs [ASA-GUIDE] and must be
         used consistently for a given objective (always True or always
         False).

      -  If the parameter 'local' is True, discovery must return a link-
         local address.  This feature is for objectives that must be
         restricted to the local link.

      -  This call may be repeated for multiple objectives.

   *  deregister_objective()

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

      -  Return value:

            errorcode (unsigned integer)

      -  The 'objective' must have been registered by the calling ASA;
         if not, this call fails.  Otherwise, it removes all state in
         the GRASP module for the given objective.

2.3.4.  Discovery

   *  discover()

      This function may be used by any ASA to discover peers handling a
      given objective.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

            timeout (unsigned integer)

            minimum_TTL (unsigned integer)

      -  Return values:

            errorcode (unsigned integer)

            locator_list (structure)

      -  This returns a list of discovered 'asa_locators' for the given
         objective.  An empty list means that no locators were
         discovered within the timeout.  Note that this structure
         includes all the fields described in Section 2.3.2.5.

      -  The parameter 'minimum_TTL' must be greater than or equal to
         zero.  Any locally cached locators for the objective whose
         remaining time to live in milliseconds is less than or equal to
         'minimum_TTL' are deleted first.  Thus, 'minimum_TTL' = 0 will
         flush all entries.  Note that this will not affect sessions
         already in progress using the deleted locators.

      -  If the parameter 'timeout' is zero, any remaining locally
         cached locators for the objective are returned immediately, and
         no other action is taken.  (Thus, a call with 'minimum_TTL' and
         'timeout' both equal to zero is pointless.)

      -  If the parameter 'timeout' is greater than zero, GRASP
         discovery is performed, and all results obtained before the
         timeout in milliseconds expires are returned.  If no results
         are obtained, an empty list is returned after the timeout.
         That is not an error condition.  GRASP discovery is not a
         deterministic process.  If there are multiple nodes handling an
         objective, none, some, or all of them will be discovered before
         the timeout expires.

      -  Asynchronous Mechanisms:

         Threaded implementation:  This should be called in a separate
            thread if asynchronous operation is required.

         Event loop implementation:  An additional in/out
            'session_handle' parameter is used.  If the 'errorcode'
            parameter has the value 2 ('noReply'), no response has been
            received so far.  The 'session_handle' parameter must be
            presented in subsequent calls.  A callback may be used in
            the case of a non-zero timeout.

2.3.5.  Negotiation

   Since the negotiation mechanism is different from a typical client/
   server exchange, Figure 2 illustrates the sequence of calls and GRASP
   messages in a negotiation.  Note that after the first protocol
   exchange, the process is symmetrical, with negotiating steps strictly
   alternating between the two sides.  Either side can end the
   negotiation.  Also, the side that is due to respond next can insert a
   delay at any time, to extend the other side's timeout.  This would be
   used, for example, if an ASA needed to negotiate with a third party
   before continuing with the current negotiation.

   The loop count embedded in the objective that is the subject of
   negotiation is initialized by the ASA that starts a negotiation and
   is then decremented by the GRASP core at each step, prior to sending
   each M_NEGOTIATE message.  If it reaches zero, the negotiation will
   fail, and each side will receive an error code.

Initiator                         Responder
---------                         ---------

                                  listen_negotiate() \ Await request

request_negotiate()
          M_REQ_NEG      ->       negotiate_step()   \ Open session,
                         <-      M_NEGOTIATE         / start negotiation
negotiate_step()
        M_NEGOTIATE      ->       negotiate_step()   \ Continue
                         <-      M_NEGOTIATE         / negotiation
                         ...
negotiate_wait()                                     \ Insert
        M_WAIT           ->                          / delay
negotiate_step()
        M_NEGOTIATE      ->       negotiate_step()   \ Continue
                         <-      M_NEGOTIATE         / negotiation
negotiate_step()
        M_NEGOTIATE      ->       end_negotiate()    \ End
                         <-      M_END               / negotiation

                                                     \ Process results

                    Figure 2: Negotiation Sequence

   As the negotiation proceeds, each side will update the value of the
   objective in accordance with its particular semantics, defined in the
   specification of the objective.  Although many objectives will have
   values that can be ordered, so that negotiation can be a simple
   bidding process, it is not a requirement.

   Failure to agree, a timeout, or loop count exhaustion may all end a
   negotiation session, but none of these cases are protocol failures.

   *  request_negotiate()

      This function is used by any ASA to initiate negotiation of a
      GRASP objective as a requester (client).

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

            peer (asa_locator)

            timeout (unsigned integer)

      -  Return values:

            errorcode (unsigned integer)

            session_handle (structure) (undefined unless successful)

            proffered_objective (structure) (undefined unless
            successful)

            reason (string) (empty unless negotiation declined)

      -  This function opens a negotiation session between two ASAs.
         Note that GRASP currently does not support multiparty
         negotiation, which would need to be added as an extended
         function.

      -  The 'objective' parameter must include the requested value, and
         its loop count should be set to a suitable starting value by
         the ASA.  If not, the GRASP default will apply.

      -  Note that a given negotiation session may or may not be a dry-
         run negotiation; the two modes must not be mixed in a single
         session.

      -  The 'peer' parameter is the target node; it must be an
         'asa_locator' as returned by discover().  If 'peer' is null,
         GRASP discovery is automatically performed first to find a
         suitable peer (i.e., any node that supports the objective in
         question).

      -  The 'timeout' parameter is described in Section 2.3.2.3.

      -  If the 'errorcode' return value is 0, the negotiation has
         successfully started.  There are then two cases:

         1.  The 'session_handle' parameter is null.  In this case, the
             negotiation has succeeded with one exchange of messages,
             and the peer has accepted the request.  The returned
             'proffered_objective' contains the value accepted by the
             peer, which is therefore equal to the value in the
             requested 'objective'.  For this reason, no session handle
             is needed, since the session has ended.

         2.  The 'session_handle' parameter is not null.  In this case,
             negotiation must continue.  The 'session_handle' must be
             presented in all subsequent negotiation steps.  The
             returned 'proffered_objective' contains the first value
             proffered by the negotiation peer in the first exchange of
             messages; in other words, it is a counter-offer.  The
             contents of this instance of the objective must be used to
             prepare the next negotiation step (see negotiate_step()
             below) because it contains the updated loop count, sent by
             the negotiation peer.  The GRASP code automatically
             decrements the loop count by 1 at each step and returns an
             error if it becomes zero.  Since this terminates the
             negotiation, the other end will experience a timeout, which
             will terminate the other end of the session.

