idnits 2.17.1 draft-iab-extension-recs-08.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 == Line 402 has weird spacing: '...es used by th...' == Line 705 has weird spacing: '...mber to decla...' == Line 706 has weird spacing: '...message to b...' == Line 707 has weird spacing: '... allows mail ...' == Line 709 has weird spacing: '...hich is presu...' == The document seems to contain a disclaimer for pre-RFC5378 work, but was first submitted on or after 10 November 2008. The disclaimer is usually necessary only for documents that revise or obsolete older RFCs, and that take significant amounts of text from those RFCs. If you can contact all authors of the source material and they are willing to grant the BCP78 rights to the IETF Trust, you can and should remove the disclaimer. Otherwise, the disclaimer is needed and you can ignore this comment. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (22 October 2011) is 4570 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 5226 (Obsoleted by RFC 8126) -- Obsolete informational reference (is this intentional?): RFC 822 (Obsoleted by RFC 2822) -- Obsolete informational reference (is this intentional?): RFC 1341 (Obsoleted by RFC 1521) -- Obsolete informational reference (is this intentional?): RFC 1521 (Obsoleted by RFC 2045, RFC 2046, RFC 2047, RFC 2048, RFC 2049) -- Obsolete informational reference (is this intentional?): RFC 2246 (Obsoleted by RFC 4346) -- Obsolete informational reference (is this intentional?): RFC 2284 (Obsoleted by RFC 3748) -- Obsolete informational reference (is this intentional?): RFC 2616 (Obsoleted by RFC 7230, RFC 7231, RFC 7232, RFC 7233, RFC 7234, RFC 7235) -- Obsolete informational reference (is this intentional?): RFC 2671 (Obsoleted by RFC 6891) -- Obsolete informational reference (is this intentional?): RFC 2822 (Obsoleted by RFC 5322) -- Obsolete informational reference (is this intentional?): RFC 3427 (Obsoleted by RFC 5727) -- Obsolete informational reference (is this intentional?): RFC 3588 (Obsoleted by RFC 6733) -- Obsolete informational reference (is this intentional?): RFC 4366 (Obsoleted by RFC 5246, RFC 6066) -- Obsolete informational reference (is this intentional?): RFC 5246 (Obsoleted by RFC 8446) -- Obsolete informational reference (is this intentional?): RFC 5405 (Obsoleted by RFC 8085) -- Obsolete informational reference (is this intentional?): RFC 5996 (Obsoleted by RFC 7296) == Outdated reference: A later version (-04) exists of draft-ietf-idr-optional-transitive-03 Summary: 1 error (**), 0 flaws (~~), 8 warnings (==), 15 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group B. Carpenter 3 Internet-Draft B. Aboba (ed) 4 Intended Status: Informational S. Cheshire 5 Expires: May 23, 2012 Internet Architecture Board 6 22 October 2011 8 Design Considerations for Protocol Extensions 9 draft-iab-extension-recs-08 11 Abstract 13 This document discusses issues related to the extensibility of 14 Internet protocols, with a focus on architectural design 15 considerations. It is intended to assist designers of both base 16 protocols and extensions. Case studies are included. 18 Status of this Memo 20 This Internet-Draft is submitted to IETF in full conformance with the 21 provisions of BCP 78 and BCP 79. 23 Internet-Drafts are working documents of the Internet Engineering 24 Task Force (IETF), its areas, and its working groups. Note that 25 other groups may also distribute working documents as Internet- 26 Drafts. 28 Internet-Drafts are draft documents valid for a maximum of six months 29 and may be updated, replaced, or obsoleted by other documents at any 30 time. It is inappropriate to use Internet-Drafts as reference 31 material or to cite them other than as "work in progress." 33 The list of current Internet-Drafts can be accessed at 34 http://www.ietf.org/ietf/1id-abstracts.txt. 36 The list of Internet-Draft Shadow Directories can be accessed at 37 http://www.ietf.org/shadow.html. 39 This Internet-Draft will expire on May 23, 2012. 41 Copyright Notice 43 Copyright (c) 2011 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 This document may contain material from IETF Documents or IETF 57 Contributions published or made publicly available before November 58 10, 2008. The person(s) controlling the copyright in some of this 59 material may not have granted the IETF Trust the right to allow 60 modifications of such material outside the IETF Standards Process. 