idnits 2.17.1 draft-ietf-anima-grasp-10.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (March 10, 2017) is 2603 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) -- Looks like a reference, but probably isn't: '7' on line 3431 -- Looks like a reference, but probably isn't: '13767778' on line 3431 -- Looks like a reference, but probably isn't: '34965' on line 3431 == Outdated reference: A later version (-11) exists of draft-greevenbosch-appsawg-cbor-cddl-09 ** Downref: Normative reference to an Informational draft: draft-greevenbosch-appsawg-cbor-cddl (ref. 'I-D.greevenbosch-appsawg-cbor-cddl') ** Obsolete normative reference: RFC 5246 (Obsoleted by RFC 8446) ** Obsolete normative reference: RFC 6347 (Obsoleted by RFC 9147) ** Obsolete normative reference: RFC 7049 (Obsoleted by RFC 8949) == Outdated reference: A later version (-30) exists of draft-ietf-anima-autonomic-control-plane-05 == Outdated reference: A later version (-45) exists of draft-ietf-anima-bootstrapping-keyinfra-04 == Outdated reference: A later version (-10) exists of draft-ietf-anima-reference-model-02 == Outdated reference: A later version (-10) exists of draft-ietf-anima-stable-connectivity-02 == Outdated reference: A later version (-06) exists of draft-liu-anima-grasp-api-03 -- Obsolete informational reference (is this intentional?): RFC 3315 (Obsoleted by RFC 8415) -- Obsolete informational reference (is this intentional?): RFC 5226 (Obsoleted by RFC 8126) Summary: 4 errors (**), 0 flaws (~~), 7 warnings (==), 6 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group C. Bormann 3 Internet-Draft Universitaet Bremen TZI 4 Intended status: Standards Track B. Carpenter, Ed. 5 Expires: September 11, 2017 Univ. of Auckland 6 B. Liu, Ed. 7 Huawei Technologies Co., Ltd 8 March 10, 2017 10 A Generic Autonomic Signaling Protocol (GRASP) 11 draft-ietf-anima-grasp-10 13 Abstract 15 This document establishes requirements for a signaling protocol that 16 enables autonomic nodes and autonomic service agents to dynamically 17 discover peers, to synchronize state with them, and to negotiate 18 parameter settings with them. The document then defines a general 19 protocol for discovery, synchronization and negotiation, while the 20 technical objectives for specific scenarios are to be described in 21 separate documents. An Appendix briefly discusses existing protocols 22 with comparable features. 24 Status of This Memo 26 This Internet-Draft is submitted in full conformance with the 27 provisions of BCP 78 and BCP 79. 29 Internet-Drafts are working documents of the Internet Engineering 30 Task Force (IETF). Note that other groups may also distribute 31 working documents as Internet-Drafts. The list of current Internet- 32 Drafts is at http://datatracker.ietf.org/drafts/current/. 34 Internet-Drafts are draft documents valid for a maximum of six months 35 and may be updated, replaced, or obsoleted by other documents at any 36 time. It is inappropriate to use Internet-Drafts as reference 37 material or to cite them other than as "work in progress." 39 This Internet-Draft will expire on September 11, 2017. 41 Copyright Notice 43 Copyright (c) 2017 IETF Trust and the persons identified as the 44 document authors. All rights reserved. 46 This document is subject to BCP 78 and the IETF Trust's Legal 47 Provisions Relating to IETF Documents 48 (http://trustee.ietf.org/license-info) in effect on the date of 49 publication of this document. Please review these documents 50 carefully, as they describe your rights and restrictions with respect 51 to this document. Code Components extracted from this document must 52 include Simplified BSD License text as described in Section 4.e of 53 the Trust Legal Provisions and are provided without warranty as 54 described in the Simplified BSD License. 56 Table of Contents 58 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 59 2. Requirement Analysis of Discovery, Synchronization and 60 Negotiation . . . . . . . . . . . . . . . . . . . . . . . . . 5 61 2.1. Requirements for Discovery . . . . . . . . . . . . . . . 5 62 2.2. Requirements for Synchronization and Negotiation 63 Capability . . . . . . . . . . . . . . . . . . . . . . . 6 64 2.3. Specific Technical Requirements . . . . . . . . . . . . . 9 65 3. GRASP Protocol Overview . . . . . . . . . . . . . . . . . . . 10 66 3.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 10 67 3.2. High Level Deployment Model . . . . . . . . . . . . . . . 12 68 3.3. High Level Design Choices . . . . . . . . . . . . . . . . 13 69 3.4. Quick Operating Overview . . . . . . . . . . . . . . . . 16 70 3.5. GRASP Protocol Basic Properties and Mechanisms . . . . . 16 71 3.5.1. Required External Security Mechanism . . . . . . . . 16 72 3.5.2. Constrained Instances . . . . . . . . . . . . . . . . 17 73 3.5.3. Transport Layer Usage . . . . . . . . . . . . . . . . 19 74 3.5.4. Discovery Mechanism and Procedures . . . . . . . . . 20 75 3.5.5. Negotiation Procedures . . . . . . . . . . . . . . . 23 76 3.5.6. Synchronization and Flooding Procedures . . . . . . . 25 77 3.6. GRASP Constants . . . . . . . . . . . . . . . . . . . . . 27 78 3.7. Session Identifier (Session ID) . . . . . . . . . . . . . 28 79 3.8. GRASP Messages . . . . . . . . . . . . . . . . . . . . . 29 80 3.8.1. Message Overview . . . . . . . . . . . . . . . . . . 29 81 3.8.2. GRASP Message Format . . . . . . . . . . . . . . . . 29 82 3.8.3. Message Size . . . . . . . . . . . . . . . . . . . . 30 83 3.8.4. Discovery Message . . . . . . . . . . . . . . . . . . 30 84 3.8.5. Discovery Response Message . . . . . . . . . . . . . 31 85 3.8.6. Request Messages . . . . . . . . . . . . . . . . . . 32 86 3.8.7. Negotiation Message . . . . . . . . . . . . . . . . . 33 87 3.8.8. Negotiation End Message . . . . . . . . . . . . . . . 34 88 3.8.9. Confirm Waiting Message . . . . . . . . . . . . . 34 89 3.8.10. Synchronization Message . . . . . . . . . . . . . . . 34 90 3.8.11. Flood Synchronization Message . . . . . . . . . . . . 35 91 3.8.12. Invalid Message . . . . . . . . . . . . . . . . . . . 36 92 3.8.13. No Operation Message . . . . . . . . . . . . . . . . 36 93 3.9. GRASP Options . . . . . . . . . . . . . . . . . . . . . . 36 94 3.9.1. Format of GRASP Options . . . . . . . . . . . . . . . 36 95 3.9.2. Divert Option . . . . . . . . . . . . . . . . . . . . 37 96 3.9.3. Accept Option . . . . . . . . . . . . . . . . . . . . 37 97 3.9.4. Decline Option . . . . . . . . . . . . . . . . . . . 37 98 3.9.5. Locator Options . . . . . . . . . . . . . . . . . . . 38 99 3.10. Objective Options . . . . . . . . . . . . . . . . . . . . 40 100 3.10.1. Format of Objective Options . . . . . . . . . . . . 40 101 3.10.2. Objective flags . . . . . . . . . . . . . . . . . . 41 102 3.10.3. General Considerations for Objective Options . . . . 41 103 3.10.4. Organizing of Objective Options . . . . . . . . . . 42 104 3.10.5. Experimental and Example Objective Options . . . . . 43 105 4. Implementation Status [RFC Editor: please remove] . . . . . . 44 106 4.1. BUPT C++ Implementation . . . . . . . . . . . . . . . . . 44 107 4.2. Python Implementation . . . . . . . . . . . . . . . . . . 44 108 5. Security Considerations . . . . . . . . . . . . . . . . . . . 45 109 6. CDDL Specification of GRASP . . . . . . . . . . . . . . . . . 47 110 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 50 111 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 52 112 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 52 113 9.1. Normative References . . . . . . . . . . . . . . . . . . 52 114 9.2. Informative References . . . . . . . . . . . . . . . . . 53 115 Appendix A. Open Issues [RFC Editor: This section should be 116 empty. Please remove] . . . . . . . . . . . . . . . 56 117 Appendix B. Closed Issues [RFC Editor: Please remove] . . . . . 57 118 Appendix C. Change log [RFC Editor: Please remove] . . . . . . . 65 119 Appendix D. Example Message Formats . . . . . . . . . . . . . . 71 120 D.1. Discovery Example . . . . . . . . . . . . . . . . . . . . 71 121 D.2. Flood Example . . . . . . . . . . . . . . . . . . . . . . 72 122 D.3. Synchronization Example . . . . . . . . . . . . . . . . . 72 123 D.4. Simple Negotiation Example . . . . . . . . . . . . . . . 72 124 D.5. Complete Negotiation Example . . . . . . . . . . . . . . 73 125 Appendix E. Capability Analysis of Current Protocols . . . . . . 74 126 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 76 128 1. Introduction 130 The success of the Internet has made IP-based networks bigger and 131 more complicated. Large-scale ISP and enterprise networks have 132 become more and more problematic for human based management. Also, 133 operational costs are growing quickly. Consequently, there are 134 increased requirements for autonomic behavior in the networks. 135 General aspects of autonomic networks are discussed in [RFC7575] and 136 [RFC7576]. 138 One approach is to largely decentralize the logic of network 139 management by migrating it into network elements. A reference model 140 for autonomic networking on this basis is given in 141 [I-D.ietf-anima-reference-model]. The reader should consult this 142 document to understand how various autonomic components fit together. 143 In order to fulfill autonomy, devices that embody Autonomic Service 144 Agents (ASAs, [RFC7575]) have specific signaling requirements. In 145 particular they need to discover each other, to synchronize state 146 with each other, and to negotiate parameters and resources directly 147 with each other. There is no limitation on the types of parameters 148 and resources concerned, which can include very basic information 149 needed for addressing and routing, as well as anything else that 150 might be configured in a conventional non-autonomic network. The 151 atomic unit of discovery, synchronization or negotiation is referred 152 to as a technical objective, i.e, a configurable parameter or set of 153 parameters (defined more precisely in Section 3.1). 155 Following this Introduction, Section 2 describes the requirements for 156 discovery, synchronization and negotiation. Negotiation is an 157 iterative process, requiring multiple message exchanges forming a 158 closed loop between the negotiating entities. In fact, these 159 entities are ASAs, normally but not necessarily in different network 160 devices. State synchronization, when needed, can be regarded as a 161 special case of negotiation, without iteration. Section 3.3 162 describes a behavior model for a protocol intended to support 163 discovery, synchronization and negotiation. The design of GeneRic 164 Autonomic Signaling Protocol (GRASP) in Section 3 of this document is 165 based on this behavior model. The relevant capabilities of various 166 existing protocols are reviewed in Appendix E. 168 The proposed discovery mechanism is oriented towards synchronization 169 and negotiation objectives. It is based on a neighbor discovery 170 process on the local link, but also supports diversion to peers on 171 other links. There is no assumption of any particular form of 172 network topology. When a device starts up with no pre-configuration, 173 it has no knowledge of the topology. The protocol itself is capable 174 of being used in a small and/or flat network structure such as a 175 small office or home network as well as in a large professionally 176 managed network. Therefore, the discovery mechanism needs to be able 177 to allow a device to bootstrap itself without making any prior 178 assumptions about network structure. 180 Because GRASP can be used as part of a decision process among 181 distributed devices or between networks, it must run in a secure and 182 strongly authenticated environment. 184 In realistic deployments, not all devices will support GRASP. 185 Therefore, some autonomic service agents will directly manage a group 186 of non-autonomic nodes, and other non-autonomic nodes will be managed 187 traditionally. Such mixed scenarios are not discussed in this 188 specification. 190 2. Requirement Analysis of Discovery, Synchronization and Negotiation 192 This section discusses the requirements for discovery, negotiation 193 and synchronization capabilities. The primary user of the protocol 194 is an autonomic service agent (ASA), so the requirements are mainly 195 expressed as the features needed by an ASA. A single physical device 196 might contain several ASAs, and a single ASA might manage several 197 technical objectives. If a technical objective is managed by several 198 ASAs, any necessary coordination is outside the scope of the GRASP 199 signaling protocol. Furthermore, requirements for ASAs themselves, 200 such as the processing of Intent [RFC7575], are out of scope for the 201 present document. 203 2.1. Requirements for Discovery 205 D1. ASAs may be designed to manage any type of configurable device 206 or software, as required in Section 2.2. A basic requirement is 207 therefore that the protocol can represent and discover any kind of 208 technical objective among arbitrary subsets of participating nodes. 210 In an autonomic network we must assume that when a device starts up 211 it has no information about any peer devices, the network structure, 212 or what specific role it must play. The ASA(s) inside the device are 213 in the same situation. In some cases, when a new application session 214 starts up within a device, the device or ASA may again lack 215 information about relevant peers. For example, it might be necessary 216 to set up resources on multiple other devices, coordinated and 217 matched to each other so that there is no wasted resource. Security 218 settings might also need updating to allow for the new device or 219 user. The relevant peers may be different for different technical 220 objectives. Therefore discovery needs to be repeated as often as 221 necessary to find peers capable of acting as counterparts for each 222 objective that a discovery initiator needs to handle. From this 223 background we derive the next three requirements: 225 D2. When an ASA first starts up, it may have no knowledge of the 226 specific network to which it is attached. Therefore the discovery 227 process must be able to support any network scenario, assuming only 228 that the device concerned is bootstrapped from factory condition. 230 D3. When an ASA starts up, it must require no configured location 231 information about any peers in order to discover them. 233 D4. If an ASA supports multiple technical objectives, relevant peers 234 may be different for different discovery objectives, so discovery 235 needs to be performed separately to find counterparts for each 236 objective. Thus, there must be a mechanism by which an ASA can 237 separately discover peer ASAs for each of the technical objectives 238 that it needs to manage, whenever necessary. 240 D5. Following discovery, an ASA will normally perform negotiation or 241 synchronization for the corresponding objectives. The design should 242 allow for this by conveniently linking discovery to negotiation and 243 synchronization. It may provide an optional mechanism to combine 244 discovery and negotiation/synchronization in a single protocol 245 exchange. 247 D6. Some objectives may only be significant on the local link, but 248 others may be significant across the routed network and require off- 249 link operations. Thus, the relevant peers might be immediate 250 neighbors on the same layer 2 link, or they might be more distant and 251 only accessible via layer 3. The mechanism must therefore provide 252 both on-link and off-link discovery of ASAs supporting specific 253 technical objectives. 255 D7. The discovery process should be flexible enough to allow for 256 special cases, such as the following: 258 o During initialization, a device must be able to establish mutual 259 trust with the rest of the network and participate in an 260 authentication mechanism. Although this will inevitably start 261 with a discovery action, it is a special case precisely because 262 trust is not yet established. This topic is the subject of 263 [I-D.ietf-anima-bootstrapping-keyinfra]. We require that once 264 trust has been established for a device, all ASAs within the 265 device inherit the device's credentials and are also trusted. 266 This does not preclude the device having multiple credentials. 268 o Depending on the type of network involved, discovery of other 269 central functions might be needed, such as the Network Operations 270 Center (NOC) [I-D.ietf-anima-stable-connectivity]. The protocol 271 must be capable of supporting such discovery during 272 initialization, as well as discovery during ongoing operation. 274 D8. The discovery process must not generate excessive traffic and 275 must take account of sleeping nodes. 277 D9. There must be a mechanism for handling stale discovery results. 279 2.2. Requirements for Synchronization and Negotiation Capability 281 As background, consider the example of routing protocols, the closest 282 approximation to autonomic networking already in widespread use. 283 Routing protocols use a largely autonomic model based on distributed 284 devices that communicate repeatedly with each other. The focus is 285 reachability, so routing protocols primarily consider simple link 286 status and metrics, and an underlying assumption is that nodes need a 287 consistent, although partial, view of the network topology in order 288 for the routing algorithm to converge. Also, routing is mainly based 289 on simple information synchronization between peers, rather than on 290 bi-directional negotiation. 292 By contrast, autonomic networks need to be able to manage many 293 different types of parameter and consider many more dimensions, such 294 as latency, load, unused or limited resources, conflicting resource 295 requests, security settings, power saving, load balancing, etc. 296 Status information and resource metrics need to be shared between 297 nodes for dynamic adjustment of resources and for monitoring 298 purposes. While this might be achieved by existing protocols when 299 they are available, the new protocol needs to be able to support 300 parameter exchange, including mutual synchronization, even when no 301 negotiation as such is required. In general, these parameters do not 302 apply to all participating nodes, but only to a subset. 304 SN1. A basic requirement for the protocol is therefore the ability 305 to represent, discover, synchronize and negotiate almost any kind of 306 network parameter among selected subsets of participating nodes. 308 SN2. Negotiation is an iterative request/response process that must 309 be guaranteed to terminate (with success or failure). While tie- 310 breaking rules must be defined specifically for each use case, the 311 protocol should have some general mechanisms in support of loop and 312 deadlock prevention, such as hop count limits or timeouts. 314 SN3. Synchronization must be possible for groups of nodes ranging 315 from small to very large. 317 SN4. To avoid "reinventing the wheel", the protocol should be able 318 to encapsulate the data formats used by existing configuration 319 protocols (such as NETCONF/YANG) in cases where that is convenient. 321 SN5. Human intervention in complex situations is costly and error- 322 prone. Therefore, synchronization or negotiation of parameters 323 without human intervention is desirable whenever the coordination of 324 multiple devices can improve overall network performance. It follows 325 that the protocol's resource requirements must be appropriate for any 326 device that would otherwise need human intervention. The issue of 327 running in constrained nodes is discussed in 328 [I-D.ietf-anima-reference-model]. 