             This function must be followed by calls to 'negotiate_step'
             and/or 'negotiate_wait' and/or 'end_negotiate' until the
             negotiation ends. 'request_negotiate' may then be called
             again to start a new negotiation.

      -  If the 'errorcode' parameter has the value 1 ('declined'), the
         negotiation has been declined by the peer (M_END and O_DECLINE
         features of GRASP).  The 'reason' string is then available for
         information and diagnostic use, but it may be a null string.
         For this and any other error code, an exponential backoff is
         recommended before any retry (see Section 3).

      -  Asynchronous Mechanisms:

         Threaded implementation:  This should be called in a separate
            thread if asynchronous operation is required.

         Event loop implementation:  The 'session_handle' parameter is
            used to distinguish multiple simultaneous sessions.  If the
            'errorcode' parameter has the value 2 ('noReply'), no
            response has been received so far.  The 'session_handle'
            parameter must be presented in subsequent calls.

      -  Use of dry-run mode must be consistent within a GRASP session.
         The state of the 'dry' flag in the initial request_negotiate()
         call must be the same in all subsequent negotiation steps of
         the same session.  The semantics of the dry-run mode are built
         into the ASA; GRASP merely carries the flag bit.

      -  Special note for the ACP infrastructure ASA: It is likely that
         this ASA will need to discover and negotiate with its peers in
         each of its on-link neighbors.  It will therefore need to know
         not only the link-local IP address but also the physical
         interface and transport port for connecting to each neighbor.
         One implementation approach to this is to include these details
         in the 'session_handle' data structure, which is opaque to
         normal ASAs.

   *  listen_negotiate()

      This function is used by an ASA to start acting as a negotiation
      responder (listener) for a given GRASP objective.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

      -  Return values:

            errorcode (unsigned integer)

            session_handle (structure) (undefined unless successful)

            requested_objective (structure) (undefined unless
            successful)

      -  This function instructs GRASP to listen for negotiation
         requests for the given 'objective'.  It also enables discovery
         responses for the objective, as mentioned under
         register_objective() in Section 2.3.3.

      -  Asynchronous Mechanisms:

         Threaded implementation:  It will block waiting for an incoming
            request, so it should be called in a separate thread if
            asynchronous operation is required.  Unless there is an
            unexpected failure, this call only returns after an incoming
            negotiation request.  If the ASA supports multiple
            simultaneous transactions, a new sub-thread must be spawned
            for each new session, so that listen_negotiate() can be
            called again immediately.

         Event loop implementation:  A 'session_handle' parameter is
            used to distinguish individual sessions.  If the ASA
            supports multiple simultaneous transactions, a new event
            must be inserted in the event loop for each new session, so
            that listen_negotiate() can be reactivated immediately.

      -  This call only returns (threaded model) or triggers (event
         loop) after an incoming negotiation request.  When this occurs,
         'requested_objective' contains the first value requested by the
         negotiation peer.  The contents of this instance of the
         objective must be used in the subsequent negotiation call
         because it contains the loop count sent by the negotiation
         peer.  The 'session_handle' must be presented in all subsequent
         negotiation steps.

      -  This function must be followed by calls to 'negotiate_step'
         and/or 'negotiate_wait' and/or 'end_negotiate' until the
         negotiation ends.

      -  If an ASA is capable of handling multiple negotiations
         simultaneously, it may call 'listen_negotiate' simultaneously
         from multiple threads, or insert multiple events.  The API and
         GRASP implementation must support re-entrant use of the
         listening state and the negotiation calls.  Simultaneous
         sessions will be distinguished by the threads or events
         themselves, the GRASP session handles, and the underlying
         unicast transport sockets.

   *  stop_listen_negotiate()

      This function is used by an ASA to stop acting as a responder
      (listener) for a given GRASP objective.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

      -  Return value:

            errorcode (unsigned integer)

      -  Instructs GRASP to stop listening for negotiation requests for
         the given objective, i.e., cancels 'listen_negotiate'.

      -  Asynchronous Mechanisms:

         Threaded implementation:  Must be called from a different
            thread than 'listen_negotiate'.

         Event loop implementation:  No special considerations.

   *  negotiate_step()

      This function is used by either ASA in a negotiation session to
      make the next step in negotiation.

      -  Input parameters:

            asa_handle (unsigned integer)

            session_handle (structure)

            objective (structure)

            timeout (unsigned integer) as described in Section 2.3.2.3

      -  Return values:

            Exactly as for 'request_negotiate'

      -  Executes the next negotiation step with the peer.  The
         'objective' parameter contains the next value being proffered
         by the ASA in this step.  It must also contain the latest
         'loop_count' value received from request_negotiate() or
         negotiate_step().

      -  Asynchronous Mechanisms:

         Threaded implementation:  Usually called in the same thread as
            the preceding 'request_negotiate' or 'listen_negotiate',
            with the same value of 'session_handle'.

         Event loop implementation:  Must use the same value of
            'session_handle' returned by the preceding
            'request_negotiate' or 'listen_negotiate'.

   *  negotiate_wait()

      This function is used by either ASA in a negotiation session to
      delay the next step in negotiation.

      -  Input parameters:

            asa_handle (unsigned integer)

            session_handle (structure)

            timeout (unsigned integer)

      -  Return value:

            errorcode (unsigned integer)

      -  Requests the remote peer to delay the negotiation session by
         'timeout' milliseconds, thereby extending the original timeout.
         This function simply triggers a GRASP Confirm Waiting message
         (see [RFC8990] for details).

      -  Asynchronous Mechanisms:

         Threaded implementation:  Called in the same thread as the
            preceding 'request_negotiate' or 'listen_negotiate', with
            the same value of 'session_handle'.

         Event loop implementation:  Must use the same value of
            'session_handle' returned by the preceding
            'request_negotiate' or 'listen_negotiate'.

   *  end_negotiate()

      This function is used by either ASA in a negotiation session to
      end a negotiation.

      -  Input parameters:

            asa_handle (unsigned integer)

            session_handle (structure)

            result (Boolean)

            reason (UTF-8 string)

      -  Return value:

            errorcode (unsigned integer)

      -  End the negotiation session:

         'result' = True for accept (successful negotiation), and False
         for decline (failed negotiation).

         'reason' = string describing reason for decline (may be null;
         ignored if accept).

      -  Asynchronous Mechanisms:

         Threaded implementation:  Called in the same thread as the
            preceding 'request_negotiate' or 'listen_negotiate', with
            the same value of 'session_handle'.

         Event loop implementation:  Must use the same value of
            'session_handle' returned by the preceding
            'request_negotiate' or 'listen_negotiate'.

2.3.6.  Synchronization and Flooding

   *  synchronize()

      This function is used by any ASA to cause synchronization of a
      GRASP objective as a requester (client).

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

            peer (asa_locator)

            timeout (unsigned integer)

      -  Return values:

            errorcode (unsigned integer)

            result (structure) (undefined unless successful)

      -  This call requests the synchronized value of the given
         'objective'.

      -  If the 'peer' parameter is null, and the objective is already
         available in the local cache, the flooded objective is returned
         immediately in the 'result' parameter.  In this case, the
         'timeout' is ignored.

      -  If the 'peer' parameter is not null, or a cached value is not
         available, synchronization with a discovered ASA is performed.
         If successful, the retrieved objective is returned in the
         'result' value.

      -  The 'peer' parameter is an 'asa_locator' as returned by
         discover().  If 'peer' is null, GRASP discovery is
         automatically performed first to find a suitable peer (i.e.,
         any node that supports the objective in question).

      -  The 'timeout' parameter is described in Section 2.3.2.3.

      -  This call should be repeated whenever the latest value is
         needed.

      -  Asynchronous Mechanisms:

         Threaded implementation:  Call in a separate thread if
            asynchronous operation is required.

         Event loop implementation:  An additional in/out
            'session_handle' parameter is used, as in
            request_negotiate().  If the 'errorcode' parameter has the
            value 2 ('noReply'), no response has been received so far.
            The 'session_handle' parameter must be presented in
            subsequent calls.

      -  In the case of failure, an exponential backoff is recommended
         before retrying (Section 3).

   *  listen_synchronize()

      This function is used by an ASA to start acting as a
      synchronization responder (listener) for a given GRASP objective.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

      -  Return value:

            errorcode (unsigned integer)

      -  This instructs GRASP to listen for synchronization requests for
         the given objective and to respond with the value given in the
         'objective' parameter.  It also enables discovery responses for
         the objective, as mentioned under register_objective() in
         Section 2.3.3.

      -  This call is non-blocking and may be repeated whenever the
         value changes.

   *  stop_listen_synchronize()

      This function is used by an ASA to stop acting as a
      synchronization responder (listener) for a given GRASP objective.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

      -  Return value:

            errorcode (unsigned integer)

      -  This call instructs GRASP to stop listening for synchronization
         requests for the given 'objective', i.e., it cancels a previous
         listen_synchronize.

   *  flood()

      This function is used by an ASA to flood one or more GRASP
      objectives throughout the Autonomic Network.

      Note that each GRASP node caches all flooded objectives that it
      receives, until each one's time to live expires.  Cached
      objectives are tagged with their origin as well as an expiry time,
      so multiple copies of the same objective may be cached
      simultaneously.  Further details are given in "Flood
      Synchronization Message" (Section 2.8.11 of [RFC8990]).

      -  Input parameters:

            asa_handle (unsigned integer)

            ttl (unsigned integer)

            tagged_objective_list (structure)

      -  Return value:

            errorcode (unsigned integer)

      -  This call instructs GRASP to flood the given synchronization
         objective(s) and their value(s) and associated locator(s) to
         all GRASP nodes.

      -  The 'ttl' parameter is the valid lifetime (time to live) of the
         flooded data in milliseconds (0 = infinity).

      -  The 'tagged_objective_list' parameter is a list of one or more
         'tagged_objective' couplets.  The 'locator' parameter that tags
         each objective is normally null but may be a valid
         'asa_locator'.  Infrastructure ASAs needing to flood an
         {address, protocol, port} 3-tuple with an objective create an
         asa_locator object to do so.  If the IP address in that locator
         is the unspecified address ('::'), it is replaced by the link-
         local address of the sending node in each copy of the flood
         multicast, which will be forced to have a loop count of 1.
         This feature is for objectives that must be restricted to the
         local link.

      -  The function checks that the ASA registered each objective.

      -  This call may be repeated whenever any value changes.

   *  get_flood()

      This function is used by any ASA to obtain the current value of a
      flooded GRASP objective.

      -  Input parameters:

            asa_handle (unsigned integer)

            objective (structure)

      -  Return values:

            errorcode (unsigned integer)

            tagged_objective_list (structure) (undefined unless
            successful)

      -  This call instructs GRASP to return the given synchronization
         objective if it has been flooded and its lifetime has not
         expired.

      -  The 'tagged_objective_list' parameter is a list of
         'tagged_objective' couplets, each one being a copy of the
         flooded objective and a corresponding locator.  Thus, if the
         same objective has been flooded by multiple ASAs, the recipient
         can distinguish the copies.

      -  Note that this call is for advanced ASAs.  In a simple case, an
         ASA can simply call synchronize() in order to get a valid
         flooded objective.

   *  expire_flood()

      This function may be used by an ASA to expire specific entries in
      the local GRASP flood cache.

      -  Input parameters:

            asa_handle (unsigned integer)

            tagged_objective (structure)

      -  Return value:

            errorcode (unsigned integer)

      -  This is a call that can only be used after a preceding call to
         get_flood() by an ASA that is capable of deciding that the
         flooded value is stale or invalid.  Use with care.

      -  The 'tagged_objective' parameter is the one to be expired.

2.3.7.  Invalid Message Function

   *  send_invalid()

      This function may be used by any ASA to stop an ongoing GRASP
      session.