330 SN6. Human intervention in large networks is often replaced by use 331 of a top-down network management system (NMS). It therefore follows 332 that the protocol, as part of the Autonomic Networking 333 Infrastructure, should be capable of running in any device that would 334 otherwise be managed by an NMS, and that it can co-exist with an NMS, 335 and with protocols such as SNMP and NETCONF. 337 SN7. Some features are expected to be implemented by individual 338 ASAs, but the protocol must be general enough to allow them: 340 o Dependencies and conflicts: In order to decide upon a 341 configuration for a given device, the device may need information 342 from neighbors. This can be established through the negotiation 343 procedure, or through synchronization if that is sufficient. 344 However, a given item in a neighbor may depend on other 345 information from its own neighbors, which may need another 346 negotiation or synchronization procedure to obtain or decide. 347 Therefore, there are potential dependencies and conflicts among 348 negotiation or synchronization procedures. Resolving dependencies 349 and conflicts is a matter for the individual ASAs involved. To 350 allow this, there need to be clear boundaries and convergence 351 mechanisms for negotiations. Also some mechanisms are needed to 352 avoid loop dependencies or uncontrolled growth in a tree of 353 dependencies. It is the ASA designer's responsibility to avoid or 354 detect looping dependencies or excessive growth of dependency 355 trees. The protocol's role is limited to bilateral signaling 356 between ASAs, and the avoidance of loops during bilateral 357 signaling. 359 o Recovery from faults and identification of faulty devices should 360 be as automatic as possible. The protocol's role is limited to 361 discovery, synchronization and negotiation. These processes can 362 occur at any time, and an ASA may need to repeat any of these 363 steps when the ASA detects an event such as a negotiation 364 counterpart failing. 366 o Since a major goal is to minimize human intervention, it is 367 necessary that the network can in effect "think ahead" before 368 changing its parameters. One aspect of this is an ASA that relies 369 on a knowledge base to predict network behavior. This is out of 370 scope for the signaling protocol. However, another aspect is 371 forecasting the effect of a change by a "dry run" negotiation 372 before actually installing the change. Signaling a dry run is 373 therefore a desirable feature of the protocol. 375 Note that management logging, monitoring, alerts and tools for 376 intervention are required. However, these can only be features of 377 individual ASAs, not of the protocol itself. Another document 378 [I-D.ietf-anima-stable-connectivity] discusses how such agents may be 379 linked into conventional OAM systems via an Autonomic Control Plane 380 [I-D.ietf-anima-autonomic-control-plane]. 382 SN8. The protocol will be able to deal with a wide variety of 383 technical objectives, covering any type of network parameter. 384 Therefore the protocol will need a flexible and easily extensible 385 format for describing objectives. At a later stage it may be 386 desirable to adopt an explicit information model. One consideration 387 is whether to adopt an existing information model or to design a new 388 one. 390 2.3. Specific Technical Requirements 392 T1. It should be convenient for ASA designers to define new 393 technical objectives and for programmers to express them, without 394 excessive impact on run-time efficiency and footprint. In 395 particular, it should be convenient for ASAs to be implemented 396 independently of each other as user space programs rather than as 397 kernel code, where such a programming model is possible. The classes 398 of device in which the protocol might run is discussed in 399 [I-D.ietf-anima-reference-model]. 401 T2. The protocol should be easily extensible in case the initially 402 defined discovery, synchronization and negotiation mechanisms prove 403 to be insufficient. 405 T3. To be a generic platform, the protocol payload format should be 406 independent of the transport protocol or IP version. In particular, 407 it should be able to run over IPv6 or IPv4. However, some functions, 408 such as multicasting on a link, might need to be IP version 409 dependent. By default, IPv6 should be preferred. 411 T4. The protocol must be able to access off-link counterparts via 412 routable addresses, i.e., must not be restricted to link-local 413 operation. 415 T5. It must also be possible for an external discovery mechanism to 416 be used, if appropriate for a given technical objective. In other 417 words, GRASP discovery must not be a prerequisite for GRASP 418 negotiation or synchronization. 420 T6. The protocol must be capable of supporting multiple simultaneous 421 operations with one or more peers, especially when wait states occur. 423 T7. Intent: Although the distribution of Intent is out of scope for 424 this document, the protocol must not by design exclude its use for 425 Intent distribution. 427 T8. Management monitoring, alerts and intervention: Devices should 428 be able to report to a monitoring system. Some events must be able 429 to generate operator alerts and some provision for emergency 430 intervention must be possible (e.g. to freeze synchronization or 431 negotiation in a mis-behaving device). These features might not use 432 the signaling protocol itself, but its design should not exclude such 433 use. 435 T9. Because this protocol may directly cause changes to device 436 configurations and have significant impacts on a running network, all 437 protocol exchanges need to be fully secured against forged messages 438 and man-in-the middle attacks, and secured as much as reasonably 439 possible against denial of service attacks. There must also be an 440 encryption mechanism to resist unwanted monitoring. However, it is 441 not required that the protocol itself provides these security 442 features; it may depend on an existing secure environment. 444 3. GRASP Protocol Overview 446 3.1. Terminology 448 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 449 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 450 "OPTIONAL" in this document are to be interpreted as described in 451 [RFC2119] when they appear in ALL CAPS. When these words are not in 452 ALL CAPS (such as "should" or "Should"), they have their usual 453 English meanings, and are not to be interpreted as [RFC2119] key 454 words. 456 This document uses terminology defined in [RFC7575]. 458 The following additional terms are used throughout this document: 460 o Discovery: a process by which an ASA discovers peers according to 461 a specific discovery objective. The discovery results may be 462 different according to the different discovery objectives. The 463 discovered peers may later be used as negotiation counterparts or 464 as sources of synchronization data. 466 o Negotiation: a process by which two ASAs interact iteratively to 467 agree on parameter settings that best satisfy the objectives of 468 both ASAs. 470 o State Synchronization: a process by which ASAs interact to receive 471 the current state of parameter values stored in other ASAs. This 472 is a special case of negotiation in which information is sent but 473 the ASAs do not request their peers to change parameter settings. 474 All other definitions apply to both negotiation and 475 synchronization. 477 o Technical Objective (usually abbreviated as Objective): A 478 technical objective is a data structure, whose main contents are a 479 name and a value. The value consists of a single configurable 480 parameter or a set of parameters of some kind. The exact format 481 of an objective is defined in Section 3.10.1. An objective occurs 482 in three contexts: Discovery, Negotiation and Synchronization. 483 Normally, a given objective will not occur in negotiation and 484 synchronization contexts simultaneously. 486 * One ASA may support multiple independent objectives. 488 * The parameter(s) in the value of a given objective apply to a 489 specific service or function or action. They may in principle 490 be anything that can be set to a specific logical, numerical or 491 string value, or a more complex data structure, by a network 492 node. Each node is expected to contain one or more ASAs which 493 may each manage subsidiary non-autonomic nodes. 495 * Discovery Objective: an objective in the process of discovery. 496 Its value may be undefined. 498 * Synchronization Objective: an objective whose specific 499 technical content needs to be synchronized among two or more 500 ASAs. Thus, each ASA will maintain its own copy of the 501 objective. 503 * Negotiation Objective: an objective whose specific technical 504 content needs to be decided in coordination with another ASA. 505 Again, each ASA will maintain its own copy of the objective. 507 A detailed discussion of objectives, including their format, is 508 found in Section 3.10. 510 o Discovery Initiator: an ASA that starts discovery by sending a 511 discovery message referring to a specific discovery objective. 513 o Discovery Responder: a peer that either contains an ASA supporting 514 the discovery objective indicated by the discovery initiator, or 515 caches the locator(s) of the ASA(s) supporting the objective. It 516 sends a Discovery Response, as described later. 518 o Synchronization Initiator: an ASA that starts synchronization by 519 sending a request message referring to a specific synchronization 520 objective. 522 o Synchronization Responder: a peer ASA which responds with the 523 value of a synchronization objective. 525 o Negotiation Initiator: an ASA that starts negotiation by sending a 526 request message referring to a specific negotiation objective. 528 o Negotiation Counterpart: a peer with which the Negotiation 529 Initiator negotiates a specific negotiation objective. 531 o GRASP Instance: This refers to an instantiation of a GRASP 532 protocol engine, likely including multiple threads or processes as 533 well as dynamic data structures such as a discovery cache, running 534 in a given security environment on a single device. 536 o Network Interface: Unless otherwise stated, this refers to a 537 network interface - which might be physical or virtual - that a 538 specific instance of GRASP is currently using. A device might 539 have other interfaces that are not used by GRASP. 541 3.2. High Level Deployment Model 543 A GRASP implementation will be part of the Autonomic Networking 544 Infrastructure in an autonomic node, which must also provide an 545 appropriate security environment. In accordance with 546 [I-D.ietf-anima-reference-model], this SHOULD be the Autonomic 547 Control Plane (ACP) [I-D.ietf-anima-autonomic-control-plane]. It is 548 expected that GRASP will access the ACP by using a typical socket 549 programming interface. There will also be one or more Autonomic 550 Service Agents (ASAs). In the minimal case of a single-purpose 551 device, these components might be fully integrated. A more common 552 model is expected to be a multi-purpose device capable of containing 553 several ASAs. In this case it is expected that the ACP, GRASP and 554 the ASAs will be implemented as separate processes, which are 555 probably multi-threaded to support asynchronous and simultaneous 556 operations. 558 In some scenarios, a limited negotiation model might be deployed 559 based on a limited trust relationship such as that between two 560 administrative domains. ASAs might then exchange limited information 561 and negotiate some particular configurations. 563 A suitable Application Programming Interface (API) will be needed 564 between GRASP and the ASAs. In some implementations, ASAs would run 565 in user space with a GRASP library providing the API, and this 566 library would in turn communicate via system calls with core GRASP 567 functions. Details of the API are out of scope for the present 568 document. For further details of possible deployment models, see 569 [I-D.ietf-anima-reference-model]. 571 An instance of GRASP must be aware of the network interfaces it will 572 use, and of the appropriate global-scope and link-local addresses. 574 In the presence of the ACP, such information will be available from 575 the adjacency table discussed in [I-D.ietf-anima-reference-model]. 576 In other cases, GRASP must determine such information for itself. 577 Details depend on the device and operating system. In the rest of 578 this document, the term 'interfaces' refers only to the set of 579 network interfaces that a specific instance of GRASP is currently 580 using. 582 Because GRASP needs to work with very high reliability, especially 583 during bootstrapping and during fault conditions, it is essential 584 that every implementation is as robust as possible. For example, 585 discovery failures, or any kind of socket exception at any time, must 586 not cause irrecoverable failures in GRASP itself, and must return 587 suitable error codes through the API so that ASAs can also recover. 589 GRASP must not depend upon non-volatile data storage. All run time 590 error conditions, and events such as address renumbering, network 591 interface failures, and CPU sleep/wake cycles, must be handled in 592 such a way that GRASP will still operate correctly and securely 593 (Section 3.5.1) afterwards. 595 An autonomic node will normally run a single instance of GRASP, used 596 by multiple ASAs. Possible exceptions are mentioned below. 598 3.3. High Level Design Choices 600 This section describes a behavior model and design choices for GRASP, 601 supporting discovery, synchronization and negotiation, to act as a 602 platform for different technical objectives. 604 o A generic platform: 606 The protocol design is generic and independent of the 607 synchronization or negotiation contents. The technical contents 608 will vary according to the various technical objectives and the 609 different pairs of counterparts. 611 o Normally, a single main instance of the GRASP protocol engine will 612 exist in an autonomic node, and each ASA will run as an 613 independent asynchronous process. However, scenarios where 614 multiple instances of GRASP run in a single node, perhaps with 615 different security properties, are possible (Section 3.5.2). In 616 this case, each instance MUST listen independently for GRASP link- 617 local multicasts, and all instances MUST be woken by each such 618 multicast, in order for discovery and flooding to work correctly. 620 o Security infrastructure: 622 As noted above, the protocol itself has no built-in security 623 functionality, and relies on a separate secure infrastructure. 625 o Discovery, synchronization and negotiation are designed together: 627 The discovery method and the synchronization and negotiation 628 methods are designed in the same way and can be combined when this 629 is useful, allowing a rapid mode of operation described in 630 Section 3.5.4. These processes can also be performed 631 independently when appropriate. 633 * Thus, for some objectives, especially those concerned with 634 application layer services, another discovery mechanism such as 635 the future DNS Service Discovery [RFC7558] MAY be used. The 636 choice is left to the designers of individual ASAs. 638 o A uniform pattern for technical objectives: 640 The synchronization and negotiation objectives are defined 641 according to a uniform pattern. The values that they contain 642 could be carried either in a simple binary format or in a complex 643 object format. The basic protocol design uses the Concise Binary 644 Object Representation (CBOR) [RFC7049], which is readily 645 extensible for unknown future requirements. 647 o A flexible model for synchronization: 649 GRASP supports synchronization between two nodes, which could be 650 used repeatedly to perform synchronization among a small number of 651 nodes. It also supports an unsolicited flooding mode when large 652 groups of nodes, possibly including all autonomic nodes, need data 653 for the same technical objective. 655 * There may be some network parameters for which a more 656 traditional flooding mechanism such as DNCP [RFC7787] is 657 considered more appropriate. GRASP can coexist with DNCP. 659 o A simple initiator/responder model for negotiation: 661 Multi-party negotiations are very complicated to model and cannot 662 readily be guaranteed to converge. GRASP uses a simple bilateral 663 model and can support multi-party negotiations by indirect steps. 665 o Organizing of synchronization or negotiation content: 667 The technical content transmitted by GRASP will be organized 668 according to the relevant function or service. The objectives for 669 different functions or services are kept separate, because they 670 may be negotiated or synchronized with different counterparts or 671 have different response times. Thus a normal arrangement would be 672 a single ASA managing a small set of closely related objectives, 673 with a version of that ASA in each relevant autonomic node. 674 Further discussion of this aspect is out of scope for the current 675 document. 677 o Requests and responses in negotiation procedures: 679 The initiator can negotiate a specific negotiation objective with 680 relevant counterpart ASAs. It can request relevant information 681 from a counterpart so that it can coordinate its local 682 configuration. It can request the counterpart to make a matching 683 configuration. It can request simulation or forecast results by 684 sending some dry run conditions. 686 Beyond the traditional yes/no answer, the responder can reply with 687 a suggested alternative value for the objective concerned. This 688 would start a bi-directional negotiation ending in a compromise 689 between the two ASAs. 691 o Convergence of negotiation procedures: 693 To enable convergence, when a responder suggests a new value or 694 condition in a negotiation step reply, it should be as close as 695 possible to the original request or previous suggestion. The 696 suggested value of later negotiation steps should be chosen 697 between the suggested values from the previous two steps. GRASP 698 provides mechanisms to guarantee convergence (or failure) in a 699 small number of steps, namely a timeout and a maximum number of 700 iterations. 702 o Extensibility: 704 GRASP does not have a version number, and could be extended by 705 adding new message types and options. In normal use, new 706 semantics will be added by defining new synchronization or 707 negotiation objectives. 709 3.4. Quick Operating Overview 711 An instance of GRASP is expected to run as a separate core module, 712 providing an API (such as [I-D.liu-anima-grasp-api]) to interface to 713 various ASAs. These ASAs may operate without special privilege, 714 unless they need it for other reasons (such as configuring IP 715 addresses or manipulating routing tables). 717 The GRASP mechanisms used by the ASA are built around GRASP 718 objectives defined as data structures containing administrative 719 information such as the objective's unique name, and its current 720 value. The format and size of the value is not restricted by the 721 protocol, except that it must be possible to serialise it for 722 transmission in CBOR, which is no restriction at all in practice. 724 GRASP provides the following mechanisms: 726 o A discovery mechanism (M_DISCOVERY, M_RESPONSE), by which an ASA 727 can discover other ASAs supporting a given objective. 