      -  Input parameters:

            asa_handle (unsigned integer)

            session_handle (structure)

            info (bytes)

      -  Return value:

            errorcode (unsigned integer)

      -  Sends a GRASP Invalid message (M_INVALID), as described in
         [RFC8990].  It should not be used if end_negotiate() would be
         sufficient.  Note that this message may be used in response to
         any unicast GRASP message that the receiver cannot interpret
         correctly.  In most cases, this message will be generated
         internally by a GRASP implementation.

         'info' = optional diagnostic data supplied by the ASA.  It may
         be raw bytes from the invalid message.

3.  Security Considerations

   Security considerations for the GRASP protocol are discussed in
   [RFC8990].  These include denial-of-service issues, even though these
   are considered a low risk in the ACP.  In various places, GRASP
   recommends an exponential backoff.  An ASA using the API should use
   exponential backoff after failed discover(), req_negotiate(), or
   synchronize() operations.  The timescale for such backoffs depends on
   the semantics of the GRASP objective concerned.  Additionally, a
   flood() operation should not be repeated at shorter intervals than is
   useful.  The appropriate interval depends on the semantics of the
   GRASP objective concerned.  These precautions are intended to assist
   the detection of denial-of-service attacks.

   As a general precaution, all ASAs able to handle multiple negotiation
   or synchronization requests in parallel may protect themselves
   against a denial-of-service attack by limiting the number of requests
   they handle simultaneously and silently discarding excess requests.
   It might also be useful for the GRASP core to limit the number of
   objectives registered by a given ASA, the total number of ASAs
   registered, and the total number of simultaneous sessions, to protect
   system resources.  During times of high autonomic activity, such as
   recovery from widespread faults, ASAs may experience many GRASP
   session failures.  Guidance on making ASAs suitably robust is given
   in [ASA-GUIDE].

   As noted earlier, the trust model is that all ASAs in a given
   Autonomic Network communicate via a secure autonomic control plane;
   therefore, they trust each other's messages.  Specific authorization
   of ASAs to use particular GRASP objectives is a subject for future
   study, also briefly discussed in [RFC8990].

   The careful reader will observe that a malicious ASA could extend a
   negotiation session indefinitely by use of the negotiate_wait()
   function or by manipulating the loop count of an objective.  A
   robustly implemented ASA could detect such behavior by a peer and
   break off negotiation.

   The 'asa_handle' is used in the API as a first line of defense
   against a malware process attempting to imitate a legitimately
   registered ASA.  The 'session_handle' is used in the API as a first
   line of defense against a malware process attempting to hijack a
   GRASP session.  Both these handles are likely to be created using
   GRASP's 32-bit pseudorandom Session ID.  By construction, GRASP
   avoids the risk of Session ID collisions (see "Session Identifier
   (Session ID)", Section 2.7 of [RFC8990]).  There remains a finite
   probability that an attacker could guess a Session ID,
   session_handle, or asa_handle.  However, this would only be of value
   to an attacker that had already penetrated the ACP, which would allow
   many other simpler forms of attack than hijacking GRASP sessions.

4.  IANA Considerations

   This document has no IANA actions.

5.  References

5.1.  Normative References

   [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,
              <https://www.rfc-editor.org/info/rfc8949>.

   [RFC8990]  Bormann, C., Carpenter, B., Ed., and B. Liu, Ed., "GeneRic
              Autonomic Signaling Protocol (GRASP)", RFC 8990,
              DOI 10.17487/RFC8990, May 2021,
              <https://www.rfc-editor.org/info/rfc8990>.

5.2.  Informative References

   [ANIMA-COORD]
              Ciavaglia, L. and P. Peloso, "Autonomic Functions
              Coordination", Work in Progress, Internet-Draft, draft-
              ciavaglia-anima-coordination-01, 21 March 2016,
              <https://tools.ietf.org/html/draft-ciavaglia-anima-
              coordination-01>.

   [ASA-GUIDE]
              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-
              ietf-anima-asa-guidelines-00>.

   [GRASP-DISTRIB]
              Liu, B., Xiao, X., Hecker, A., Jiang, S., Despotovic, Z.,
              and B. Carpenter, "Information Distribution over GRASP",
              Work in Progress, Internet-Draft, draft-ietf-anima-grasp-
              distribution-02, 8 March 2021,
              <https://tools.ietf.org/html/draft-ietf-anima-grasp-
              distribution-02>.

   [libcbor]  Kalvoda, P., "libcbor - libcbor 0.8.0 documentation",
              April 2021, <https://libcbor.readthedocs.io/>.

   [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,
              <https://www.rfc-editor.org/info/rfc8993>.

   [RFC8994]  Eckert, T., Ed., Behringer, M., Ed., and S. Bjarnason, "An
              Autonomic Control Plane (ACP)", RFC 8994,
              DOI 10.17487/RFC8994, May 2021,
              <https://www.rfc-editor.org/info/rfc8994>.

   [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.  Error Codes

   This appendix lists the error codes defined so far on the basis of
   implementation experience, with suggested symbolic names and
   corresponding descriptive strings in English.  It is expected that
   complete API implementations will provide for localization of these
   descriptive strings, and that additional error codes will be needed
   according to implementation details.

   The error codes that may only be returned by one or two functions are
   annotated accordingly, and the others may be returned by numerous
   functions.  The 'noSecurity' error will be returned to most calls if
   GRASP is running in an insecure mode (i.e., with no secure substrate
   such as the ACP), except for the specific DULL usage mode described
   in "Discovery Unsolicited Link-Local (DULL) GRASP" (Section 2.5.2 of
   [RFC8990].