729 o A negotiation request mechanism (M_REQ_NEG), by which an ASA can 730 start negotiation of an objective with a counterpart ASA. Once a 731 negotiation has started, the process is symmetrical, and there is 732 a negotiation step message (M_NEGOTIATE) for each ASA to use in 733 turn. Two other functions support negotiating steps (M_WAIT, 734 M_END). 736 o A synchronization mechanism (M_REQ_SYN), by which an ASA can 737 request the current value of an objective from a counterpart ASA. 738 With this, there is a corresponding response function (M_SYNCH) 739 for an ASA that wishes to respond to synchronization requests. 741 o A flood mechanism (M_FLOOD), by which an ASA can cause the current 742 value of an objective to be flooded throughout the autonomic 743 network so that any ASA can receive it. One application of this 744 is to act as an announcement, avoiding the need for discovery of a 745 widely applicable objective. 747 Some example messages and simple message flows are provided in 748 Appendix D. 750 3.5. GRASP Protocol Basic Properties and Mechanisms 752 3.5.1. Required External Security Mechanism 754 The protocol SHOULD always run within a secure Autonomic Control 755 Plane (ACP) [I-D.ietf-anima-autonomic-control-plane]. The ACP is 756 assumed to carry all messages securely, including link-local 757 multicast when it is virtualized over the ACP. A GRASP instance MUST 758 verify whether the ACP is operational. 760 If there is no ACP, one of the following alternatives applies: 762 1. The protocol instance MUST use another form of strong 763 authentication and SHOULD use a form of strong encryption. An 764 exception is that during initialization of nodes there will be a 765 transition period during which it might not be practical to run 766 with strong encryption. This period MUST be as short as 767 possible, changing to a fully secure setup as soon as possible. 768 See Section 3.5.2.1 for further discussion. 770 2. The protocol instance MUST operate as described in 771 Section 3.5.2.2 or Section 3.5.2.3. 773 Network interfaces could be at different security levels, for example 774 being part of the ACP or not. All the interfaces supported by a 775 given GRASP instance MUST be at the same security level. 777 The ACP, or in its absence another security mechanism, sets the 778 boundary within which nodes are trusted as GRASP peers. A GRASP 779 implementation MUST refuse to execute GRASP synchronization and 780 negotiation functions if there is neither an operational ACP nor 781 another secure environment. 783 Link-local multicast is used for discovery messages. Responses to 784 discovery messages MUST be secured, with one exception mentioned in 785 the next section. 787 3.5.2. Constrained Instances 789 This section describes some cases where additional instances of GRASP 790 subject to certain constraints are appropriate. 792 3.5.2.1. No ACP 794 As mentioned in Section 3.3, some GRASP operations might be performed 795 across an administrative domain boundary by mutual agreement, without 796 the benefit of an ACP. Such operations MUST be confined to a 797 separate instance of GRASP with its own copy of all GRASP data 798 structures. Messages MUST be authenticated and SHOULD be encrypted. 799 TLS [RFC5246] and DTLS [RFC6347] based on a Public Key Infrastructure 800 (PKI) [RFC5280] are RECOMMENDED for this purpose. Further details 801 are out of scope for this document. 803 3.5.2.2. Discovery Unsolicited Link-Local 805 Some services may need to use insecure GRASP discovery, response and 806 flood messages without being able to use pre-existing security 807 associations. Such operations being intrinsically insecure, they 808 need to be confined to link-local use to minimize the risk of 809 malicious actions. Possible examples include discovery of candidate 810 ACP neighbors [I-D.ietf-anima-autonomic-control-plane], discovery of 811 bootstrap proxies [I-D.ietf-anima-bootstrapping-keyinfra] or perhaps 812 initialization services in networks using GRASP without being fully 813 autonomic (e.g., no ACP). Such usage MUST be limited to link-local 814 operations and MUST be confined to a separate insecure instance of 815 GRASP with its own copy of all GRASP data structures. This instance 816 is nicknamed DULL - Discovery Unsolicited Link-Local. 818 The detailed rules for the DULL instance of GRASP are as follows: 820 o An initiator MUST only send Discovery or Flood Synchronization 821 link-local multicast messages with a loop count of 1. Other GRASP 822 message types MUST NOT be sent. 824 o A responder MUST silently discard any message whose loop count is 825 not 1. 827 o A responder MUST silently discard any message referring to a GRASP 828 Objective that is not directly part of a service that requires 829 this insecure mode. 831 o A responder MUST NOT relay any multicast messages. 833 o A Discovery Response MUST indicate a link-local address. 835 o A Discovery Response MUST NOT include a Divert option. 837 o A node MUST silently discard any message whose source address is 838 not link-local. 840 To minimize traffic possibly observed by third parties, GRASP traffic 841 SHOULD be minimized by using only Flood Synchronization to announce 842 objectives and their associated locators, rather than by using 843 Discovery and Response. Further details are out of scope for this 844 document 846 3.5.2.3. Secure Only Neighbor Negotiation 848 Some services might use insecure on-link operations as in DULL, but 849 also use unicast synchronization or negotiation operations protected 850 by TLS. A separate instance of GRASP is used, with its own copy of 851 all GRASP data structures. This instance is nicknamed SONN - Secure 852 Only Neighbor Negotiation. 854 The detailed rules for the SONN instance of GRASP are as follows: 856 o All types of GRASP message are permitted. 858 o An initiator MUST send any Discovery or Flood Synchronization 859 link-local multicast messages with a loop count of 1. 861 o A responder MUST silently discard any Discovery or Flood 862 Synchronization message whose loop count is not 1. 864 o A responder MUST silently discard any message referring to a GRASP 865 Objective that is not directly part of the service concerned. 867 o A responder MUST NOT relay any multicast messages. 869 o A Discovery Response MUST indicate a link-local address. 871 o A Discovery Response MUST NOT include a Divert option. 873 o A node MUST silently discard any message whose source address is 874 not link-local. 876 Further details are out of scope for this document. 878 3.5.3. Transport Layer Usage 880 GRASP discovery and flooding messages are designed for use over link- 881 local multicast UDP. They MUST NOT be fragmented, and therefore MUST 882 NOT exceed the link MTU size. 884 All other GRASP messages are unicast and could in principle run over 885 any transport protocol. An implementation MUST support use of TCP. 886 It MAY support use of another transport protocol. However, GRASP 887 itself does not provide for error detection or retransmission. Use 888 of an unreliable transport protocol is therefore NOT RECOMMENDED. 890 Nevertheless, when running within a secure ACP on reliable 891 infrastructure, UDP MAY be used for unicast messages not exceeding 892 the minimum IPv6 path MTU; however, TCP MUST be used for longer 893 messages. In other words, IPv6 fragmentation is avoided. If a node 894 receives a UDP message but the reply is too long, it MUST open a TCP 895 connection to the peer for the reply. Note that when the network is 896 under heavy load or in a fault condition, UDP might become 897 unreliable. Since this is when autonomic functions are most 898 necessary, automatic fallback to TCP MUST be implemented. The 899 simplest implementation is therefore to use only TCP. 901 For considerations when running without an ACP, see Section 3.5.2.1. 903 For link-local multicast, the GRASP protocol listens to the well- 904 known GRASP Listen Port (Section 3.6). For unicast transport 905 sessions used for discovery responses, synchronization and 906 negotiation, the ASA concerned normally listens on its own 907 dynamically assigned ports, which are communicated to its peers 908 during discovery. However, a minimal implementation MAY use the 909 GRASP Listen Port for this purpose. 911 3.5.4. Discovery Mechanism and Procedures 913 3.5.4.1. Separated discovery and negotiation mechanisms 915 Although discovery and negotiation or synchronization are defined 916 together in GRASP, they are separate mechanisms. The discovery 917 process could run independently from the negotiation or 918 synchronization process. Upon receiving a Discovery (Section 3.8.4) 919 message, the recipient node should return a response message in which 920 it either indicates itself as a discovery responder or diverts the 921 initiator towards another more suitable ASA. However, this response 922 may be delayed if the recipient needs to relay the discovery onwards, 923 as described below. 925 The discovery action (M_DISCOVERY) will normally be followed by a 926 negotiation (M_REQ_NEG) or synchronization (M_REQ_SYN) action. The 927 discovery results could be utilized by the negotiation protocol to 928 decide which ASA the initiator will negotiate with. 930 The initiator of a discovery action for a given objective need not be 931 capable of responding to that objective as a Negotiation Counterpart, 932 as a Synchronization Responder or as source for flooding. For 933 example, an ASA might perform discovery even if it only wishes to act 934 a Synchronization Initiator or Negotiation Initiator. Such an ASA 935 does not itself need to respond to discovery messages. 937 It is also entirely possible to use GRASP discovery without any 938 subsequent negotiation or synchronization action. In this case, the 939 discovered objective is simply used as a name during the discovery 940 process and any subsequent operations between the peers are outside 941 the scope of GRASP. 943 3.5.4.2. Discovery Overview 945 A complete discovery process will start with a multicast (of 946 M_DISCOVERY) on the local link. On-link neighbors supporting the 947 discovery objective will respond directly (with M_RESPONSE). A 948 neighbor with multiple interfaces will respond with a cached 949 discovery response if any. However, it SHOULD NOT respond with a 950 cached response on an interface if it learnt that information from 951 the same interface, because the peer in question will answer directly 952 if still operational. If it has no cached response, it will relay 953 the discovery on its other interfaces, for example reaching a higher- 954 level gateway in a hierarchical network. If a node receiving the 955 relayed discovery supports the discovery objective, it will respond 956 to the relayed discovery. If it has a cached response, it will 957 respond with that. If not, it will repeat the discovery process, 958 which thereby becomes iterative. The loop count and timeout will 959 ensure that the process ends. 961 A Discovery message MAY be sent unicast (via UDP or TCP) to a peer 962 node, which SHOULD then proceed exactly as if the message had been 963 multicast, except that when TCP is used, the response will be on the 964 same socket as the query. However, this mode does not guarantee 965 successful discovery in the general case. 967 3.5.4.3. Discovery Procedures 969 Discovery starts as an on-link operation. The Divert option can tell 970 the discovery initiator to contact an off-link ASA for that discovery 971 objective. A Discovery message is sent by a discovery initiator via 972 UDP to the ALL_GRASP_NEIGHBORS link-local multicast address 973 (Section 3.6). Every network device that supports GRASP always 974 listens to a well-known UDP port to capture the discovery messages. 975 Because this port is unique in a device, this is a function of the 976 GRASP instance and not of an individual ASA. As a result, each ASA 977 will need to register the objectives that it supports with the local 978 GRASP instance. 980 If an ASA in a neighbor device supports the requested discovery 981 objective, the device SHOULD respond to the link-local multicast with 982 a unicast Discovery Response message (Section 3.8.5) with locator 983 option(s), unless it is temporarily unavailable. Otherwise, if the 984 neighbor has cached information about an ASA that supports the 985 requested discovery objective (usually because it discovered the same 986 objective before), it SHOULD respond with a Discovery Response 987 message with a Divert option pointing to the appropriate Discovery 988 Responder. 990 If a device has no information about the requested discovery 991 objective, and is not acting as a discovery relay (see below) it MUST 992 silently discard the Discovery message. 994 If no discovery response is received within a reasonable timeout 995 (default GRASP_DEF_TIMEOUT milliseconds, Section 3.6), the Discovery 996 message MAY be repeated, with a newly generated Session ID 997 (Section 3.7). An exponential backoff SHOULD be used for subsequent 998 repetitions, to limit the load during busy periods. Frequent 999 repetition might be symptomatic of a denial of service attack. 1001 After a GRASP device successfully discovers a locator for a Discovery 1002 Responder supporting a specific objective, it MUST cache this 1003 information, including the interface index via which it was 1004 discovered. This cache record MAY be used for future negotiation or 1005 synchronization, and the locator SHOULD be passed on when appropriate 1006 as a Divert option to another Discovery Initiator. 1008 The cache mechanism MUST include a lifetime for each entry. The 1009 lifetime is derived from a time-to-live (ttl) parameter in each 1010 Discovery Response message. Cached entries MUST be ignored or 1011 deleted after their lifetime expires. In some environments, 1012 unplanned address renumbering might occur. In such cases, the 1013 lifetime SHOULD be short compared to the typical address lifetime and 1014 a mechanism to flush the discovery cache MUST be implemented. The 1015 discovery mechanism needs to track the node's current address to 1016 ensure that Discovery Responses always indicate the correct address. 1018 If multiple Discovery Responders are found for the same objective, 1019 they SHOULD all be cached, unless this creates a resource shortage. 1020 The method of choosing between multiple responders is an 1021 implementation choice. This choice MUST be available to each ASA but 1022 the GRASP implementation SHOULD provide a default choice. 1024 Because Discovery Responders will be cached in a finite cache, they 1025 might be deleted at any time. In this case, discovery will need to 1026 be repeated. If an ASA exits for any reason, its locator might still 1027 be cached for some time, and attempts to connect to it will fail. 1028 ASAs need to be robust in these circumstances. 1030 3.5.4.4. Discovery Relaying 1032 A GRASP instance with multiple link-layer interfaces (typically 1033 running in a router) MUST support discovery on all interfaces. We 1034 refer to this as a 'relaying instance'. 1036 Constrained Instances (Section 3.5.2) are always single-interface 1037 instances and therefore MUST NOT perform discovery relaying. 1039 If a relaying instance receives a Discovery message on a given 1040 interface for a specific objective that it does not support and for 1041 which it has not previously cached a Discovery Responder, it MUST 1042 relay the query by re-issuing a new Discovery message as a link-local 1043 multicast on its other interfaces. 1045 The relayed discovery message MUST have the same Session ID as the 1046 incoming discovery message and MUST be tagged with the IP address of 1047 its original initiator (see Section 3.8.4). Note that this initiator 1048 address is only used to allow for disambiguation of the Session ID 1049 and is never used to address Response packets, which are sent to the 1050 relaying instance, not the original initiator. 1052 Since the relay device is unaware of the timeout set by the original 1053 initiator it SHOULD set a timeout at least equal to GRASP_DEF_TIMEOUT 1054 milliseconds. 1056 The relaying instance MUST decrement the loop count within the 1057 objective, and MUST NOT relay the Discovery message if the result is 1058 zero. Also, it MUST limit the total rate at which it relays 1059 discovery messages to a reasonable value, in order to mitigate 1060 possible denial of service attacks. It MUST cache the Session ID 1061 value and initiator address of each relayed Discovery message until 1062 any Discovery Responses have arrived or the discovery process has 1063 timed out. To prevent loops, it MUST NOT relay a Discovery message 1064 which carries a given cached Session ID and initiator address more 1065 than once. These precautions avoid discovery loops and mitigate 1066 potential overload. 1068 The discovery results received by the relaying instance MUST in turn 1069 be sent as a Discovery Response message to the Discovery message that 1070 caused the relay action. 1072 3.5.4.5. Rapid Mode (Discovery/Negotiation binding) 1074 A Discovery message MAY include a Negotiation Objective option. This 1075 allows a rapid mode of negotiation described in Section 3.5.5. A 1076 similar mechanism is defined for synchronization in Section 3.5.6. 1078 Note that rapid mode is currently limited to a single objective for 1079 simplicity of design and implementation. A possible future extension 1080 is to allow multiple objectives in rapid mode for greater efficiency. 1082 3.5.5. Negotiation Procedures 1084 A negotiation initiator sends a negotiation request (using M_REQ_NEG) 1085 to a counterpart ASA, including a specific negotiation objective. It 1086 may request the negotiation counterpart to make a specific 1087 configuration. Alternatively, it may request a certain simulation or 1088 forecast result by sending a dry run configuration. The details, 1089 including the distinction between dry run and an actual configuration 1090 change, will be defined separately for each type of negotiation 1091 objective. 1093 If no reply message of any kind is received within a reasonable 1094 timeout (default GRASP_DEF_TIMEOUT milliseconds, Section 3.6), the 1095 negotiation request MAY be repeated, with a newly generated Session 1096 ID (Section 3.7). An exponential backoff SHOULD be used for 1097 subsequent repetitions. 1099 If the counterpart can immediately apply the requested configuration, 1100 it will give an immediate positive (O_ACCEPT) answer (using M_END). 1101 This will end the negotiation phase immediately. Otherwise, it will 1102 negotiate (using M_NEGOTIATE). It will reply with a proposed 1103 alternative configuration that it can apply (typically, a 1104 configuration that uses fewer resources than requested by the 1105 negotiation initiator). This will start a bi-directional negotiation 1106 (using M_NEGOTIATE) to reach a compromise between the two ASAs. 1108 The negotiation procedure is ended when one of the negotiation peers 1109 sends a Negotiation Ending (M_END) message, which contains an accept 1110 (O_ACCEPT) or decline (O_DECLINE) option and does not need a response 1111 from the negotiation peer. Negotiation may also end in failure 1112 (equivalent to a decline) if a timeout is exceeded or a loop count is 1113 exceeded. 1115 A negotiation procedure concerns one objective and one counterpart. 1116 Both the initiator and the counterpart may take part in simultaneous 1117 negotiations with various other ASAs, or in simultaneous negotiations 1118 about different objectives. Thus, GRASP is expected to be used in a 1119 multi-threaded mode. Certain negotiation objectives may have 1120 restrictions on multi-threading, for example to avoid over-allocating 1121 resources. 1123 Some configuration actions, for example wavelength switching in 1124 optical networks, might take considerable time to execute. The ASA 1125 concerned needs to allow for this by design, but GRASP does allow for 1126 a peer to insert latency in a negotiation process if necessary 1127 (Section 3.8.9, M_WAIT). 1129 3.5.5.1. Rapid Mode (Discovery/Negotiation Linkage) 1131 A Discovery message MAY include a Negotiation Objective option. In 1132 this case it is as if the initiator sent the sequence M_DISCOVERY, 1133 immediately followed by M_REQ_NEG. This has implications for the 1134 construction of the GRASP core, as it must carefully pass the 1135 contents of the Negotiation Objective option to the ASA so that it 1136 may evaluate the objective directly. When a Negotiation Objective 1137 option is present the ASA replies with an M_NEGOTIATE message (or 1138 M_END with O_ACCEPT if it is immediately satisfied with the 1139 proposal), rather than with an M_RESPONSE. However, if the recipient 1140 node does not support rapid mode, discovery will continue normally. 1142 It is possible that a Discovery Response will arrive from a responder 1143 that does not support rapid mode, before such a Negotiation message 1144 arrives. In this case, rapid mode will not occur. 1146 This rapid mode could reduce the interactions between nodes so that a 1147 higher efficiency could be achieved. However, a network in which 1148 some nodes support rapid mode and others do not will have complex 1149 timing-dependent behaviors. Therefore, the rapid negotiation 1150 function SHOULD be disabled by default. 1152 3.5.6. Synchronization and Flooding Procedures 1154 3.5.6.1. Unicast Synchronization 1156 A synchronization initiator sends a synchronization request to a 1157 counterpart, including a specific synchronization objective. The 1158 counterpart responds with a Synchronization message (Section 3.8.10) 1159 containing the current value of the requested synchronization 1160 objective. No further messages are needed. 1162 If no reply message of any kind is received within a reasonable 1163 timeout (default GRASP_DEF_TIMEOUT milliseconds, Section 3.6), the 1164 synchronization request MAY be repeated, with a newly generated 1165 Session ID (Section 3.7). An exponential backoff SHOULD be used for 1166 subsequent repetitions. 1168 3.5.6.2. Flooding 1170 In the case just described, the message exchange is unicast and 1171 concerns only one synchronization objective. For large groups of 1172 nodes requiring the same data, synchronization flooding is available. 1173 For this, a flooding initiator MAY send an unsolicited Flood 1174 Synchronization message containing one or more Synchronization 1175 Objective option(s), if and only if the specification of those 1176 objectives permits it. This is sent as a multicast message to the 1177 ALL_GRASP_NEIGHBORS multicast address (Section 3.6). 1179 Receiving flood multicasts is a function of the GRASP core, as in the 1180 case of discovery multicasts (Section 3.5.4.3). 1182 To ensure that flooding does not result in a loop, the originator of 1183 the Flood Synchronization message MUST set the loop count in the 1184 objectives to a suitable value (the default is GRASP_DEF_LOOPCT). 1185 Also, a suitable mechanism is needed to avoid excessive multicast 1186 traffic. This mechanism MUST be defined as part of the specification 1187 of the synchronization objective(s) concerned. It might be a simple 1188 rate limit or a more complex mechanism such as the Trickle algorithm 1189 [RFC6206]. 1191 A GRASP device with multiple link-layer interfaces (typically a 1192 router) MUST support synchronization flooding on all interfaces. If 1193 it receives a multicast Flood Synchronization message on a given 1194 interface, it MUST relay it by re-issuing a Flood Synchronization 1195 message on its other interfaces. The relayed message MUST have the 1196 same Session ID as the incoming message and MUST be tagged with the 1197 IP address of its original initiator. 1199 Link-layer Flooding is supported by GRASP by setting the loop count 1200 to 1, and sending with a link-local source address. Floods with 1201 link-local source addresses and a loop count other than 1 are 1202 invalid, and such messages MUST be discarded. 1204 The relaying device MUST decrement the loop count within the first 1205 objective, and MUST NOT relay the Flood Synchronization message if 1206 the result is zero. Also, it MUST limit the total rate at which it 1207 relays Flood Synchronization messages to a reasonable value, in order 1208 to mitigate possible denial of service attacks. It MUST cache the 1209 Session ID value and initiator address of each relayed Flood 1210 Synchronization message for a time not less than twice 1211 GRASP_DEF_TIMEOUT milliseconds. To prevent loops, it MUST NOT relay 1212 a Flood Synchronization message which carries a given cached Session 1213 ID and initiator address more than once. These precautions avoid 1214 synchronization loops and mitigate potential overload. 1216 Note that this mechanism is unreliable in the case of sleeping nodes, 1217 or new nodes that join the network, or nodes that rejoin the network 1218 after a fault. An ASA that initiates a flood SHOULD repeat the flood 1219 at a suitable frequency and SHOULD also act as a synchronization 1220 responder for the objective(s) concerned. Thus nodes that require an 1221 objective subject to flooding can either wait for the next flood or 1222 request unicast synchronization for that objective. 1224 The multicast messages for synchronization flooding are subject to 1225 the security rules in Section 3.5.1. In practice this means that 1226 they MUST NOT be transmitted and MUST be ignored on receipt unless 1227 there is an operational ACP or equivalent strong security in place. 1228 However, because of the security weakness of link-local multicast 1229 (Section 5), synchronization objectives that are flooded SHOULD NOT 1230 contain unencrypted private information and SHOULD be validated by 1231 the recipient ASA. 1233 3.5.6.3. Rapid Mode (Discovery/Synchronization Linkage) 1235 A Discovery message MAY include a Synchronization Objective option. 1236 In this case the Discovery message also acts as a Request 1237 Synchronization message to indicate to the Discovery Responder that 1238 it could directly reply to the Discovery Initiator with a 1239 Synchronization message Section 3.8.10 with synchronization data for 1240 rapid processing, if the discovery target supports the corresponding 1241 synchronization objective. The design implications are similar to 1242 those discussed in Section 3.5.5.1. 1244 It is possible that a Discovery Response will arrive from a responder 1245 that does not support rapid mode, before such a Synchronization 1246 message arrives. In this case, rapid mode will not occur. 1248 This rapid mode could reduce the interactions between nodes so that a 1249 higher efficiency could be achieved. However, a network in which 1250 some nodes support rapid mode and others do not will have complex 1251 timing-dependent behaviors. Therefore, the rapid synchronization 1252 function SHOULD be configured off by default and MAY be configured on 1253 or off by Intent. 1255 3.6. GRASP Constants 1257 o ALL_GRASP_NEIGHBORS 1259 A link-local scope multicast address used by a GRASP-enabled 1260 device to discover GRASP-enabled neighbor (i.e., on-link) devices. 1261 All devices that support GRASP are members of this multicast 1262 group. 1264 * IPv6 multicast address: TBD1 1266 * IPv4 multicast address: TBD2 1268 o GRASP_LISTEN_PORT (TBD3) 1270 A well-known UDP user port that every GRASP-enabled network device 1271 MUST always listen to for link-local multicasts. This user port 1272 MAY also be used to listen for TCP or UDP unicast messages in a 1273 simple implementation of GRASP (Section 3.5.3). 1275 o GRASP_DEF_TIMEOUT (60000 milliseconds) 1276 The default timeout used to determine that a discovery etc. has 1277 failed to complete. 1279 o GRASP_DEF_LOOPCT (6) 1281 The default loop count used to determine that a negotiation has 1282 failed to complete, and to avoid looping messages. 1284 o GRASP_DEF_MAX_SIZE (2048) 1286 The default maximum message size in bytes. 1288 3.7. Session Identifier (Session ID) 1290 This is an up to 32-bit opaque value used to distinguish multiple 1291 sessions between the same two devices. A new Session ID MUST be 1292 generated by the initiator for every new Discovery, Flood 1293 Synchronization or Request message. All responses and follow-up 1294 messages in the same discovery, synchronization or negotiation 1295 procedure MUST carry the same Session ID. 1297 The Session ID SHOULD have a very low collision rate locally. It 1298 MUST be generated by a pseudo-random algorithm using a locally 1299 generated seed which is unlikely to be used by any other device in 1300 the same network [RFC4086]. When allocating a new Session ID, GRASP 1301 MUST check that the value is not already in use and SHOULD check that 1302 it has not been used recently, by consulting a cache of current and 1303 recent sessions. In the unlikely event of a clash, GRASP MUST 1304 generate a new value. 1306 However, there is a finite probability that two nodes might generate 1307 the same Session ID value. For that reason, when a Session ID is 1308 communicated via GRASP, the receiving node MUST tag it with the 1309 initiator's IP address to allow disambiguation. In the highly 1310 unlikely event of two peers opening sessions with the same Session ID 1311 value, this tag will allow the two sessions to be distinguished. 1312 Multicast GRASP messages and their responses, which may be relayed 1313 between links, therefore include a field that carries the initiator's 1314 global IP address. 1316 There is a highly unlikely race condition in which two peers start 1317 simultaneous negotiation sessions with each other using the same 1318 Session ID value. Depending on various implementation choices, this 1319 might lead to the two sessions being confused. See Section 3.8.6 for 1320 details of how to avoid this. 1322 3.8. GRASP Messages 1324 3.8.1. Message Overview 1326 This section defines the GRASP message format and message types. 1327 Message types not listed here are reserved for future use. 1329 The messages currently defined are: 1331 Discovery and Discovery Response (M_DISCOVERY, M_RESPONSE). 1333 Request Negotiation, Negotiation, Confirm Waiting and Negotiation 1334 End (M_REQ_NEG, M_NEGOTIATE, M_WAIT, M_END). 1336 Request Synchronization, Synchronization, and Flood 1337 Synchronization (M_REQ_SYN, M_SYNCH, M_FLOOD. 1339 No Operation and Invalid (M_NOOP, M_INVALID). 1341 3.8.2. GRASP Message Format 1343 GRASP messages share an identical header format and a variable format 1344 area for options. GRASP message headers and options are transmitted 1345 in Concise Binary Object Representation (CBOR) [RFC7049]. In this 1346 specification, they are described using CBOR data definition language 1347 (CDDL) [I-D.greevenbosch-appsawg-cbor-cddl]. Fragmentary CDDL is 1348 used to describe each item in this section. A complete and normative 1349 CDDL specification of GRASP is given in Section 6, including 1350 constants such as message types. 1352 Every GRASP message, except the No Operation message, carries a 1353 Session ID (Section 3.7). Options are then presented serially in the 1354 options field. 1356 In fragmentary CDDL, every GRASP message follows the pattern: 1358 grasp-message = (message .within message-structure) / noop-message 1360 message-structure = [MESSAGE_TYPE, session-id, ?initiator, 1361 *grasp-option] 1363 MESSAGE_TYPE = 1..255 1364 session-id = 0..4294967295 ;up to 32 bits 1365 grasp-option = any 1367 The MESSAGE_TYPE indicates the type of the message and thus defines 1368 the expected options. Any options received that are not consistent 1369 with the MESSAGE_TYPE SHOULD be silently discarded. 1371 The No Operation (noop) message is described in Section 3.8.13. 1373 The various MESSAGE_TYPE values are defined in Section 6. 1375 All other message elements are described below and formally defined 1376 in Section 6. 1378 If an unrecognized MESSAGE_TYPE is received in a unicast message, an 1379 Invalid message (Section 3.8.12) MAY be returned. Otherwise the 1380 message MAY be logged and MUST be discarded. If an unrecognized 1381 MESSAGE_TYPE is received in a multicast message, it MAY be logged and 1382 MUST be silently discarded. 1384 3.8.3. Message Size 1386 GRASP nodes MUST be able to receive unicast messages of at least 1387 GRASP_DEF_MAX_SIZE bytes. GRASP nodes MUST NOT send unicast messages 1388 longer than GRASP_DEF_MAX_SIZE bytes unless a longer size is 1389 explicitly allowed for the objective concerned. For example, GRASP 1390 negotiation itself could be used to agree on a longer message size. 1392 The message parser used by GRASP should be configured to know about 1393 the GRASP_DEF_MAX_SIZE, or any larger negotiated message size, so 1394 that it may defend against overly long messages. 1396 The maximum size of multicast messages (M_DISCOVERY and M_FLOOD) 1397 depends on the link layer technology or link adaptation layer in use. 1399 3.8.4. Discovery Message 1401 In fragmentary CDDL, a Discovery message follows the pattern: 1403 discovery-message = [M_DISCOVERY, session-id, initiator, objective] 1405 A discovery initiator sends a Discovery message to initiate a 1406 discovery process for a particular objective option. 1408 The discovery initiator sends all Discovery messages via UDP to port 1409 GRASP_LISTEN_PORT at the link-local ALL_GRASP_NEIGHBORS multicast 1410 address on each link-layer interface in use by GRASP. It then 1411 listens for unicast TCP responses on a given port, and stores the 1412 discovery results (including responding discovery objectives and 1413 corresponding unicast locators). 1415 The listening port used for TCP MUST be the same port as used for 1416 sending the Discovery UDP multicast, on a given interface. In a low- 1417 end implementation this MAY be GRASP_LISTEN_PORT. In a more complex 1418 implementation, the GRASP discovery mechanism will find, for each 1419 interface, a dynamic port that it can bind to for both UDP and TCP 1420 before initiating any discovery. 1422 The 'initiator' field in the message is a globally unique IP address 1423 of the initiator, for the sole purpose of disambiguating the Session 1424 ID in other nodes. If for some reason the initiator does not have a 1425 globally unique IP address, it MUST use a link-local address for this 1426 purpose that is highly likely to be unique, for example using 1427 [RFC7217]. 1429 A Discovery message MUST include exactly one of the following: 1431 o a discovery objective option (Section 3.10.1). Its loop count 1432 MUST be set to a suitable value to prevent discovery loops 1433 (default value is GRASP_DEF_LOOPCT). If the discovery initiator 1434 requires only on-link responses, the loop count MUST be set to 1. 1436 o a negotiation objective option (Section 3.10.1). This is used 1437 both for the purpose of discovery and to indicate to the discovery 1438 target that it MAY directly reply to the discovery initiatior with 1439 a Negotiation message for rapid processing, if it could act as the 1440 corresponding negotiation counterpart. The sender of such a 1441 Discovery message MUST initialize a negotiation timer and loop 1442 count in the same way as a Request Negotiation message 1443 (Section 3.8.6). 1445 o a synchronization objective option (Section 3.10.1). This is used 1446 both for the purpose of discovery and to indicate to the discovery 1447 target that it MAY directly reply to the discovery initiator with 1448 a Synchronization message for rapid processing, if it could act as 1449 the corresponding synchronization counterpart. Its loop count 1450 MUST be set to a suitable value to prevent discovery loops 1451 (default value is GRASP_DEF_LOOPCT). 1453 As mentioned in Section 3.5.4.2, a Discovery message MAY be sent 1454 unicast to a peer node, which SHOULD then proceed exactly as if the 1455 message had been multicast. 1457 3.8.5. Discovery Response Message 1459 In fragmentary CDDL, a Discovery Response message follows the 1460 pattern: 1462 response-message = [M_RESPONSE, session-id, initiator, ttl, 1463 (+locator-option // divert-option), ?objective)] 1465 ttl = 0..4294967295 ; in milliseconds 1467 A node which receives a Discovery message SHOULD send a Discovery 1468 Response message if and only if it can respond to the discovery. 1470 It MUST contain the same Session ID and initiator as the Discovery 1471 message. 1473 It MUST contain a time-to-live (ttl) for the validity of the 1474 response, given as a positive integer value in milliseconds. Zero 1475 is treated as the default value GRASP_DEF_TIMEOUT (Section 3.6). 1477 It MAY include a copy of the discovery objective from the 1478 Discovery message. 1480 It is sent to the sender of the Discovery message via TCP at the port 1481 used to send the Discovery message (as explained in Section 3.8.4). 1482 In the case of a relayed Discovery message, the Discovery Response is 1483 thus sent to the relay, not the original initiator. 1485 If the responding node supports the discovery objective of the 1486 discovery, it MUST include at least one kind of locator option 1487 (Section 3.9.5) to indicate its own location. A sequence of multiple 1488 kinds of locator options (e.g. IP address option and FQDN option) is 1489 also valid. 1491 If the responding node itself does not support the discovery 1492 objective, but it knows the locator of the discovery objective, then 1493 it SHOULD respond to the discovery message with a divert option 1494 (Section 3.9.2) embedding a locator option or a combination of 1495 multiple kinds of locator options which indicate the locator(s) of 1496 the discovery objective. 1498 More details on the processing of Discovery Responses are given in 1499 Section 3.5.4. 1501 3.8.6. Request Messages 1503 In fragmentary CDDL, Request Negotiation and Request Synchronization 1504 messages follow the patterns: 1506 request-negotiation-message = [M_REQ_NEG, session-id, objective] 1508 request-synchronization-message = [M_REQ_SYN, session-id, objective] 1510 A negotiation or synchronization requesting node sends the 1511 appropriate Request message to the unicast address (directly stored 1512 or resolved from an FQDN or URI) of the negotiation or 1513 synchronization counterpart, using the appropriate protocol and port 1514 numbers (selected from the discovery results). 1516 A Request message MUST include the relevant objective option. In the 1517 case of Request Negotiation, the objective option MUST include the 1518 requested value. 1520 When an initiator sends a Request Negotiation message, it MUST 1521 initialize a negotiation timer for the new negotiation thread. The 1522 default is GRASP_DEF_TIMEOUT milliseconds. Unless this timeout is 1523 modified by a Confirm Waiting message (Section 3.8.9), the initiator 1524 will consider that the negotiation has failed when the timer expires. 1526 Similarly, when an initiator sends a Request Synchronization, it 1527 SHOULD initialize a synchronization timer. The default is 1528 GRASP_DEF_TIMEOUT milliseconds. The initiator will consider that 1529 synchronization has failed if there is no response before the timer 1530 expires. 1532 When an initiator sends a Request message, it MUST initialize the 1533 loop count of the objective option with a value defined in the 1534 specification of the option or, if no such value is specified, with 1535 GRASP_DEF_LOOPCT. 