       +================+=======+=================================+
       | Name           | Error | Description                     |
       |                | Code  |                                 |
       +================+=======+=================================+
       | ok             | 0     | "OK"                            |
       +----------------+-------+---------------------------------+
       | declined       | 1     | "Declined" (req_negotiate,      |
       |                |       | negotiate_step)                 |
       +----------------+-------+---------------------------------+
       | noReply        | 2     | "No reply" (indicates waiting   |
       |                |       | state in event loop calls)      |
       +----------------+-------+---------------------------------+
       | unspec         | 3     | "Unspecified error"             |
       +----------------+-------+---------------------------------+
       | ASAfull        | 4     | "ASA registry full"             |
       |                |       | (register_asa)                  |
       +----------------+-------+---------------------------------+
       | dupASA         | 5     | "Duplicate ASA name"            |
       |                |       | (register_asa)                  |
       +----------------+-------+---------------------------------+
       | noASA          | 6     | "ASA not registered"            |
       +----------------+-------+---------------------------------+
       | notYourASA     | 7     | "ASA registered but not by you" |
       |                |       | (deregister_asa)                |
       +----------------+-------+---------------------------------+
       | notBoth        | 8     | "Objective cannot support both  |
       |                |       | negotiation and                 |
       |                |       | synchronization" (register_obj) |
       +----------------+-------+---------------------------------+
       | notDry         | 9     | "Dry-run allowed only with      |
       |                |       | negotiation" (register_obj)     |
       +----------------+-------+---------------------------------+
       | notOverlap     | 10    | "Overlap not supported by this  |
       |                |       | implementation" (register_obj)  |
       +----------------+-------+---------------------------------+
       | objFull        | 11    | "Objective registry full"       |
       |                |       | (register_obj)                  |
       +----------------+-------+---------------------------------+
       | objReg         | 12    | "Objective already registered"  |
       |                |       | (register_obj)                  |
       +----------------+-------+---------------------------------+
       | notYourObj     | 13    | "Objective not registered by    |
       |                |       | this ASA"                       |
       +----------------+-------+---------------------------------+
       | notObj         | 14    | "Objective not found"           |
       +----------------+-------+---------------------------------+
       | notNeg         | 15    | "Objective not negotiable"      |
       |                |       | (req_negotiate,                 |
       |                |       | listen_negotiate)               |
       +----------------+-------+---------------------------------+
       | noSecurity     | 16    | "No security"                   |
       +----------------+-------+---------------------------------+
       | noDiscReply    | 17    | "No reply to discovery"         |
       |                |       | (req_negotiate)                 |
       +----------------+-------+---------------------------------+
       | sockErrNegRq   | 18    | "Socket error sending           |
       |                |       | negotiation request"            |
       |                |       | (req_negotiate)                 |
       +----------------+-------+---------------------------------+
       | noSession      | 19    | "No session"                    |
       +----------------+-------+---------------------------------+
       | noSocket       | 20    | "No socket"                     |
       +----------------+-------+---------------------------------+
       | loopExhausted  | 21    | "Loop count exhausted"          |
       |                |       | (negotiate_step)                |
       +----------------+-------+---------------------------------+
       | sockErrNegStep | 22    | "Socket error sending           |
       |                |       | negotiation step"               |
       |                |       | (negotiate_step)                |
       +----------------+-------+---------------------------------+
       | noPeer         | 23    | "No negotiation peer"           |
       |                |       | (req_negotiate, negotiate_step) |
       +----------------+-------+---------------------------------+
       | CBORfail       | 24    | "CBOR decode failure"           |
       |                |       | (req_negotiate, negotiate_step, |
       |                |       | synchronize)                    |
       +----------------+-------+---------------------------------+
       | invalidNeg     | 25    | "Invalid Negotiate message"     |
       |                |       | (req_negotiate, negotiate_step) |
       +----------------+-------+---------------------------------+
       | invalidEnd     | 26    | "Invalid end message"           |
       |                |       | (req_negotiate, negotiate_step) |
       +----------------+-------+---------------------------------+
       | noNegReply     | 27    | "No reply to negotiation step"  |
       |                |       | (req_negotiate, negotiate_step) |
       +----------------+-------+---------------------------------+
       | noValidStep    | 28    | "No valid reply to negotiation  |
       |                |       | step" (req_negotiate,           |
       |                |       | negotiate_step)                 |
       +----------------+-------+---------------------------------+
       | sockErrWait    | 29    | "Socket error sending wait      |
       |                |       | message" (negotiate_wait)       |
       +----------------+-------+---------------------------------+
       | sockErrEnd     | 30    | "Socket error sending end       |
       |                |       | message" (end_negotiate,        |
       |                |       | send_invalid)                   |
       +----------------+-------+---------------------------------+
       | IDclash        | 31    | "Incoming request Session ID    |
       |                |       | clash" (listen_negotiate)       |
       +----------------+-------+---------------------------------+
       | notSynch       | 32    | "Not a synchronization          |
       |                |       | objective" (synchronize,        |
       |                |       | get_flood)                      |
       +----------------+-------+---------------------------------+
       | notFloodDisc   | 33    | "Not flooded and no reply to    |
       |                |       | discovery" (synchronize)        |
       +----------------+-------+---------------------------------+
       | sockErrSynRq   | 34    | "Socket error sending synch     |
       |                |       | request" (synchronize)          |
       +----------------+-------+---------------------------------+
       | noListener     | 35    | "No synch listener"             |
       |                |       | (synchronize)                   |
       +----------------+-------+---------------------------------+
       | noSynchReply   | 36    | "No reply to synchronization    |
       |                |       | request" (synchronize)          |
       +----------------+-------+---------------------------------+
       | noValidSynch   | 37    | "No valid reply to              |
       |                |       | synchronization request"        |
       |                |       | (synchronize)                   |
       +----------------+-------+---------------------------------+
       | invalidLoc     | 38    | "Invalid locator" (flood)       |
       +----------------+-------+---------------------------------+

                           Table 1: Error Codes

Acknowledgements

   Excellent suggestions were made by Ignas Bagdonas, Carsten Bormann,
   Laurent Ciavaglia, Roman Danyliw, Toerless Eckert, Benjamin Kaduk,
   Erik Kline, Murray Kucherawy, Paul Kyzivat, Guangpeng Li, Michael
   Richardson, Joseph Salowey, Éric Vyncke, Magnus Westerlund, Rob
   Wilton, and other participants in the ANIMA WG and the IESG.