1537 If a node receives a Request message for an objective for which no 1538 ASA is currently listening, it MUST immediately close the relevant 1539 socket to indicate this to the initiator. This is to avoid 1540 unnecessary timeouts if, for example, an ASA exits prematurely but 1541 the GRASP core is listening on its behalf. 1543 To avoid the highly unlikely race condition in which two nodes 1544 simultaneously request sessions with each other using the same 1545 Session ID (Section 3.7), when a node receives a Request message, it 1546 MUST verify that the received Session ID is not already locally 1547 active. In case of a clash, it MUST discard the Request message, in 1548 which case the initiator will detect a timeout. 1550 3.8.7. Negotiation Message 1552 In fragmentary CDDL, a Negotiation message follows the pattern: 1554 negotiate-message = [M_NEGOTIATE, session-id, objective] 1556 A negotiation counterpart sends a Negotiation message in response to 1557 a Request Negotiation message, a Negotiation message, or a Discovery 1558 message in Rapid Mode. A negotiation process MAY include multiple 1559 steps. 1561 The Negotiation message MUST include the relevant Negotiation 1562 Objective option, with its value updated according to progress in the 1563 negotiation. The sender MUST decrement the loop count by 1. If the 1564 loop count becomes zero the message MUST NOT be sent. In this case 1565 the negotiation session has failed and will time out. 1567 3.8.8. Negotiation End Message 1569 In fragmentary CDDL, a Negotiation End message follows the pattern: 1571 end-message = [M_END, session-id, accept-option / decline-option] 1573 A negotiation counterpart sends an Negotiation End message to close 1574 the negotiation. It MUST contain either an accept or a decline 1575 option, defined in Section 3.9.3 and Section 3.9.4. It could be sent 1576 either by the requesting node or the responding node. 1578 3.8.9. Confirm Waiting Message 1580 In fragmentary CDDL, a Confirm Waiting message follows the pattern: 1582 wait-message = [M_WAIT, session-id, waiting-time] 1583 waiting-time = 0..4294967295 ; in milliseconds 1585 A responding node sends a Confirm Waiting message to ask the 1586 requesting node to wait for a further negotiation response. It might 1587 be that the local process needs more time or that the negotiation 1588 depends on another triggered negotiation. This message MUST NOT 1589 include any other options. When received, the waiting time value 1590 overwrites and restarts the current negotiation timer 1591 (Section 3.8.6). 1593 The responding node SHOULD send a Negotiation, Negotiation End or 1594 another Confirm Waiting message before the negotiation timer expires. 1595 If not, the initiator MUST abandon or restart the negotiation 1596 procedure, to avoid an indefinite wait. 1598 3.8.10. Synchronization Message 1600 In fragmentary CDDL, a Synchronization message follows the pattern: 1602 synch-message = [M_SYNCH, session-id, objective] 1604 A node which receives a Request Synchronization, or a Discovery 1605 message in Rapid Mode, sends back a unicast Synchronization message 1606 with the synchronization data, in the form of a GRASP Option for the 1607 specific synchronization objective present in the Request 1608 Synchronization. 1610 3.8.11. Flood Synchronization Message 1612 In fragmentary CDDL, a Flood Synchronization message follows the 1613 pattern: 1615 flood-message = [M_FLOOD, session-id, initiator, ttl, 1616 +[objective, (locator-option / [])]] 1618 ttl = 0..4294967295 ; in milliseconds 1620 A node MAY initiate flooding by sending an unsolicited Flood 1621 Synchronization Message with synchronization data. This MAY be sent 1622 to port GRASP_LISTEN_PORT at the link-local ALL_GRASP_NEIGHBORS 1623 multicast address, in accordance with the rules in Section 3.5.6. 1625 The initiator address is provided, as described for Discovery 1626 messages (Section 3.8.4), only to disambiguate the Session ID. 1628 The message MUST contain a time-to-live (ttl) for the validity of 1629 the contents, given as a positive integer value in milliseconds. 1630 There is no default; zero indicates an indefinite lifetime. 1632 The synchronization data are in the form of GRASP Option(s) for 1633 specific synchronization objective(s). The loop count(s) MUST be 1634 set to a suitable value to prevent flood loops (default value is 1635 GRASP_DEF_LOOPCT). 1637 Each objective option MAY be followed by a locator option 1638 associated with the flooded objective. In its absence, an empty 1639 option MUST be included to indicate a null locator. 1641 A node that receives a Flood Synchronization message MUST cache the 1642 received objectives for use by local ASAs. Each cached objective 1643 MUST be tagged with the locator option sent with it, or with a null 1644 tag if an empty locator option was sent. If a subsequent Flood 1645 Synchronization message carrying the same objective arrives with the 1646 same tag, the corresponding cached copy of the objective MUST be 1647 overwritten. If a subsequent Flood Synchronization message carrying 1648 the same objective arrives with a different tag, a new cached entry 1649 MUST be created. 1651 Note: the purpose of this mechanism is to allow the recipient of 1652 flooded values to distinguish between different senders of the same 1653 objective, and if necessary communicate with them using the locator, 1654 protocol and port included in the locator option. Many objectives 1655 will not need this mechanism, so they will be flooded with a null 1656 locator. 1658 Cached entries MUST be ignored or deleted after their lifetime 1659 expires. 1661 3.8.12. Invalid Message 1663 In fragmentary CDDL, an Invalid message follows the pattern: 1665 invalid-message = [M_INVALID, session-id, ?any] 1667 This message MAY be sent by an implementation in response to an 1668 incoming unicast message that it considers invalid. The session-id 1669 MUST be copied from the incoming message. The content SHOULD be 1670 diagnostic information such as a partial copy of the invalid message. 1671 An M_INVALID message MAY be silently ignored by a recipient. 1672 However, it could be used in support of extensibility, since it 1673 indicates that the remote node does not support a new or obsolete 1674 message or option. 1676 An M_INVALID message MUST NOT be sent in response to an M_INVALID 1677 message. 1679 3.8.13. No Operation Message 1681 In fragmentary CDDL, a No Operation message follows the pattern: 1683 noop-message = [M_NOOP] 1685 This message MAY be sent by an implementation that for practical 1686 reasons needs to initialize a socket. It MUST be silently ignored by 1687 a recipient. 1689 3.9. GRASP Options 1691 This section defines the GRASP options for the negotiation and 1692 synchronization protocol signaling. Additional options may be 1693 defined in the future. 1695 3.9.1. Format of GRASP Options 1697 GRASP options are CBOR objects that MUST start with an unsigned 1698 integer identifying the specific option type carried in this option. 1699 These option types are formally defined in Section 6. Apart from 1700 that the only format requirement is that each option MUST be a well- 1701 formed CBOR object. In general a CBOR array format is RECOMMENDED to 1702 limit overhead. 1704 GRASP options may be defined to include encapsulated GRASP options. 1706 3.9.2. Divert Option 1708 The Divert option is used to redirect a GRASP request to another 1709 node, which may be more appropriate for the intended negotiation or 1710 synchronization. It may redirect to an entity that is known as a 1711 specific negotiation or synchronization counterpart (on-link or off- 1712 link) or a default gateway. The divert option MUST only be 1713 encapsulated in Discovery Response messages. If found elsewhere, it 1714 SHOULD be silently ignored. 1716 A discovery initiator MAY ignore a Divert option if it only requires 1717 direct discovery responses. 1719 In fragmentary CDDL, the Divert option follows the pattern: 1721 divert-option = [O_DIVERT, +locator-option] 1723 The embedded Locator Option(s) (Section 3.9.5) point to diverted 1724 destination target(s) in response to a Discovery message. 1726 3.9.3. Accept Option 1728 The accept option is used to indicate to the negotiation counterpart 1729 that the proposed negotiation content is accepted. 1731 The accept option MUST only be encapsulated in Negotiation End 1732 messages. If found elsewhere, it SHOULD be silently ignored. 1734 In fragmentary CDDL, the Accept option follows the pattern: 1736 accept-option = [O_ACCEPT] 1738 3.9.4. Decline Option 1740 The decline option is used to indicate to the negotiation counterpart 1741 the proposed negotiation content is declined and end the negotiation 1742 process. 1744 The decline option MUST only be encapsulated in Negotiation End 1745 messages. If found elsewhere, it SHOULD be silently ignored. 1747 In fragmentary CDDL, the Decline option follows the pattern: 1749 decline-option = [O_DECLINE, ?reason] 1750 reason = text ;optional error message 1752 Note: there might be scenarios where an ASA wants to decline the 1753 proposed value and restart the negotiation process. In this case it 1754 is an implementation choice whether to send a Decline option or to 1755 continue with a Negotiate message, with an objective option that 1756 contains a null value, or one that contains a new value that might 1757 achieve convergence. 1759 3.9.5. Locator Options 1761 These locator options are used to present reachability information 1762 for an ASA, a device or an interface. They are Locator IPv6 Address 1763 Option, Locator IPv4 Address Option, Locator FQDN (Fully Qualified 1764 Domain Name) Option and URI (Uniform Resource Identifier) Option. 1766 Since ASAs will normally run as independent user programs, locator 1767 options need to indicate the network layer locator plus the transport 1768 protocol and port number for reaching the target. For this reason, 1769 the Locator Options for IP addresses and FQDNs include this 1770 information explicitly. In the case of the URI Option, this 1771 information can be encoded in the URI itself. 1773 Note: It is assumed that all locators are in scope throughout the 1774 GRASP domain. GRASP is not intended to work across disjoint 1775 addressing or naming realms. 1777 3.9.5.1. Locator IPv6 address option 1779 In fragmentary CDDL, the IPv6 address option follows the pattern: 1781 ipv6-locator-option = [O_IPv6_LOCATOR, ipv6-address, 1782 transport-proto, port-number] 1783 ipv6-address = bytes .size 16 1785 transport-proto = IPPROTO_TCP / IPPROTO_UDP 1786 IPPROTO_TCP = 6 1787 IPPROTO_UDP = 17 1788 port-number = 0..65535 1790 The content of this option is a binary IPv6 address followed by the 1791 protocol number and port number to be used. 1793 Note 1: The IPv6 address MUST normally have global scope. However, 1794 during initialization, a link-local address MAY be used for specific 1795 objectives only (Section 3.5.2). In this case the corresponding 1796 Discovery Response message MUST be sent via the interface to which 1797 the link-local address applies. 1799 Note 2: A link-local IPv6 address MUST NOT be used when this option 1800 is included in a Divert option. 1802 3.9.5.2. Locator IPv4 address option 1804 In fragmentary CDDL, the IPv4 address option follows the pattern: 1806 ipv4-locator-option = [O_IPv4_LOCATOR, ipv4-address, 1807 transport-proto, port-number] 1808 ipv4-address = bytes .size 4 1810 The content of this option is a binary IPv4 address followed by the 1811 protocol number and port number to be used. 1813 Note: If an operator has internal network address translation for 1814 IPv4, this option MUST NOT be used within the Divert option. 1816 3.9.5.3. Locator FQDN option 1818 In fragmentary CDDL, the FQDN option follows the pattern: 1820 fqdn-locator-option = [O_FQDN_LOCATOR, text, 1821 transport-proto, port-number] 1823 The content of this option is the Fully Qualified Domain Name of the 1824 target followed by the protocol number and port number to be used. 1826 Note 1: Any FQDN which might not be valid throughout the network in 1827 question, such as a Multicast DNS name [RFC6762], MUST NOT be used 1828 when this option is used within the Divert option. 1830 Note 2: Normal GRASP operations are not expected to use this option. 1831 It is intended for special purposes such as discovering external 1832 services. 1834 3.9.5.4. Locator URI option 1836 In fragmentary CDDL, the URI option follows the pattern: 1838 uri-locator = [O_URI_LOCATOR, text] 1840 The content of this option is the Uniform Resource Identifier of the 1841 target [RFC3986]. 1843 Note 1: Any URI which might not be valid throughout the network in 1844 question, such as one based on a Multicast DNS name [RFC6762], MUST 1845 NOT be used when this option is used within the Divert option. 1847 Note 2: Normal GRASP operations are not expected to use this option. 1848 It is intended for special purposes such as discovering external 1849 services. 1851 3.10. Objective Options 1853 3.10.1. Format of Objective Options 1855 An objective option is used to identify objectives for the purposes 1856 of discovery, negotiation or synchronization. All objectives MUST be 1857 in the following format, described in fragmentary CDDL: 1859 objective = [objective-name, objective-flags, loop-count, ?any] 1861 objective-name = text 1862 loop-count = 0..255 1864 All objectives are identified by a unique name which is a UTF-8 1865 string, to be compared byte by byte. 1867 The names of generic objectives MUST NOT include a colon (":") and 1868 MUST be registered with IANA (Section 7). 1870 The names of privately defined objectives MUST include at least one 1871 colon (":"). The string preceding the last colon in the name MUST be 1872 globally unique and in some way identify the entity or person 1873 defining the objective. The following three methods MAY be used to 1874 create such a globally unique string: 1876 1. The unique string is a decimal number representing a registered 1877 32 bit Private Enterprise Number (PEN) [I-D.liang-iana-pen] that 1878 uniquely identifies the enterprise defining the objective. 1880 2. The unique string is a fully qualified domain name that uniquely 1881 identifies the entity or person defining the objective. 1883 3. The unique string is an email address that uniquely identifies 1884 the entity or person defining the objective. 1886 The GRASP protocol treats the objective name as an opaque string. 1887 For example, "EX1", "411:EX1", "example.com:EX1", "example.org:EX1 1888 and "user@example.org:EX1" would be five different objectives. 1890 The 'objective-flags' field is described below. 1892 The 'loop-count' field is used for terminating negotiation as 1893 described in Section 3.8.7. It is also used for terminating 1894 discovery as described in Section 3.5.4, and for terminating flooding 1895 as described in Section 3.5.6.2. It is placed in the objective 1896 rather than in the GRASP message format because, as far as the ASA is 1897 concerned, it is a property of the objective itself. 1899 The 'any' field is to express the actual value of a negotiation or 1900 synchronization objective. Its format is defined in the 1901 specification of the objective and may be a simple value or a data 1902 structure of any kind. It is optional because it is optional in a 1903 Discovery or Discovery Response message. 1905 3.10.2. Objective flags 1907 An objective may be relevant for discovery only, for discovery and 1908 negotiation, or for discovery and synchronization. This is expressed 1909 in the objective by logical flag bits: 1911 objective-flags = uint .bits objective-flag 1912 objective-flag = &( 1913 F_DISC: 0 ; valid for discovery 1914 F_NEG: 1 ; valid for negotiation 1915 F_SYNCH: 2 ; valid for synchronization 1916 F_NEG_DRY: 3 ; negotiation is dry-run 1917 ) 1919 These bits are independent and may be combined appropriately, e.g. 1920 (F_DISC and F_SYNCH) or (F_DISC and F_NEG) or (F_DISC and F_NEG and 1921 F_NEG_DRY). 1923 Note that for a given negotiation session, an objective must be 1924 either used for negotiation, or for dry-run negotiation. Mixing the 1925 two modes in a single negotiation is not possible. 1927 3.10.3. General Considerations for Objective Options 1929 As mentioned above, Objective Options MUST be assigned a unique name. 1930 As long as privately defined Objective Options obey the rules above, 1931 this document does not restrict their choice of name, but the entity 1932 or person concerned SHOULD publish the names in use. 1934 All Objective Options MUST respect the CBOR patterns defined above as 1935 "objective" and MUST replace the "any" field with a valid CBOR data 1936 definition for the relevant use case and application. 1938 An Objective Option that contains no additional fields beyond its 1939 "loop-count" can only be a discovery objective and MUST only be used 1940 in Discovery and Discovery Response messages. 1942 The Negotiation Objective Options contain negotiation objectives, 1943 which vary according to different functions/services. They MUST be 1944 carried by Discovery, Request Negotiation or Negotiation messages 1945 only. The negotiation initiator MUST set the initial "loop-count" to 1946 a value specified in the specification of the objective or, if no 1947 such value is specified, to GRASP_DEF_LOOPCT. 1949 For most scenarios, there should be initial values in the negotiation 1950 requests. Consequently, the Negotiation Objective options MUST 1951 always be completely presented in a Request Negotiation message, or 1952 in a Discovery message in rapid mode. If there is no initial value, 1953 the value field SHOULD be set to the 'null' value defined by CBOR. 1955 Synchronization Objective Options are similar, but MUST be carried by 1956 Discovery, Discovery Response, Request Synchronization, or Flood 1957 Synchronization messages only. They include value fields only in 1958 Synchronization or Flood Synchronization messages. 1960 3.10.4. Organizing of Objective Options 1962 Generic objective options MUST be specified in documents available to 1963 the public and SHOULD be designed to use either the negotiation or 1964 the synchronization mechanism described above. 1966 As noted earlier, one negotiation objective is handled by each GRASP 1967 negotiation thread. Therefore, a negotiation objective, which is 1968 based on a specific function or action, SHOULD be organized as a 1969 single GRASP option. It is NOT RECOMMENDED to organize multiple 1970 negotiation objectives into a single option, nor to split a single 1971 function or action into multiple negotiation objectives. 1973 It is important to understand that GRASP negotiation does not support 1974 transactional integrity. If transactional integrity is needed for a 1975 specific objective, this must be ensured by the ASA. For example, an 1976 ASA might need to ensure that it only participates in one negotiation 1977 thread at the same time. Such an ASA would need to stop listening 1978 for incoming negotiation requests before generating an outgoing 1979 negotiation request. 1981 A synchronization objective SHOULD be organized as a single GRASP 1982 option. 1984 Some objectives will support more than one operational mode. An 1985 example is a negotiation objective with both a "dry run" mode (where 1986 the negotiation is to find out whether the other end can in fact make 1987 the requested change without problems) and a "live" mode. Such modes 1988 will be defined in the specification of such an objective. These 1989 objectives SHOULD include flags indicating the applicable mode(s). 1991 An issue requiring particular attention is that GRASP itself is a 1992 stateless protocol. Any state associated with a dry run operation, 1993 such as temporarily reserving a resource for subsequent use in a live 1994 run, is entirely a matter for the designer of the ASA concerned. 1996 As indicated in Section 3.1, an objective's value may include 1997 multiple parameters. Parameters might be categorized into two 1998 classes: the obligatory ones presented as fixed fields; and the 1999 optional ones presented in some other form of data structure embedded 2000 in CBOR. The format might be inherited from an existing management 2001 or configuration protocol, with the objective option acting as a 2002 carrier for that format. The data structure might be defined in a 2003 formal language, but that is a matter for the specifications of 2004 individual objectives. There are many candidates, according to the 2005 context, such as ABNF, RBNF, XML Schema, YANG, etc. The GRASP 2006 protocol itself is agnostic on these questions. The only restriction 2007 is that the format can be mapped into CBOR. 2009 It is NOT RECOMMENDED to mix parameters that have significantly 2010 different response time characteristics in a single objective. 2011 Separate objectives are more suitable for such a scenario. 2013 All objectives MUST support GRASP discovery. However, as mentioned 2014 in Section 3.3, it is acceptable for an ASA to use an alternative 2015 method of discovery. 2017 Normally, a GRASP objective will refer to specific technical 2018 parameters as explained in Section 3.1. However, it is acceptable to 2019 define an abstract objective for the purpose of managing or 2020 coordinating ASAs. It is also acceptable to define a special-purpose 2021 objective for purposes such as trust bootstrapping or formation of 2022 the ACP. 2024 To guarantee convergence, a limited number of rounds or a timeout is 2025 needed for each negotiation objective. Therefore, the definition of 2026 each negotiation objective SHOULD clearly specify this, for example a 2027 default loop count and timeout, so that the negotiation can always be 2028 terminated properly. If not, the GRASP defaults will apply. 2030 There must be a well-defined procedure for concluding that a 2031 negotiation cannot succeed, and if so deciding what happens next 2032 (e.g., deadlock resolution, tie-breaking, or revert to best-effort 2033 service). This MUST be specified for individual negotiation 2034 objectives. 2036 3.10.5. Experimental and Example Objective Options 2038 The names "EX0" through "EX9" have been reserved for experimental 2039 options. Multiple names have been assigned because a single 2040 experiment may use multiple options simultaneously. These 2041 experimental options are highly likely to have different meanings 2042 when used for different experiments. Therefore, they SHOULD NOT be 2043 used without an explicit human decision and SHOULD NOT be used in 2044 unmanaged networks such as home networks. 2046 These names are also RECOMMENDED for use in documentation examples. 2048 4. Implementation Status [RFC Editor: please remove] 2050 Two prototype implementations of GRASP have been made. 2052 4.1. BUPT C++ Implementation 2054 o Name: BaseNegotiator.cpp, msg.cpp, Client.cpp, Server.cpp 2056 o Description: C++ implementation of GRASP core and API 2058 o Maturity: Prototype code, interoperable between Ubuntu. 2060 o Coverage: Corresponds to draft-carpenter-anima-gdn-protocol-03. 2061 Since it was implemented based on the old version draft, the most 2062 significant limitations comparing to current protocol design 2063 include: 2065 * Not support CBOR 2067 * Not support Flooding 2069 * Not support loop avoidance 2071 * only coded for IPv6, any IPv4 is accidental 2073 o Licensing: Huawei License. 2075 o Experience: https://github.com/liubingpang/IETF-Anima-Signaling- 2076 Protocol/blob/master/README.md 2078 o Contact: https://github.com/liubingpang/IETF-Anima-Signaling- 2079 Protocol 2081 4.2. Python Implementation 2083 o Name: graspy 2085 o Description: Python 3 implementation of GRASP core and API. 2087 o Maturity: Prototype code, interoperable between Windows 7 and 2088 Linux. 2090 o Coverage: Corresponds to draft-ietf-anima-grasp-10. Limitations 2091 include: 2093 * insecure: uses a dummy ACP module and does not implement TLS 2095 * only coded for IPv6, any IPv4 is accidental 2097 * FQDN and URI locators incompletely supported 2099 * no code for rapid mode 2101 * relay code is lazy (no rate control) 2103 * all unicast transactions use TCP (no unicast UDP). 2104 Experimental code for unicast UDP proved to be complex and 2105 brittle. 2107 * optional Objective option in Response messages not implemented 2109 * workarounds for defects in Python socket module and Windows 2110 socket peculiarities 2112 o Licensing: Simplified BSD 2114 o Experience: https://www.cs.auckland.ac.nz/~brian/graspy/graspy.pdf 2116 o Contact: https://www.cs.auckland.ac.nz/~brian/graspy/ 2118 5. Security Considerations 2120 A successful attack on negotiation-enabled nodes would be extremely 2121 harmful, as such nodes might end up with a completely undesirable 2122 configuration that would also adversely affect their peers. GRASP 2123 nodes and messages therefore require full protection. As explained 2124 in Section 3.5.1, GRASP MUST run within a secure environment such as 2125 the Autonomic Control Plane [I-D.ietf-anima-autonomic-control-plane], 2126 except for the constrained instances described in Section 3.5.2. 2128 - Authentication 2130 A cryptographically authenticated identity for each device is 2131 needed in an autonomic network. It is not safe to assume that a 2132 large network is physically secured against interference or that 2133 all personnel are trustworthy. Each autonomic node MUST be 2134 capable of proving its identity and authenticating its messages. 2135 GRASP relies on a separate external certificate-based security 2136 mechanism to support authentication, data integrity protection, 2137 and anti-replay protection. 2139 Since GRASP must be deployed in an existing secure environment, 2140 the protocol itself specifies nothing concerning the trust anchor 2141 and certification authority. 2143 If GRASP is used temporarily without an external security 2144 mechanism, for example during system bootstrap (Section 3.5.1), 2145 the Session ID (Section 3.7) will act as a nonce to provide 2146 limited protection against third parties injecting responses. A 2147 full analysis of the secure bootstrap process is in 2148 [I-D.ietf-anima-bootstrapping-keyinfra]. 2150 - Authorization and Roles 2152 The GRASP protocol is agnostic about the roles and capabilities of 2153 individual ASAs and about which objectives a particular ASA is 2154 authorized to support. An implementation might support 2155 precautions such as allowing only one ASA in a given node to 2156 modify a given objective, but this may not be appropriate in all 2157 cases. For example, it might be operationally useful to allow an 2158 old and a new version of the same ASA to run simultaneously during 2159 an overlap period. These questions are out of scope for the 2160 present specification. 2162 - Privacy and confidentiality 2164 Generally speaking, no personal information is expected to be 2165 involved in the signaling protocol, so there should be no direct 2166 impact on personal privacy. Nevertheless, traffic flow paths, 2167 VPNs, etc. could be negotiated, which could be of interest for 2168 traffic analysis. Also, operators generally want to conceal 2169 details of their network topology and traffic density from 2170 outsiders. Therefore, since insider attacks cannot be excluded in 2171 a large network, the security mechanism for the protocol MUST 2172 provide message confidentiality. This is why Section 3.5.1 2173 requires either an ACP or an alternative security mechanism. 2175 - Link-local multicast security 2177 GRASP has no reasonable alternative to using link-local multicast 2178 for Discovery or Flood Synchronization messages and these messages 2179 are sent in clear and with no authentication. They are therefore 2180 available to on-link eavesdroppers, and could be forged by on-link 2181 attackers. In the case of Discovery, the Discovery Responses are 2182 unicast and will therefore be protected (Section 3.5.1), and an 2183 untrusted forger will not be able to receive responses. In the 2184 case of Flood Synchronization, an on-link eavesdropper will be 2185 able to receive the flooded objectives but there is no response 2186 message to consider. Some precautions for Flood Synchronization 2187 messages are suggested in Section 3.5.6.2. 2189 - DoS Attack Protection 2191 GRASP discovery partly relies on insecure link-local multicast. 2192 Since routers participating in GRASP sometimes relay discovery 2193 messages from one link to another, this could be a vector for 2194 denial of service attacks. Some mitigations are specified in 2195 Section 3.5.4. However, malicious code installed inside the 2196 Autonomic Control Plane could always launch DoS attacks consisting 2197 of spurious discovery messages, or of spurious discovery 2198 responses. It is important that firewalls prevent any GRASP 2199 messages from entering the domain from an unknown source. 2201 - Security during bootstrap and discovery 2203 A node cannot trust GRASP traffic from other nodes until the 2204 security environment (such as the ACP) has identified the trust 2205 anchor and can authenticate traffic by validating certificates for 2206 other nodes. Also, until it has succesfully enrolled 2207 [I-D.ietf-anima-bootstrapping-keyinfra] a node cannot assume that 2208 other nodes are able to authenticate its own traffic. Therefore, 2209 GRASP discovery during the bootstrap phase for a new device will 2210 inevitably be insecure. Secure synchronization and negotiation 2211 will be impossible until enrollment is complete. Further details 2212 are given in Section 3.5.2. 2214 - Security of discovered locators 2216 When GRASP discovery returns an IP address, it MUST be that of a 2217 node within the secure environment (Section 3.5.1). If it returns 2218 an FQDN or a URI, the ASA that receives it MUST NOT assume that 2219 the target of the locator is within the secure environment. 2221 6. CDDL Specification of GRASP 2223 2224 grasp-message = (message .within message-structure) / noop-message 2226 message-structure = [MESSAGE_TYPE, session-id, ?initiator, 2227 *grasp-option] 2229 MESSAGE_TYPE = 0..255 2230 session-id = 0..4294967295 ;up to 32 bits 2231 grasp-option = any 2233 message /= discovery-message 2234 discovery-message = [M_DISCOVERY, session-id, initiator, objective] 2236 message /= response-message ;response to Discovery 2237 response-message = [M_RESPONSE, session-id, initiator, ttl, 2238 (+locator-option // divert-option), ?objective] 2240 message /= synch-message ;response to Synchronization request 2241 synch-message = [M_SYNCH, session-id, objective] 2243 message /= flood-message 2244 flood-message = [M_FLOOD, session-id, initiator, ttl, 2245 +[objective, (locator-option / [])]] 2247 message /= request-negotiation-message 2248 request-negotiation-message = [M_REQ_NEG, session-id, objective] 2250 message /= request-synchronization-message 2251 request-synchronization-message = [M_REQ_SYN, session-id, objective] 2253 message /= negotiation-message 2254 negotiation-message = [M_NEGOTIATE, session-id, objective] 2256 message /= end-message 2257 end-message = [M_END, session-id, accept-option / decline-option ] 2259 message /= wait-message 2260 wait-message = [M_WAIT, session-id, waiting-time] 2262 message /= invalid-message 2263 invalid-message = [M_INVALID, session-id, ?any] 2265 noop-message = [M_NOOP] 2267 divert-option = [O_DIVERT, +locator-option] 2269 accept-option = [O_ACCEPT] 2271 decline-option = [O_DECLINE, ?reason] 2272 reason = text ;optional error message 2274 waiting-time = 0..4294967295 ; in milliseconds 2275 ttl = 0..4294967295 ; in milliseconds 2277 locator-option /= [O_IPv4_LOCATOR, ipv4-address, 2278 transport-proto, port-number] 2279 ipv4-address = bytes .size 4 2281 locator-option /= [O_IPv6_LOCATOR, ipv6-address, 2282 transport-proto, port-number] 2283 ipv6-address = bytes .size 16 2285 locator-option /= [O_FQDN_LOCATOR, text, transport-proto, port-number] 2287 transport-proto = IPPROTO_TCP / IPPROTO_UDP 2288 IPPROTO_TCP = 6 2289 IPPROTO_UDP = 17 2290 port-number = 0..65535 2292 locator-option /= [O_URI_LOCATOR, text] 2294 initiator = ipv4-address / ipv6-address 2296 objective-flags = uint .bits objective-flag 2298 objective-flag = &( 2299 F_DISC: 0 ; valid for discovery 2300 F_NEG: 1 ; valid for negotiation 2301 F_SYNCH: 2 ; valid for synchronization 2302 F_NEG_DRY: 3 ; negotiation is dry-run 2303 ) 2305 objective = [objective-name, objective-flags, loop-count, ?any] 2307 objective-name = text ;see specification for uniqueness rules 2309 loop-count = 0..255 2311 ; Constants for message types and option types 2313 M_NOOP = 0 2314 M_DISCOVERY = 1 2315 M_RESPONSE = 2 2316 M_REQ_NEG = 3 2317 M_REQ_SYN = 4 2318 M_NEGOTIATE = 5 2319 M_END = 6 2320 M_WAIT = 7 2321 M_SYNCH = 8 2322 M_FLOOD = 9 2323 M_INVALID = 99 2325 O_DIVERT = 100 2326 O_ACCEPT = 101 2327 O_DECLINE = 102 2328 O_IPv6_LOCATOR = 103 2329 O_IPv4_LOCATOR = 104 2330 O_FQDN_LOCATOR = 105 2331 O_URI_LOCATOR = 106 2332 2334 7. IANA Considerations 2336 This document defines the GeneRic Autonomic Signaling Protocol 2337 (GRASP). 2339 Section 3.6 explains the following link-local multicast addresses, 2340 which IANA is requested to assign for use by GRASP: 2342 ALL_GRASP_NEIGHBORS multicast address (IPv6): (TBD1). Assigned in 2343 the IPv6 Link-Local Scope Multicast Addresses registry. 2345 ALL_GRASP_NEIGHBORS multicast address (IPv4): (TBD2). Assigned in 2346 the IPv4 Multicast Local Network Control Block. 2348 Section 3.6 explains the following User Port, which IANA is requested 2349 to assign for use by GRASP for both UDP and TCP: 2351 GRASP_LISTEN_PORT: (TBD3) 2352 Service Name: Generic Autonomic Signaling Protocol (GRASP) 2353 Transport Protocols: UDP, TCP 2354 Assignee: iesg@ietf.org 2355 Contact: chair@ietf.org 2356 Description: See Section 3.6 2357 Reference: RFC XXXX (this document) 2359 The IANA is requested to create a GRASP Parameter Registry including 2360 two registry tables. These are the GRASP Messages and Options 2361 Table and the GRASP Objective Names Table. 2363 GRASP Messages and Options Table. The values in this table are names 2364 paired with decimal integers. Future values MUST be assigned using 2365 the Standards Action policy defined by [RFC5226]. The following 2366 initial values are assigned by this document: 2368 M_NOOP = 0 2369 M_DISCOVERY = 1 2370 M_RESPONSE = 2 2371 M_REQ_NEG = 3 2372 M_REQ_SYN = 4 2373 M_NEGOTIATE = 5 2374 M_END = 6 2375 M_WAIT = 7 2376 M_SYNCH = 8 2377 M_FLOOD = 9 2378 M_INVALID = 99 2380 O_DIVERT = 100 2381 O_ACCEPT = 101 2382 O_DECLINE = 102 2383 O_IPv6_LOCATOR = 103 2384 O_IPv4_LOCATOR = 104 2385 O_FQDN_LOCATOR = 105 2386 O_URI_LOCATOR = 106 2388 GRASP Objective Names Table. The values in this table are UTF-8 2389 strings. Future values MUST be assigned using the Specification 2390 Required policy defined by [RFC5226]. 2392 To assist expert review of a new objective, the specification should 2393 include a precise description of the format of the new objective, 2394 with sufficient explanation of its semantics to allow independent 2395 implementations. See Section 3.10.3 for more details. If the new 2396 objective is similar in name or purpose to a previously registered 2397 objective, the specification should explain why a new objective is 2398 justified. 2400 The following initial values are assigned by this document: 2402 EX0 2403 EX1 2404 EX2 2405 EX3 2406 EX4 2407 EX5 2408 EX6 2409 EX7 2410 EX8 2411 EX9 2413 8. Acknowledgements 2415 A major contribution to the original version of this document was 2416 made by Sheng Jiang. Significant review inputs were received from 2417 Toerless Eckert, Joel Halpern, Barry Leiba, Charles E. Perkins, and 2418 Michael Richardson. 2420 Valuable comments were received from Michael Behringer, Jeferson 2421 Campos Nobre, Laurent Ciavaglia, Zongpeng Du, Yu Fu, Zhenbin Li, 2422 Dimitri Papadimitriou, Pierre Peloso, Reshad Rahman, Markus Stenberg, 2423 Rene Struik, Dacheng Zhang, and other participants in the NMRG 2424 research group and the ANIMA working group. 2426 9. References 2428 9.1. Normative References 2430 [I-D.greevenbosch-appsawg-cbor-cddl] 2431 Vigano, C. and H. Birkholz, "CBOR data definition language 2432 (CDDL): a notational convention to express CBOR data 2433 structures", draft-greevenbosch-appsawg-cbor-cddl-09 (work 2434 in progress), September 2016. 2436 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 2437 Requirement Levels", BCP 14, RFC 2119, 2438 DOI 10.17487/RFC2119, March 1997, 2439 . 2441 [RFC3986] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform 2442 Resource Identifier (URI): Generic Syntax", STD 66, 2443 RFC 3986, DOI 10.17487/RFC3986, January 2005, 2444 . 2446 [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker, 2447 "Randomness Requirements for Security", BCP 106, RFC 4086, 2448 DOI 10.17487/RFC4086, June 2005, 2449 . 2451 [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security 2452 (TLS) Protocol Version 1.2", RFC 5246, 2453 DOI 10.17487/RFC5246, August 2008, 2454 . 2456 [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S., 2457 Housley, R., and W. Polk, "Internet X.509 Public Key 2458 Infrastructure Certificate and Certificate Revocation List 2459 (CRL) Profile", RFC 5280, DOI 10.17487/RFC5280, May 2008, 2460 . 2462 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 2463 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 2464 January 2012, . 2466 [RFC7049] Bormann, C. and P. Hoffman, "Concise Binary Object 2467 Representation (CBOR)", RFC 7049, DOI 10.17487/RFC7049, 2468 October 2013, . 2470 [RFC7217] Gont, F., "A Method for Generating Semantically Opaque 2471 Interface Identifiers with IPv6 Stateless Address 2472 Autoconfiguration (SLAAC)", RFC 7217, 2473 DOI 10.17487/RFC7217, April 2014, 2474 . 2476 9.2. Informative References 2478 [I-D.chaparadza-intarea-igcp] 2479 Behringer, M., Chaparadza, R., Petre, R., Li, X., and H. 2480 Mahkonen, "IP based Generic Control Protocol (IGCP)", 2481 draft-chaparadza-intarea-igcp-00 (work in progress), July 2482 2011. 2484 [I-D.ietf-anima-autonomic-control-plane] 2485 Behringer, M., Eckert, T., and S. Bjarnason, "An Autonomic 2486 Control Plane", draft-ietf-anima-autonomic-control- 2487 plane-05 (work in progress), January 2017. 2489 [I-D.ietf-anima-bootstrapping-keyinfra] 2490 Pritikin, M., Richardson, M., Behringer, M., Bjarnason, 2491 S., and K. Watsen, "Bootstrapping Remote Secure Key 2492 Infrastructures (BRSKI)", draft-ietf-anima-bootstrapping- 2493 keyinfra-04 (work in progress), October 2016. 2495 [I-D.ietf-anima-reference-model] 2496 Behringer, M., Carpenter, B., Eckert, T., Ciavaglia, L., 2497 Pierre, P., Liu, B., Nobre, J., and J. Strassner, "A 2498 Reference Model for Autonomic Networking", draft-ietf- 2499 anima-reference-model-02 (work in progress), July 2016. 2501 [I-D.ietf-anima-stable-connectivity] 2502 Eckert, T. and M. Behringer, "Using Autonomic Control 2503 Plane for Stable Connectivity of Network OAM", draft-ietf- 2504 anima-stable-connectivity-02 (work in progress), February 2505 2017. 2507 [I-D.liang-iana-pen] 2508 Liang, P., Melnikov, A., and D. Conrad, "Private 2509 Enterprise Number (PEN) practices and Internet Assigned 2510 Numbers Authority (IANA) registration considerations", 2511 draft-liang-iana-pen-06 (work in progress), July 2015. 2513 [I-D.liu-anima-grasp-api] 2514 Carpenter, B., Liu, B., Wang, W., and X. Gong, "Generic 2515 Autonomic Signaling Protocol Application Program Interface 2516 (GRASP API)", draft-liu-anima-grasp-api-03 (work in 2517 progress), February 2017. 