Authors' Addresses

   Brian E. Carpenter
   School of Computer Science
   University of Auckland
   PB 92019
   Auckland 1142
   New Zealand

   Email: brian.e.carpenter@gmail.com


   Bing Liu (editor)
   Huawei Technologies
   Q14, Huawei Campus
   No.156 Beiqing Road
   Hai-Dian District, Beijing
   100095
   China

   Email: leo.liubing@huawei.com


   Wendong Wang
   BUPT University
   Beijing University of Posts & Telecom.
   No.10 Xitucheng Road
   Hai-Dian District, Beijing 100876
   China

   Email: wdwang@bupt.edu.cn


   Xiangyang Gong
   BUPT University
   Beijing University of Posts & Telecom.
   No.10 Xitucheng Road
   Hai-Dian District, Beijing 100876
   China

   Email: xygong@bupt.edu.cn

ISO Plain Language

On this page
PROBLEMAutonomic service agents have to find peers, agree on values, and exchange state, and doing that directly in terms of protocol messages means every agent implements the protocol details itself.
MECHANISMThe RFC describes an abstract application programming interface through which an agent registers itself and its objectives, discovers peers, negotiates, synchronizes and floods, with the GRASP core handling the protocol underneath.
WHAT IT ISThe document is a conceptual outline rather than a formal specification, so language bindings and the exact details are left to implementations, and it leaves authorization of agents to particular objectives for future study.

§1What the API Is For

An Autonomic Service Agent (ASA) is the atomic entity of an autonomic function, instantiated on autonomic nodes, and those nodes are members of a secure Autonomic Control Plane (ACP). When agents communicate with each other they should use the GeneRic Autonomic Signaling Protocol (GRASP). GRASP relies on the message confidentiality and integrity provided by the ACP, so all nodes in a given Autonomic Network share the same trust boundary. Nodes that have not successfully joined the ACP cannot send, receive or intercept GRASP messages and cannot usurp ACP addresses. An ASA runs in an ACP node and therefore benefits from the node's security properties, and all agents within a given Autonomic Network therefore trust each other's messages. For these reasons, the API defined in the document has no explicit security features.

An important feature of GRASP is the concept of a GRASP objective, a data structure encoded like all GRASP messages in CBOR whose main contents are a name and a value. When an objective is passed from one agent to another, its value is either conveyed in one direction, by synchronization or flooding, or negotiated bilaterally. The semantics of the value are opaque to GRASP and therefore to the API, and each objective must be accurately specified in a dedicated specification that defines the syntax and semantics of the value, whether and how it supports negotiation, whether it supports a dry-run mode, and any other details needed for interoperability. The use of CBOR, with CDDL as the data definition language, allows the value to be passed between agents regardless of the programming languages in use. Data storage and consistency during negotiation are the responsibility of the agents involved.

A GRASP implementation could contain several sub-layers. The bottom layer is the GRASP base protocol module, responsible only for sending and receiving GRASP messages and maintaining shared data structures. Above that is the basic API this document describes, and the upper layer contains extended API functions built on the GRASP basic protocol. Multiple agents in a single node share the same instance of GRASP, much as multiple applications share a single TCP/IP stack, and this is hidden from individual agents by the API.

It is desirable that agents be portable user-space programs using a system-independent API, so in many implementations the GRASP code is split between user space and kernel space: in user space, library functions provide the API and communicate directly with the agents, and in kernel space a daemon, or a set of sub-services, provides the GRASP core functions that are independent of specific agents, such as multicast handling and relaying, and common data structures such as the discovery cache. The library and the core would communicate through an interprocess communication or system call mechanism, the details of which are system-dependent.

The functions provided by the API do not map one-to-one onto GRASP messages. They are intended to offer convenient support for message sequences, such as a discovery request followed by responses from several peers or a negotiation request followed by various possible responses, so that agent programmers can write code based on their application requirements rather than needing to understand protocol details. The document gives a conceptual outline of the API and is not a formal specification for any particular programming language or operating system.

An agent is a small program that does one job, and a whole set of them shares a network that is supposed to look after itself. GRASP is the protocol they use to find each other and agree on things. This RFC is the interface between an agent and that protocol, so an agent programmer writes calls like discover and negotiate rather than building and parsing messages.

§2.2Asynchronous Operations and Overlapping Sessions

GRASP depends on asynchronous operations and wait states, and some of its messages are not idempotent, meaning that repeating a message may cause repeated changes of state in the recipient agent. Many agents will need to support several concurrent operations: an agent might need to negotiate one objective with a peer while discovering and synchronizing a different objective with a different peer, or an agent acting as a resource manager might need to run simultaneous negotiations for a given objective with several peers. Such an agent will probably need to support uninterruptible atomic changes to its internal data structures, using a mechanism provided by the operating system and programming language in use.

Three techniques are considered. In multithreading, the operating system and language provide the support for asynchronous operations, including thread creation, context switching, queues, locks and implicit wait states, so API calls can be treated as simple synchronous function calls within their own thread, even where the function includes wait states, blocking and queueing. Concurrent operations each run in their own thread: the discover() call may not return until discovery results have arrived or a timeout has occurred, so if the agent has other work to do, the call must be in a thread of its own. In an event loop with polling, blocking calls are not acceptable, so all calls must be non-blocking, and the main loop supports multiple GRASP sessions in parallel by repeatedly polling each for a change of state. The API implementation provides non-blocking versions of the functions that would otherwise block, and each call returns a noReply code instead of blocking until the event it is waiting for, or a failure, has occurred. The discover() call would return noReply instead of waiting, and would be repeated in every cycle of the main loop until it completes, effectively becoming a polling call. This does not repeat a message, so the fact that some GRASP messages are not idempotent, in particular each step in a negotiation session, is not a problem. In an event loop with callbacks, the agent programmer provides a callback function for each asynchronous operation, called asynchronously when a reply is received or a failure such as a timeout occurs.

GRASP allows an agent acting as a resource allocator for an objective to run two simultaneous negotiations with the same peer for different quantities of the resource, and the allocator must conduct two separate negotiation sessions at the same time with the same peer without mixing them up. Agents could be designed to avoid this, restricting themselves to one negotiation session at a time for a given objective, but that would be a voluntary restriction not required by the protocol, and GRASP assumes that any agent managing a resource may need multiple parallel negotiations, possibly with the same peer.

In the callback model, each agent therefore has to distinguish the sessions, and the peer's IP address is not sufficient. It is also not safe to rely on transport port numbers, since future variants of GRASP might use shared ports rather than a separate port per session. GRASP therefore includes a Session ID, and the API uses a session handle to distinguish simultaneous sessions, so that any number of sessions may proceed asynchronously in parallel. A GRASP session consists of a finite sequence of messages, for discovery, synchronization or negotiation, between two agents, and is uniquely identified on the wire by a pseudorandom Session ID plus the IP address of the initiator. On the first call in a new session the API returns a session_handle that identifies the session within the API, and a likely implementation forms the handle from the underlying GRASP Session ID and IP address. The handle must be used in all subsequent calls for the same session, and the GRASP core must keep state for each operation in progress, most likely keyed by the GRASP Session ID and the source address of the session initiator.