2519 [I-D.stenberg-anima-adncp] 2520 Stenberg, M., "Autonomic Distributed Node Consensus 2521 Protocol", draft-stenberg-anima-adncp-00 (work in 2522 progress), March 2015. 2524 [RFC2205] Braden, R., Ed., Zhang, L., Berson, S., Herzog, S., and S. 2525 Jamin, "Resource ReSerVation Protocol (RSVP) -- Version 1 2526 Functional Specification", RFC 2205, DOI 10.17487/RFC2205, 2527 September 1997, . 2529 [RFC2334] Luciani, J., Armitage, G., Halpern, J., and N. Doraswamy, 2530 "Server Cache Synchronization Protocol (SCSP)", RFC 2334, 2531 DOI 10.17487/RFC2334, April 1998, 2532 . 2534 [RFC2608] Guttman, E., Perkins, C., Veizades, J., and M. Day, 2535 "Service Location Protocol, Version 2", RFC 2608, 2536 DOI 10.17487/RFC2608, June 1999, 2537 . 2539 [RFC2865] Rigney, C., Willens, S., Rubens, A., and W. Simpson, 2540 "Remote Authentication Dial In User Service (RADIUS)", 2541 RFC 2865, DOI 10.17487/RFC2865, June 2000, 2542 . 2544 [RFC3209] Awduche, D., Berger, L., Gan, D., Li, T., Srinivasan, V., 2545 and G. Swallow, "RSVP-TE: Extensions to RSVP for LSP 2546 Tunnels", RFC 3209, DOI 10.17487/RFC3209, December 2001, 2547 . 2549 [RFC3315] Droms, R., Ed., Bound, J., Volz, B., Lemon, T., Perkins, 2550 C., and M. Carney, "Dynamic Host Configuration Protocol 2551 for IPv6 (DHCPv6)", RFC 3315, DOI 10.17487/RFC3315, July 2552 2003, . 2554 [RFC3416] Presuhn, R., Ed., "Version 2 of the Protocol Operations 2555 for the Simple Network Management Protocol (SNMP)", 2556 STD 62, RFC 3416, DOI 10.17487/RFC3416, December 2002, 2557 . 2559 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 2560 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 2561 DOI 10.17487/RFC4861, September 2007, 2562 . 2564 [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an 2565 IANA Considerations Section in RFCs", BCP 26, RFC 5226, 2566 DOI 10.17487/RFC5226, May 2008, 2567 . 2569 [RFC5971] Schulzrinne, H. and R. Hancock, "GIST: General Internet 2570 Signalling Transport", RFC 5971, DOI 10.17487/RFC5971, 2571 October 2010, . 2573 [RFC6206] Levis, P., Clausen, T., Hui, J., Gnawali, O., and J. Ko, 2574 "The Trickle Algorithm", RFC 6206, DOI 10.17487/RFC6206, 2575 March 2011, . 2577 [RFC6241] Enns, R., Ed., Bjorklund, M., Ed., Schoenwaelder, J., Ed., 2578 and A. Bierman, Ed., "Network Configuration Protocol 2579 (NETCONF)", RFC 6241, DOI 10.17487/RFC6241, June 2011, 2580 . 2582 [RFC6733] Fajardo, V., Ed., Arkko, J., Loughney, J., and G. Zorn, 2583 Ed., "Diameter Base Protocol", RFC 6733, 2584 DOI 10.17487/RFC6733, October 2012, 2585 . 2587 [RFC6762] Cheshire, S. and M. Krochmal, "Multicast DNS", RFC 6762, 2588 DOI 10.17487/RFC6762, February 2013, 2589 . 2591 [RFC6763] Cheshire, S. and M. Krochmal, "DNS-Based Service 2592 Discovery", RFC 6763, DOI 10.17487/RFC6763, February 2013, 2593 . 2595 [RFC6887] Wing, D., Ed., Cheshire, S., Boucadair, M., Penno, R., and 2596 P. Selkirk, "Port Control Protocol (PCP)", RFC 6887, 2597 DOI 10.17487/RFC6887, April 2013, 2598 . 2600 [RFC7558] Lynn, K., Cheshire, S., Blanchet, M., and D. Migault, 2601 "Requirements for Scalable DNS-Based Service Discovery 2602 (DNS-SD) / Multicast DNS (mDNS) Extensions", RFC 7558, 2603 DOI 10.17487/RFC7558, July 2015, 2604 . 2606 [RFC7575] Behringer, M., Pritikin, M., Bjarnason, S., Clemm, A., 2607 Carpenter, B., Jiang, S., and L. Ciavaglia, "Autonomic 2608 Networking: Definitions and Design Goals", RFC 7575, 2609 DOI 10.17487/RFC7575, June 2015, 2610 . 2612 [RFC7576] Jiang, S., Carpenter, B., and M. Behringer, "General Gap 2613 Analysis for Autonomic Networking", RFC 7576, 2614 DOI 10.17487/RFC7576, June 2015, 2615 . 2617 [RFC7787] Stenberg, M. and S. Barth, "Distributed Node Consensus 2618 Protocol", RFC 7787, DOI 10.17487/RFC7787, April 2016, 2619 . 2621 [RFC7788] Stenberg, M., Barth, S., and P. Pfister, "Home Networking 2622 Control Protocol", RFC 7788, DOI 10.17487/RFC7788, April 2623 2016, . 2625 [RFC8040] Bierman, A., Bjorklund, M., and K. Watsen, "RESTCONF 2626 Protocol", RFC 8040, DOI 10.17487/RFC8040, January 2017, 2627 . 2629 Appendix A. Open Issues [RFC Editor: This section should be empty. 2630 Please remove] 2632 o 63. Should encryption be MUST instead of SHOULD in Section 3.5.1 2633 and Section 3.5.2.1? 2635 o 64. Should more security text be moved from the main text into 2636 the Security Considerations? 2638 o 65. Do we need to formally restrict Unicode characters allowed in 2639 objective names? 2641 o 66. Split requirements into separate document? 2643 o 67. Remove normative dependency on draft-greevenbosch-appsawg- 2644 cbor-cddl? 2646 Appendix B. Closed Issues [RFC Editor: Please remove] 2648 o 1. UDP vs TCP: For now, this specification suggests UDP and TCP 2649 as message transport mechanisms. This is not clarified yet. UDP 2650 is good for short conversations, is necessary for multicast 2651 discovery, and generally fits the discovery and divert scenarios 2652 well. However, it will cause problems with large messages. TCP 2653 is good for stable and long sessions, with a little bit of time 2654 consumption during the session establishment stage. If messages 2655 exceed a reasonable MTU, a TCP mode will be required in any case. 2656 This question may be affected by the security discussion. 2658 RESOLVED by specifying UDP for short message and TCP for longer 2659 one. 2661 o 2. DTLS or TLS vs built-in security mechanism. For now, this 2662 specification has chosen a PKI based built-in security mechanism 2663 based on asymmetric cryptography. However, (D)TLS might be chosen 2664 as security solution to avoid duplication of effort. It also 2665 allows essentially similar security for short messages over UDP 2666 and longer ones over TCP. The implementation trade-offs are 2667 different. The current approach requires expensive asymmetric 2668 cryptographic calculations for every message. (D)TLS has startup 2669 overheads but cheaper crypto per message. DTLS is less mature 2670 than TLS. 2672 RESOLVED by specifying external security (ACP or (D)TLS). 2674 o The following open issues applied only if the original security 2675 model was retained: 2677 * 2.1. For replay protection, GRASP currently requires every 2678 participant to have an NTP-synchronized clock. Is this OK for 2679 low-end devices, and how does it work during device 2680 bootstrapping? We could take the Timestamp out of signature 2681 option, to become an independent and OPTIONAL (or RECOMMENDED) 2682 option. 2684 * 2.2. The Signature Option states that this option could be any 2685 place in a message. Wouldn't it be better to specify a 2686 position (such as the end)? That would be much simpler to 2687 implement. 2689 RESOLVED by changing security model. 2691 o 3. DoS Attack Protection needs work. 2693 RESOLVED by adding text. 2695 o 4. Should we consider preferring a text-based approach to 2696 discovery (after the initial discovery needed for bootstrapping)? 2697 This could be a complementary mechanism for multicast based 2698 discovery, especially for a very large autonomic network. 2699 Centralized registration could be automatically deployed 2700 incrementally. At the very first stage, the repository could be 2701 empty; then it could be filled in by the objectives discovered by 2702 different devices (for example using Dynamic DNS Update). The 2703 more records are stored in the repository, the less the multicast- 2704 based discovery is needed. However, if we adopt such a mechanism, 2705 there would be challenges: stateful solution, and security. 2707 RESOLVED for now by adding optional use of DNS-SD by ASAs. 2708 Subsequently removed by editors as irrelevant to GRASP istelf. 2710 o 5. Need to expand description of the minimum requirements for the 2711 specification of an individual discovery, synchronization or 2712 negotiation objective. 2714 RESOLVED for now by extra wording. 2716 o 6. Use case and protocol walkthrough. A description of how a 2717 node starts up, performs discovery, and conducts negotiation and 2718 synchronisation for a sample use case would help readers to 2719 understand the applicability of this specification. Maybe it 2720 should be an artificial use case or maybe a simple real one, based 2721 on a conceptual API. However, the authors have not yet decided 2722 whether to have a separate document or have it in the protocol 2723 document. 2725 RESOLVED: recommend a separate document. 2727 o 7. Cross-check against other ANIMA WG documents for consistency 2728 and gaps. 2730 RESOLVED: Satisfied by WGLC. 2732 o 8. Consideration of ADNCP proposal. 2734 RESOLVED by adding optional use of DNCP for flooding-type 2735 synchronization. 2737 o 9. Clarify how a GDNP instance knows whether it is running inside 2738 the ACP. (Sheng) 2740 RESOLVED by improved text. 2742 o 10. Clarify how a non-ACP GDNP instance initiates (D)TLS. 2743 (Sheng) 2745 RESOLVED by improved text and declaring DTLS out of scope for this 2746 draft. 2748 o 11. Clarify how UDP/TCP choice is made. (Sheng) [Like DNS? - 2749 Brian] 2751 RESOLVED by improved text. 2753 o 12. Justify that IP address within ACP or (D)TLS environment is 2754 sufficient to prove AN identity; or explain how Device Identity 2755 Option is used. (Sheng) 2757 RESOLVED for now: we assume that all ASAs in a device are trusted 2758 as soon as the device is trusted, so they share credentials. In 2759 that case the Device Identity Option is useless. This needs to be 2760 reviewed later. 2762 o 13. Emphasise that negotiation/synchronization are independent 2763 from discovery, although the rapid discovery mode includes the 2764 first step of a negotiation/synchronization. (Sheng) 2766 RESOLVED by improved text. 2768 o 14. Do we need an unsolicited flooding mechanism for discovery 2769 (for discovery results that everyone needs), to reduce scaling 2770 impact of flooding discovery messages? (Toerless) 2772 RESOLVED: Yes, added to requirements and solution. 2774 o 15. Do we need flag bits in Objective Options to distinguish 2775 distinguish Synchronization and Negotiation "Request" or rapid 2776 mode "Discovery" messages? (Bing) 2778 RESOLVED: yes, work on the API showed that these flags are 2779 essential. 2781 o 16. (Related to issue 14). Should we revive the "unsolicited 2782 Response" for flooding synchronisation data? This has to be done 2783 carefully due to the well-known issues with flooding, but it could 2784 be useful, e.g. for Intent distribution, where DNCP doesn't seem 2785 applicable. 2787 RESOLVED: Yes, see #14. 2789 o 17. Ensure that the discovery mechanism is completely proof 2790 against loops and protected against duplicate responses. 2792 RESOLVED: Added loop count mechanism. 2794 o 18. Discuss the handling of multiple valid discovery responses. 2796 RESOLVED: Stated that the choice must be available to the ASA but 2797 GRASP implementation should pick a default. 2799 o 19. Should we use a text-oriented format such as JSON/CBOR 2800 instead of native binary TLV format? 2802 RESOLVED: Yes, changed to CBOR. 2804 o 20. Is the Divert option needed? If a discovery response 2805 provides a valid IP address or FQDN, the recipient doesn't gain 2806 any extra knowledge from the Divert. On the other hand, the 2807 presence of Divert informs the receiver that the target is off- 2808 link, which might be useful sometimes. 2810 RESOLVED: Decided to keep Divert option. 2812 o 21. Rename the protocol as GRASP (GeneRic Autonomic Signaling 2813 Protocol)? 2815 RESOLVED: Yes, name changed. 2817 o 22. Does discovery mechanism scale robustly as needed? Need hop 2818 limit on relaying? 2820 RESOLVED: Added hop limit. 2822 o 23. Need more details on TTL for caching discovery responses. 2824 RESOLVED: Done. 2826 o 24. Do we need "fast withdrawal" of discovery responses? 2828 RESOLVED: This doesn't seem necessary. If an ASA exits or stops 2829 supporting a given objective, peers will fail to start future 2830 sessions and will simply repeat discovery. 2832 o 25. Does GDNP discovery meet the needs of multi-hop DNS-SD? 2834 RESOLVED: Decided not to consider this further as a GRASP protocol 2835 issue. GRASP objectives could embed DNS-SD formats if needed. 2837 o 26. Add a URL type to the locator options (for security bootstrap 2838 etc.) 2840 RESOLVED: Done, later renamed as URI. 2842 o 27. Security of Flood multicasts (Section 3.5.6.2). 2844 RESOLVED: added text. 2846 o 28. Does ACP support secure link-local multicast? 2848 RESOLVED by new text in the Security Considerations. 2850 o 29. PEN is used to distinguish vendor options. Would it be 2851 better to use a domain name? Anything unique will do. 2853 RESOLVED: Simplified this by removing PEN field and changing 2854 naming rules for objectives. 2856 o 30. Does response to discovery require randomized delays to 2857 mitigate amplification attacks? 2859 RESOLVED: WG feedback is that it's unnecessary. 2861 o 31. We have specified repeats for failed discovery etc. Is that 2862 sufficient to deal with sleeping nodes? 2864 RESOLVED: WG feedback is that it's unnecessary to say more. 2866 o 32. We have one-to-one synchronization and flooding 2867 synchronization. Do we also need selective flooding to a subset 2868 of nodes? 2870 RESOLVED: This will be discussed as a protocol extension in a 2871 separate draft (draft-liu-anima-grasp-distribution). 2873 o 33. Clarify if/when discovery needs to be repeated. 2875 RESOLVED: Done. 2877 o 34. Clarify what is mandatory for running in ACP, expand 2878 discussion of security boundary when running with no ACP - might 2879 rely on the local PKI infrastructure. 2881 RESOLVED: Done. 2883 o 35. State that role-based authorization of ASAs is out of scope 2884 for GRASP. GRASP doesn't recognize/handle any "roles". 2886 RESOLVED: Done. 2888 o 36. Reconsider CBOR definition for PEN syntax. ( objective-name 2889 = text / [pen, text] ; pen = uint ) 2891 RESOLVED: See issue 29. 2893 o 37. Are URI locators really needed? 2895 RESOLVED: Yes, e.g. for security bootstrap discovery, but added 2896 note that addresses are the normal case (same for FQDN locators). 2898 o 38. Is Session ID sufficient to identify relayed responses? 2899 Isn't the originator's address needed too? 2901 RESOLVED: Yes, this is needed for multicast messages and their 2902 responses. 2904 o 39. Clarify that a node will contain one GRASP instance 2905 supporting multiple ASAs. 2907 RESOLVED: Done. 2909 o 40. Add a "reason" code to the DECLINE option? 2911 RESOLVED: Done. 2913 o 41. What happens if an ASA cannot conveniently use one of the 2914 GRASP mechanisms? Do we (a) add a message type to GRASP, or (b) 2915 simply pass the discovery results to the ASA so that it can open 2916 its own socket? 2918 RESOLVED: Both would be possible, but (b) is preferred. 2920 o 42. Do we need a feature whereby an ASA can bypass the ACP and 2921 use the data plane for efficiency/throughput? This would require 2922 discovery to return non-ACP addresses and would evade ACP 2923 security. 2925 RESOLVED: This is considered out of scope for GRASP, but a comment 2926 has been added in security considerations. 2928 o 43. Rapid mode synchronization and negotiation is currently 2929 limited to a single objective for simplicity of design and 2930 implementation. A future consideration is to allow multiple 2931 objectives in rapid mode for greater efficiency. 2933 RESOLVED: This is considered out of scope for this version. 2935 o 44. In requirement T9, the words that encryption "may not be 2936 required in all deployments" were removed. Is that OK?. 2938 RESOLVED: No objections. 2940 o 45. Device Identity Option is unused. Can we remove it 2941 completely?. 2943 RESOLVED: No objections. Done. 2945 o 46. The 'initiator' field in DISCOVER, RESPONSE and FLOOD 2946 messages is intended to assist in loop prevention. However, we 2947 also have the loop count for that. Also, if we create a new 2948 Session ID each time a DISCOVER or FLOOD is relayed, that ID can 2949 be disambiguated by recipients. It would be simpler to remove the 2950 initiator from the messages, making parsing more uniform. Is that 2951 OK? 2953 RESOLVED: Yes. Done. 2955 o 47. REQUEST is a dual purpose message (request negotiation or 2956 request synchronization). Would it be better to split this into 2957 two different messages (and adjust various message names 2958 accordingly)? 2960 RESOLVED: Yes. Done. 2962 o 48. Should the Appendix "Capability Analysis of Current 2963 Protocols" be deleted before RFC publication? 2965 RESOLVED: No (per WG meeting at IETF 96). 2967 o 49. Section 3.5.1 Should say more about signaling between two 2968 autonomic networks/domains. 2970 RESOLVED: Description of separate GRASP instance added. 2972 o 50. Is Rapid mode limited to on-link only? What happens if first 2973 discovery responder does not support Rapid Mode? Section 3.5.5, 2974 Section 3.5.6) 2976 RESOLVED: Not limited to on-link. First responder wins. 2978 o 51. Should flooded objectives have a time-to-live before they are 2979 deleted from the flood cache? And should they be tagged in the 2980 cache with their source locator? 2981 RESOLVED: TTL added to Flood (and Discovery Response) messages. 2982 Cached flooded objectives must be tagged with their originating 2983 ASA locator, and multiple copies must be kept if necessary. 2985 o 52. Describe in detail what is allowed and disallowed in an 2986 insecure instance of GRASP. 2988 RESOLVED: Done. 2990 o 53. Tune IANA Considerations to support early assignment request. 2992 o 54. Is there a highly unlikely race condition if two peers 2993 simultaneously choose the same Session ID and send each other 2994 simultaneous M_REQ_NEG messages? 2996 RESOLVED: Yes. Enhanced text on Session ID generation, and added 2997 precaution when receiving a Request message. 2999 o 55. Could discovery be performed over TCP? 3001 RESOLVED: Unicast discovery added as an option. 3003 o 56. Change Session-ID to 32 bits? 3005 RESOLVED: Done. 3007 o 57. Add M_INVALID message? 3009 RESOLVED: Done. 3011 o 58. Maximum message size? 3013 RESOLVED by specifying default maximum message size (2048 bytes). 3015 o 59. Add F_NEG_DRY flag to specify a "dry run" objective?. 3017 RESOLVED: Done. 3019 o 60. Change M_FLOOD syntax to associate a locator with each 3020 objective? 3022 RESOLVED: Done. 3024 o 61. Is the SONN constrained instance really needed? 3026 RESOLVED: Retained but only as an option. 3028 o 62. Is it helpful to tag descriptive text with message names 3029 (M_DISCOVER etc.)? 3031 RESOLVED: Yes, done in various parts of the text. 3033 Appendix C. Change log [RFC Editor: Please remove] 3035 draft-ietf-anima-grasp-10, 2017-03-10: 3037 Updates following IETF Last call: 3039 Protocol change: Specify that an objective with no initial value 3040 should have its value field set to CBOR 'null'. 3042 Protocol change: Specify behavior on receiving unrecognized message 3043 type. 3045 Noted that UTF-8 names are matched byte-for-byte. 3047 Added brief guidance for Expert Reviewer of new generic objectives. 3049 Numerous editorial improvements and clarifications and minor text 3050 rearrangements, none intended to change the meaning. 3052 draft-ietf-anima-grasp-09, 2016-12-15: 3054 Protocol change: Add F_NEG_DRY flag to specify a "dry run" objective. 3056 Protocol change: Change M_FLOOD syntax to associate a locator with 3057 each objective. 3059 Concentrated mentions of TLS in one section, with all details out of 3060 scope. 3062 Clarified text around constrained instances of GRASP. 3064 Strengthened text restricting LL addresses in locator options. 3066 Clarified description of rapid mode processsing. 3068 Specified that cached discovery results should not be returned on the 3069 same interface where they were learned. 3071 Shortened text in "High Level Design Choices" 3073 Dropped the word 'kernel' to avoid confusion with o/s kernel mode. 3075 Editorial improvements and clarifications. 3077 draft-ietf-anima-grasp-08, 2016-10-30: 3079 Protocol change: Added M_INVALID message. 3081 Protocol change: Increased Session ID space to 32 bits. 3083 Enhanced rules to avoid Session ID clashes. 3085 Corrected and completed description of timeouts for Request messages. 3087 Improved wording about exponential backoff and DoS. 3089 Clarified that discovery relaying is not done by limited security 3090 instances. 3092 Corrected and expanded explanation of port used for Discovery 3093 Response. 3095 Noted that Discovery message could be sent unicast in special cases. 3097 Added paragraph on extensibility. 3099 Specified default maximum message size. 3101 Added Appendix for sample messages. 3103 Added short protocol overview. 3105 Editorial fixes, including minor re-ordering for readability. 3107 draft-ietf-anima-grasp-07, 2016-09-13: 3109 Protocol change: Added TTL field to Flood message (issue 51). 3111 Protocol change: Added Locator option to Flood message (issue 51). 3113 Protocol change: Added TTL field to Discovery Response message 3114 (corrollary to issue 51). 3116 Clarified details of rapid mode (issues 43 and 50). 3118 Description of inter-domain GRASP instance added (issue 49). 3120 Description of limited security GRASP instances added (issue 52). 3122 Strengthened advice to use TCP rather than UDP. 3124 Updated IANA considerations and text about well-known port usage 3125 (issue 53). 3127 Amended text about ASA authorization and roles to allow for 3128 overlapping ASAs. 3130 Added text recommending that Flood should be repeated periodically. 3132 Editorial fixes. 3134 draft-ietf-anima-grasp-06, 2016-06-27: 3136 Added text on discovery cache timeouts. 3138 Noted that ASAs that are only initiators do not need to respond to 3139 discovery message. 3141 Added text on unexpected address changes. 3143 Added text on robust implementation. 3145 Clarifications and editorial fixes for numerous review comments 3147 Added open issues for some review comments. 3149 draft-ietf-anima-grasp-05, 2016-05-13: 3151 Noted in requirement T1 that it should be possible to implement ASAs 3152 independently as user space programs. 3154 Protocol change: Added protocol number and port to discovery 3155 response. Updated protocol description, CDDL and IANA considerations 3156 accordingly. 3158 Clarified that discovery and flood multicasts are handled by the 3159 GRASP core, not directly by ASAs. 3161 Clarified that a node may discover an objective without supporting it 3162 for synchronization or negotiation. 3164 Added Implementation Status section. 3166 Added reference to SCSP. 3168 Editorial fixes. 3170 draft-ietf-anima-grasp-04, 2016-03-11: 3172 Protocol change: Restored initiator field in certain messages and 3173 adjusted relaying rules to provide complete loop detection. 3175 Updated IANA Considerations. 3177 draft-ietf-anima-grasp-03, 2016-02-24: 3179 Protocol change: Removed initiator field from certain messages and 3180 adjusted relaying requirement to simplify loop detection. Also 3181 clarified narrative explanation of discovery relaying. 3183 Protocol change: Split Request message into two (Request Negotiation 3184 and Request Synchronization) and updated other message names for 3185 clarity. 3187 Protocol change: Dropped unused Device ID option. 3189 Further clarified text on transport layer usage. 3191 New text about multicast insecurity in Security Considerations. 3193 Various other clarifications and editorial fixes, including moving 3194 some material to Appendix. 3196 draft-ietf-anima-grasp-02, 2016-01-13: 3198 Resolved numerous issues according to WG discussions. 3200 Renumbered requirements, added D9. 3202 Protocol change: only allow one objective in rapid mode. 3204 Protocol change: added optional error string to DECLINE option. 3206 Protocol change: removed statement that seemed to say that a Request 3207 not preceded by a Discovery should cause a Discovery response. That 3208 made no sense, because there is no way the initiator would know where 3209 to send the Request. 3211 Protocol change: Removed PEN option from vendor objectives, changed 3212 naming rule accordingly. 3214 Protocol change: Added FLOOD message to simplify coding. 3216 Protocol change: Added SYNCH message to simplify coding. 3218 Protocol change: Added initiator id to DISCOVER, RESPONSE and FLOOD 3219 messages. But also allowed the relay process for DISCOVER and FLOOD 3220 to regenerate a Session ID. 3222 Protocol change: Require that discovered addresses must be global 3223 (except during bootstrap). 3225 Protocol change: Receiver of REQUEST message must close socket if no 3226 ASA is listening for the objective. 3228 Protocol change: Simplified Waiting message. 3230 Protocol change: Added No Operation message. 3232 Renamed URL locator type as URI locator type. 3234 Updated CDDL definition. 3236 Various other clarifications and editorial fixes. 3238 draft-ietf-anima-grasp-01, 2015-10-09: 3240 Updated requirements after list discussion. 3242 Changed from TLV to CBOR format - many detailed changes, added co- 3243 author. 3245 Tightened up loop count and timeouts for various cases. 3247 Noted that GRASP does not provide transactional integrity. 3249 Various other clarifications and editorial fixes. 3251 draft-ietf-anima-grasp-00, 2015-08-14: 3253 File name and protocol name changed following WG adoption. 3255 Added URL locator type. 3257 draft-carpenter-anima-gdn-protocol-04, 2015-06-21: 3259 Tuned wording around hierarchical structure. 3261 Changed "device" to "ASA" in many places. 3263 Reformulated requirements to be clear that the ASA is the main 3264 customer for signaling. 3266 Added requirement for flooding unsolicited synch, and added it to 3267 protocol spec. Recognized DNCP as alternative for flooding synch 3268 data. 3270 Requirements clarified, expanded and rearranged following design team 3271 discussion. 3273 Clarified that GDNP discovery must not be a prerequisite for GDNP 3274 negotiation or synchronization (resolved issue 13). 3276 Specified flag bits for objective options (resolved issue 15). 3278 Clarified usage of ACP vs TLS/DTLS and TCP vs UDP (resolved issues 3279 9,10,11). 3281 Updated DNCP description from latest DNCP draft. 3283 Editorial improvements. 3285 draft-carpenter-anima-gdn-protocol-03, 2015-04-20: 3287 Removed intrinsic security, required external security 3289 Format changes to allow DNCP co-existence 3291 Recognized DNS-SD as alternative discovery method. 3293 Editorial improvements 3295 draft-carpenter-anima-gdn-protocol-02, 2015-02-19: 3297 Tuned requirements to clarify scope, 3299 Clarified relationship between types of objective, 3301 Clarified that objectives may be simple values or complex data 3302 structures, 3304 Improved description of objective options, 3306 Added loop-avoidance mechanisms (loop count and default timeout, 3307 limitations on discovery relaying and on unsolicited responses), 3309 Allow multiple discovery objectives in one response, 3311 Provided for missing or multiple discovery responses, 3313 Indicated how modes such as "dry run" should be supported, 3314 Minor editorial and technical corrections and clarifications, 3316 Reorganized future work list. 3318 draft-carpenter-anima-gdn-protocol-01, restructured the logical flow 3319 of the document, updated to describe synchronization completely, add 3320 unsolicited responses, numerous corrections and clarifications, 3321 expanded future work list, 2015-01-06. 3323 draft-carpenter-anima-gdn-protocol-00, combination of draft-jiang- 3324 config-negotiation-ps-03 and draft-jiang-config-negotiation-protocol- 3325 02, 2014-10-08. 3327 Appendix D. Example Message Formats 3329 For readers unfamiliar with CBOR, this appendix shows a number of 3330 example GRASP messages conforming to the CDDL syntax given in 3331 Section 6. Each message is shown three times in the following 3332 formats: 3334 1. CBOR diagnostic notation. 3336 2. Similar, but showing the names of the constants. (Details of the 3337 flag bit encoding are omitted.) 3339 3. Hexadecimal version of the CBOR wire format. 3341 Long lines are split for display purposes only. 3343 D.1. Discovery Example 3345 The initiator (2001:db8:f000:baaa:28cc:dc4c:9703:6781) multicasts a 3346 discovery message looking for objective EX1: 3348 [1, 13948744, h'20010db8f000baaa28ccdc4c97036781', ["EX1", 5, 2, 0]] 3349 [M_DISCOVERY, 13948744, h'20010db8f000baaa28ccdc4c97036781', 3350 ["EX1", F_SYNCH_bits, 2, 0]] 3351 h'84011a00d4d7485020010db8f000baaa28ccdc4c970367818463455831050200' 3353 A peer (2001:0db8:f000:baaa:f000:baaa:f000:baaa) responds with a 3354 locator: 3356 [2, 13948744, h'20010db8f000baaa28ccdc4c97036781', 60000, 3357 [103, h'20010db8f000baaaf000baaaf000baaa', 6, 49443]] 3358 [M_RESPONSE, 13948744, h'20010db8f000baaa28ccdc4c97036781', 60000, 3359 [O_IPv6_LOCATOR, h'20010db8f000baaaf000baaaf000baaa', 3360 IPPROTO_TCP, 49443]] 3361 h'85021a00d4d7485020010db8f000baaa28ccdc4c9703678119ea6084186750 3362 20010db8f000baaaf000baaaf000baaa0619c123' 3364 D.2. Flood Example 3366 The initiator multicasts a flood message. The single objective has a 3367 null locator. There is no response: 3369 [9, 3504974, h'20010db8f000baaa28ccdc4c97036781', 10000, 3370 [["EX1", 5, 2, ["Example 1 value=", 100]],[] ] ] 3371 [M_FLOOD, 3504974, h'20010db8f000baaa28ccdc4c97036781', 10000, 3372 [["EX1", F_SYNCH_bits, 2, ["Example 1 value=", 100]],[] ] ] 3373 h'86091a00357b4e5020010db8f000baaa28ccdc4c97036781192710 3374 828463455831050282704578616d706c6520312076616c75653d186480' 3376 D.3. Synchronization Example 3378 Following successful discovery of objective EX2, the initiator 3379 unicasts a request: 3381 [4, 4038926, ["EX2", 5, 5, 0]] 3382 [M_REQ_SYN, 4038926, ["EX2", F_SYNCH_bits, 5, 0]] 3383 h'83041a003da10e8463455832050500' 3385 The peer responds with a value: 3387 [8, 4038926, ["EX2", 5, 5, ["Example 2 value=", 200]]] 3388 [M_SYNCH, 4038926, ["EX2", F_SYNCH_bits, 5, ["Example 2 value=", 200]]] 3389 h'83081a003da10e8463455832050582704578616d706c6520322076616c75653d18c8' 3391 D.4. Simple Negotiation Example 3393 Following successful discovery of objective EX3, the initiator 3394 unicasts a request: 3396 [3, 802813, ["EX3", 3, 6, ["NZD", 47]]] 3397 [M_REQ_NEG, 802813, ["EX3", F_NEG_bits, 6, ["NZD", 47]]] 3398 h'83031a000c3ffd8463455833030682634e5a44182f' 3400 The peer responds with immediate acceptance. Note that no objective 3401 is needed, because the initiator's request was accepted without 3402 change: 3404 [6, 802813, [101]] 3405 [M_END , 802813, [O_ACCEPT]] 3406 h'83061a000c3ffd811865' 3408 D.5. Complete Negotiation Example 3410 Again the initiator unicasts a request: 3412 [3, 13767778, ["EX3", 3, 6, ["NZD", 410]]] 3413 [M_REQ_NEG, 13767778, ["EX3", F_NEG_bits, 6, ["NZD", 410]]] 3414 h'83031a00d214628463455833030682634e5a4419019a' 3416 The responder starts to negotiate (making an offer): 3418 [5, 13767778, ["EX3", 3, 6, ["NZD", 80]]] 3419 [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 6, ["NZD", 80]]] 3420 h'83051a00d214628463455833030682634e5a441850' 3422 The initiator continues to negotiate (reducing its request, and note 3423 that the loop count is decremented): 3425 [5, 13767778, ["EX3", 3, 5, ["NZD", 307]]] 3426 [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 5, ["NZD", 307]]] 3427 h'83051a00d214628463455833030582634e5a44190133' 3429 The responder asks for more time: 3431 [7, 13767778, 34965] 3432 [M_WAIT, 13767778, 34965] 3433 h'83071a00d21462198895' 3435 The responder continues to negotiate (increasing its offer): 3437 [5, 13767778, ["EX3", 3, 4, ["NZD", 120]]] 3438 [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 4, ["NZD", 120]]] 3439 h'83051a00d214628463455833030482634e5a441878' 3441 The initiator continues to negotiate (reducing its request): 3443 [5, 13767778, ["EX3", 3, 3, ["NZD", 246]]] 3444 [M_NEGOTIATE, 13767778, ["EX3", F_NEG_bits, 3, ["NZD", 246]]] 3445 h'83051a00d214628463455833030382634e5a4418f6' 3447 The responder refuses to negotiate further: 3449 [6, 13767778, [102, "Insufficient funds"]] 3450 [M_END , 13767778, [O_DECLINE, "Insufficient funds"]] 3451 h'83061a00d2146282186672496e73756666696369656e742066756e6473' 3452 This negotiation has failed. If either side had sent [M_END, 3453 13767778, [O_ACCEPT]] it would have succeeded, converging on the 3454 objective value in the preceding M_NEGOTIATE. Note that apart from 3455 the initial M_REQ_NEG, the process is symmetrical. 3457 Appendix E. Capability Analysis of Current Protocols 3459 This appendix discusses various existing protocols with properties 3460 related to the requirements described in Section 2. The purpose is 3461 to evaluate whether any existing protocol, or a simple combination of 3462 existing protocols, can meet those requirements. 3464 Numerous protocols include some form of discovery, but these all 3465 appear to be very specific in their applicability. Service Location 3466 Protocol (SLP) [RFC2608] provides service discovery for managed 3467 networks, but requires configuration of its own servers. DNS-SD 3468 [RFC6763] combined with mDNS [RFC6762] provides service discovery for 3469 small networks with a single link layer. [RFC7558] aims to extend 3470 this to larger autonomous networks but this is not yet standardized. 3471 However, both SLP and DNS-SD appear to target primarily application 3472 layer services, not the layer 2 and 3 objectives relevant to basic 3473 network configuration. Both SLP and DNS-SD are text-based protocols. 3475 Routing protocols are mainly one-way information announcements. The 3476 receiver makes independent decisions based on the received 3477 information and there is no direct feedback information to the 3478 announcing peer. This remains true even though the protocol is used 3479 in both directions between peer routers; there is state 3480 synchronization, but no negotiation, and each peer runs its route 3481 calculations independently. 3483 Simple Network Management Protocol (SNMP) [RFC3416] uses a command/ 3484 response model not well suited for peer negotiation. Network 3485 Configuration Protocol (NETCONF) [RFC6241] uses an RPC model that 3486 does allow positive or negative responses from the target system, but 3487 this is still not adequate for negotiation. 3489 There are various existing protocols that have elementary negotiation 3490 abilities, such as Dynamic Host Configuration Protocol for IPv6 3491 (DHCPv6) [RFC3315], Neighbor Discovery (ND) [RFC4861], Port Control 3492 Protocol (PCP) [RFC6887], Remote Authentication Dial In User Service 3493 (RADIUS) [RFC2865], Diameter [RFC6733], etc. Most of them are 3494 configuration or management protocols. However, they either provide 3495 only a simple request/response model in a master/slave context or 3496 very limited negotiation abilities. 3498 There are some signaling protocols with an element of negotiation. 3499 For example Resource ReSerVation Protocol (RSVP) [RFC2205] was 3500 designed for negotiating quality of service parameters along the path 3501 of a unicast or multicast flow. RSVP is a very specialised protocol 3502 aimed at end-to-end flows. However, it has some flexibility, having 3503 been extended for MPLS label distribution [RFC3209]. A more generic 3504 design is General Internet Signalling Transport (GIST) [RFC5971], but 3505 it is complex, tries to solve many problems, and is also aimed at 3506 per-flow signaling across many hops rather than at device-to-device 3507 signaling. However, we cannot completely exclude extended RSVP or 3508 GIST as a synchronization and negotiation protocol. They do not 3509 appear to be directly useable for peer discovery. 3511 RESTCONF [RFC8040] is a protocol intended to convey NETCONF 3512 information expressed in the YANG language via HTTP, including the 3513 ability to transit HTML intermediaries. While this is a powerful 3514 approach in the context of centralised configuration of a complex 3515 network, it is not well adapted to efficient interactive negotiation 3516 between peer devices, especially simple ones that might not include 3517 YANG processing already. 3519 The Distributed Node Consensus Protocol (DNCP) [RFC7787] is defined 3520 as a generic form of state synchronization protocol, with a proposed 3521 usage profile being the Home Networking Control Protocol (HNCP) 3522 [RFC7788] for configuring Homenet routers. A specific application of 3523 DNCP for autonomic networking was proposed in 3524 [I-D.stenberg-anima-adncp]. 3526 DNCP "is designed to provide a way for each participating node to 3527 publish a set of TLV (Type-Length-Value) tuples, and to provide a 3528 shared and common view about the data published... DNCP is most 3529 suitable for data that changes only infrequently... If constant rapid 3530 state changes are needed, the preferable choice is to use an 3531 additional point-to-point channel..." 3533 Specific features of DNCP include: 3535 o Every participating node has a unique node identifier. 3537 o DNCP messages are encoded as a sequence of TLV objects, sent over 3538 unicast UDP or TCP, with or without (D)TLS security. 3540 o Multicast is used only for discovery of DNCP neighbors when lower 3541 security is acceptable. 3543 o Synchronization of state is maintained by a flooding process using 3544 the Trickle algorithm. There is no bilateral synchronization or 3545 negotiation capability. 3547 o The HNCP profile of DNCP is designed to operate between directly 3548 connected neighbors on a shared link using UDP and link-local IPv6 3549 addresses. 3551 DNCP does not meet the needs of a general negotiation protocol, 3552 because it is designed specifically for flooding synchronization. 3553 Also, in its HNCP profile it is limited to link-local messages and to 3554 IPv6. However, at the minimum it is a very interesting test case for 3555 this style of interaction between devices without needing a central 3556 authority, and it is a proven method of network-wide state 3557 synchronization by flooding. 3559 The Server Cache Synchronization Protocol (SCSP) [RFC2334] also 3560 describes a method for cache synchronization and cache replication 3561 among a group of nodes. 3563 A proposal was made some years ago for an IP based Generic Control 3564 Protocol (IGCP) [I-D.chaparadza-intarea-igcp]. This was aimed at 3565 information exchange and negotiation but not directly at peer 3566 discovery. However, it has many points in common with the present 3567 work. 3569 None of the above solutions appears to completely meet the needs of 3570 generic discovery, state synchronization and negotiation in a single 3571 solution. Many of the protocols assume that they are working in a 3572 traditional top-down or north-south scenario, rather than a fluid 3573 peer-to-peer scenario. Most of them are specialized in one way or 3574 another. As a result, we have not identified a combination of 3575 existing protocols that meets the requirements in Section 2. Also, 3576 we have not identified a path by which one of the existing protocols 3577 could be extended to meet the requirements. 3579 Authors' Addresses 3581 Carsten Bormann 3582 Universitaet Bremen TZI 3583 Postfach 330440 3584 D-28359 Bremen 3585 Germany 3587 Email: cabo@tzi.org 3588 Brian Carpenter (editor) 3589 Department of Computer Science 3590 University of Auckland 3591 PB 92019 3592 Auckland 1142 3593 New Zealand 3595 Email: brian.e.carpenter@gmail.com 3597 Bing Liu (editor) 3598 Huawei Technologies Co., Ltd 3599 Q14, Huawei Campus 3600 No.156 Beiqing Road 3601 Hai-Dian District, Beijing 100095 3602 P.R. China 3604 Email: leo.liubing@huawei.com