Sessions may terminate for numerous reasons: discovery succeeds or times out, negotiation succeeds or fails, a synchronization result is delivered, the other end fails to respond before a timeout, a loop count expires, or a network socket error occurs. A timeout at one end may result in a timeout or a socket error at the other end, since GRASP does not send error messages in this case. In all cases the API returns an appropriate code to the caller, which should then release any reserved resources, and after failure cases the GRASP specification recommends an exponential backoff before retrying.

Nothing in this protocol is instant, so the interface has to cope with waiting, and the RFC lays out three ways: one thread per operation, a polling loop that checks each session for progress, or callbacks that fire when something happens. Because one agent may run two negotiations with the same peer at once, each session gets its own handle, since the peer's address and port are not enough to tell them apart.

THREE WAYS TO HANDLE ASYNCHRONOUS OPERATIONS
  • MultithreadingEach operation gets its own thread, and the API calls can block inside it. The agent must put a long call such as discover() in a thread of its own.
  • Event loop with pollingAll calls are non-blocking and return a noReply code, and the main loop repeats them each cycle until they complete. No message is ever sent twice this way.
  • Event loop with callbacksThe agent supplies a callback for each operation, and GRASP calls it asynchronously when a reply arrives or a failure such as a timeout occurs.

§2.3.2The Shared Data Structures

Integers in the API are assumed to be 32-bit unsigned integers unless otherwise indicated. All functions return an unsigned errorcode as their first return value, where zero indicates success and any other value indicates failure of some kind. Three errorcodes have special importance: 1, Declined, meaning the other end has sent a Negotiation End message with a Decline option; 2, No reply, used in non-blocking calls to indicate that the other end has sent no reply so far; and 3, Unspecified error, used when no more specific error code applies. Appendix A gives a full list of currently suggested error codes, and while there is no absolute requirement for all implementations to use the same codes, this is highly recommended for portability of applications.

Whenever a timeout parameter appears it is an unsigned integer in milliseconds. If it is zero, the GRASP default timeout applies, with the exception of the discover() function, which has a different interpretation of a zero timeout. If no response is received before the timeout expires, the call fails unless otherwise noted.

An objective has a name as a UTF-8 string, three Boolean flags, a loop count and a value. The neg flag is true if the objective supports negotiation, synch if it supports synchronization, and dry if it supports dry-run negotiation, each defaulting to false. Only one of synch or neg may be true, and dry must not be true unless neg is also true. In some programming languages the preferred implementation may be to represent the flags as bits in a single byte, which is how they are encoded in GRASP messages; in others an enumeration might be preferable. The loop_count is an unsigned 8-bit integer limiting negotiation steps and defaulting to the GRASP default; the initiator of a negotiation sets it to a suitable value to prevent indefinite loops, and it is also used to limit the propagation of discovery and flood messages. The value is a data structure whose format is language dependent, with the constraint that it can be validly represented in CBOR. An important advantage of CBOR is that the value can be completely opaque to the GRASP core yet pass transparently through it, so the core can validate the syntax of a message but cannot validate the value of an objective, and it is the responsibility of each agent to validate the value it receives. Regardless of what other options exist for a language-specific representation, there must always be an option to use a raw CBOR data item as the value; the API wraps it with CBOR Tag 24 for transmission and unwraps it after reception, so agents can communicate regardless of programming language. The name and value fields are of variable length, and GRASP sets no maximum length for them, only for the total length of a GRASP message, though implementations might impose limits.

An asa_locator holds the actual locator, either an IP address or an ASCII string; the interface identifier index through which it was discovered; and the time on the local system clock when it will expire from the cache. The locator may be an IP address, a fully qualified domain name or a URI, these options being mutually exclusive, and it also records whether the locator was discovered via a Divert option, the applicable transport protocol and the applicable port number.

An asa_handle is provided as a very simple hook for an authentication and authorization scheme, which has not been defined. When an agent starts up it registers itself with the GRASP core, which returns an opaque handle that, although not cryptographically protected, would be difficult for a third party to predict, and the agent must present this handle in future calls. This prevents some elementary errors or trivial attacks, such as an agent manipulating an objective it has not registered to use. It is an up to 32-bit opaque value, only required to be unique within the local GRASP instance, valid until the agent terminates, and it should be unpredictable. A session_handle is likewise opaque to the agent and identifies calls to the API as belonging to a specific session; as well as distinguishing calls from different sessions, it allows GRASP to detect and ignore calls from non-existent or timed-out sessions.

Most calls take the same handful of structures. An objective is a name, three flags saying what it can be used for, a step limit and an opaque value the protocol carries without understanding. A locator says where a peer is and when the entry goes stale. Two handles complete the picture: one proves the agent registered, and one names the session a call belongs to.

THE ERROR CODES WITH SPECIAL MEANING
  • 0Success — The call succeeded.
  • 1Declined — The other end sent a Negotiation End message with a Decline option.
  • 2No reply — In a non-blocking call, the other end has sent no reply so far.
  • 3Unspecified error — Used when no more specific error code applies.

§2.3.3The Function Groups and the Negotiation Sequence

The functions fall into several groups. Registration lets an agent register itself with the GRASP core and register the objectives it will manipulate. Discovery lets an agent that needs to initiate negotiation or synchronization of an objective find a peer willing to respond. Negotiation lets an agent act as initiator or responder for a negotiation session; after initiation negotiation is symmetric, so most of the functions can be used by either party. Synchronization lets an agent act as initiator or as responder and data source for a synchronization session. Flooding lets an agent send and receive an objective flooded to all nodes of the ACP. There is also a function for reporting an invalid message.

In the registration group, all agents must call register_asa() before any other API call, and it returns the asa_handle the agent must present in all subsequent calls. deregister_asa() removes all state for the agent and deregisters any objectives it registered, and the same must happen automatically if an agent exits. register_objective() must be used for any objective whose value the agent needs to transmit by negotiation, synchronization or flooding, but not for an objective it only wants to receive, and it takes a ttl giving the valid lifetime in milliseconds of any discovery response generated for the objective. A discoverable flag makes the objective immediately discoverable, for objectives defined only for discovery; an overlap flag allows more than one agent to register the same objective in the same GRASP instance, which is of value for the life cycle management of agents and must be used consistently for a given objective; and a local flag requires discovery to return a link-local address, for objectives that must be restricted to the local link.

The discover() function returns a list of discovered locators for a given objective, where an empty list means none were found within the timeout, which is not an error: GRASP discovery is not a deterministic process, and if several nodes handle an objective, none, some or all of them may be discovered before the timeout expires. The minimum_TTL parameter must be at least zero, and locally cached locators whose remaining time to live is less than or equal to it are deleted first, so a value of zero flushes all entries without affecting sessions already in progress. A timeout of zero returns the remaining cached locators immediately and takes no other action, so a call with both parameters zero is pointless.

Because the negotiation mechanism differs from a typical client or server exchange, the RFC illustrates the sequence of calls and GRASP messages. After the first protocol exchange the process is symmetrical, with the negotiating steps strictly alternating between the two sides, either side can end the negotiation, and the side due to respond next can insert a delay at any time to extend the other side's timeout, for example if an agent needed to negotiate with a third party first. The loop count embedded in the objective is initialized by the agent that starts a negotiation and is decremented by the GRASP core at each step before sending each M_NEGOTIATE message; if it reaches zero the negotiation fails and each side receives an error code. As the negotiation proceeds, each side updates the value of the objective according to its own semantics, though many objectives will have orderable values so that negotiation can be a simple bidding process, but that is not a requirement. Failure to agree, a timeout, or loop count exhaustion may all end a session, and none of these is a protocol failure.

The request_negotiate() function opens a negotiation session between two agents, and GRASP currently does not support multiparty negotiation. The peer parameter is the target node, and if it is null GRASP discovery is performed first to find a suitable peer. If the return code is zero there are two cases: the session handle may be null, in which case the negotiation succeeded in one exchange and the peer accepted the request, so the returned proffered objective holds the accepted value and no handle is needed; or the handle is not null, in which case the negotiation must continue, the returned proffered objective is the peer's first counter-offer and its contents must be used to prepare the next step, because it carries the updated loop count. The function must be followed by calls to negotiate_step, negotiate_wait or end_negotiate until the negotiation ends. A dry-run mode may or may not apply to a given session, but the two modes must not be mixed in one session.

The synchronize() function requests the synchronized value of an objective, and if the peer parameter is null and the objective is already in the local cache the flooded objective is returned immediately, ignoring the timeout. listen_synchronize() and stop_listen_synchronize() start and stop acting as a synchronization responder, and the former is non-blocking and may be repeated whenever the value changes. The flood() function floods one or more objectives and their values to all GRASP nodes, with a ttl giving the valid lifetime in milliseconds of the flooded data, where zero means infinity, and the locator tagging each objective is normally null but may be a valid locator for infrastructure agents that need to flood an address, protocol and port triplet. Each GRASP node caches the flooded objectives it receives until their time to live expires, and cached objectives are tagged with their origin as well as an expiry time, so multiple copies of the same objective may be cached at once. get_flood() returns the current value of a flooded objective, with the tagging locator letting the recipient tell copies from different sources apart, and expire_flood() lets an agent expire specific entries in the local flood cache when it decides a flooded value is stale.

The calls come in five groups: register yourself and your objectives, find a peer, negotiate with one, synchronise a value with one, and publish a value to everyone. Negotiation is the interesting one, because it is not a request and a reply: the two sides alternate proposals, either can walk away, and a step limit in the objective stops the exchange going round in circles.

THE NEGOTIATION SEQUENCE FROM RFC 8991
InitiatorResponder
  1. M_REQ_NEGrequest_negotiate() opens the session.
  2. M_NEGOTIATEnegotiate_step() answers with a counter-offer. The steps alternate strictly between the two sides.
  3. M_NEGOTIATEThe counter-offer carries the updated loop count.
  4. M_WAITnegotiate_wait() inserts a delay, extending the other side's timeout.
  5. M_ENDend_negotiate() ends the negotiation.

§3Security Considerations

Security considerations for the GRASP protocol are discussed in the protocol specification itself. These include denial-of-service issues, even though these are considered a low risk in the ACP. In various places GRASP recommends an exponential backoff, and an agent using the API should use exponential backoff after failed discover, request_negotiate or synchronize operations, with the timescale depending on the semantics of the objective concerned. A flood operation should not be repeated at shorter intervals than is useful. These precautions are intended to assist the detection of denial-of-service attacks.

As a general precaution, all agents able to handle multiple negotiation or synchronization requests in parallel may protect themselves by limiting the number of requests they handle simultaneously and silently discarding excess requests. It might also be useful for the GRASP core to limit the number of objectives registered by a given agent, the total number of agents registered, and the total number of simultaneous sessions, to protect system resources. During times of high autonomic activity, such as recovery from widespread faults, agents may experience many session failures.

The trust model is that all agents in a given Autonomic Network communicate via a secure autonomic control plane and therefore trust each other's messages. Specific authorization of agents to use particular GRASP objectives is a subject for future study. A malicious agent could extend a negotiation session indefinitely by use of the negotiate_wait function or by manipulating the loop count of an objective, and a robustly implemented agent could detect such behaviour by a peer and break off the negotiation.

The asa_handle is the first line of defence against a malware process attempting to imitate a legitimately registered agent, and the session_handle is the first line of defence against one attempting to hijack a GRASP session. Both are likely to be created using GRASP's 32-bit pseudorandom Session ID, and by construction GRASP avoids the risk of Session ID collisions. There remains a finite probability that an attacker could guess a Session ID, session handle or asa_handle, but this would only be of value to an attacker that had already penetrated the ACP, which would allow many other simpler forms of attack.

Because the network is assumed to be secure already, the API has no security features of its own, and the RFC is mostly about not being a nuisance. Failed calls should be retried with a growing delay, floods should not be repeated too often, and an agent that can handle many requests at once should set a limit and drop the excess, so that a misbehaving peer cannot take the node down.

§6References