61 Without obtaining an adequate license from the person(s) controlling 62 the copyright in such materials, this document may not be modified 63 outside the IETF Standards Process, and derivative works of it may 64 not be created outside the IETF Standards Process, except to format 65 it for publication as an RFC or to translate it into languages other 66 than English. 68 Table of Contents 70 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 71 1.1 Requirements Language . . . . . . . . . . . . . . . . . . 5 72 2. Routine and Major Extensions . . . . . . . . . . . . . . . . . 5 73 2.1 When is an Extension Routine? . . . . . . . . . . . . . . 5 74 2.2 What Constitutes a Major Extension? . . . . . . . . . . . 6 75 3. Architectural Principles . . . . . . . . . . . . . . . . . . . 7 76 3.1 Limited Extensibility . . . . . . . . . . . . . . . . . . 7 77 3.2 Design for Global Interoperability . . . . . . . . . . . . 8 78 3.3 Architectural Compatibility . . . . . . . . . . . . . . . 9 79 3.4 Protocol Variations . . . . . . . . . . . . . . . . . . . 10 80 3.5 Testability . . . . . . . . . . . . . . . . . . . . . . . 12 81 3.6 Parameter Parameter Registration . . . . . . . . . . . . . 12 82 3.7 Extensions to Critical Protocols . . . . . . . . . . . . . 14 83 4. Considerations for the Base Protocol . . . . . . . . . . . . . 15 84 4.1 Version Numbers . . . . . . . . . . . . . . . . . . . . . 16 85 4.2 Reserved Fields . . . . . . . . . . . . . . . . . . . . . 19 86 4.3 Encoding Formats . . . . . . . . . . . . . . . . . . . . . 19 87 4.4 Parameter Space Design . . . . . . . . . . . . . . . . . . 20 88 4.5 Cryptographic Agility . . . . . . . . . . . . . . . . . . 22 89 4.6 Transport . . . . . . . . . . . . . . . . . . . . . . . . 23 90 4.7 Handling of Unknown Extensions . . . . . . . . . . . . . . 24 91 5. Security Considerations . . . . . . . . . . . . . . . . . . . 25 92 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 26 93 7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 26 94 7.1 Normative References . . . . . . . . . . . . . . . . . . . 26 95 7.2 Informative References . . . . . . . . . . . . . . . . . . 26 96 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 30 97 IAB Members . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 98 Appendix A - Examples . . . . . . . . . . . . . . . . . . . . . . 31 99 A.1 Already documented cases . . . . . . . . . . . . . . . . . 31 100 A.2 RADIUS Extensions . . . . . . . . . . . . . . . . . . . . 31 101 A.3 TLS Extensions . . . . . . . . . . . . . . . . . . . . . . 32 102 A.4 L2TP Extensions . . . . . . . . . . . . . . . . . . . . . 34 103 Change log . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 104 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 36 105 1. Introduction 107 When developing protocols, IETF Working Groups (WGs) often include 108 mechanisms whereby these protocols can be extended in the future. It 109 is a good principle to design extensibility into protocols; as 110 described in "What Makes for a Successful Protocol" [RFC5218], a 111 common definition of a successful protocol is one that becomes widely 112 used in ways not originally anticipated. Well-designed extensibility 113 mechanisms facilitate the evolution of protocols and help make it 114 easier to roll out incremental changes in an interoperable fashion. 116 When an initial protocol design is extended, there is always a risk 117 of unintended consequences, such as interoperability problems or 118 security vulnerabilities. This risk is especially high if the 119 extension is performed by a different team than the original 120 designers, who may stray outside implicit design constraints or 121 assumptions. As a result, extensions should be done carefully and 122 with a full understanding of the base protocol, existing 123 implementations, and current operational practice. 125 The proliferation of extensions, even well designed ones, can be 126 costly. As noted in RFC 5321 [RFC5321] Section 2.2.1: 128 Experience with many protocols has shown that protocols with few 129 options tend towards ubiquity, whereas protocols with many options 130 tend towards obscurity. 132 Each and every extension, regardless of its benefits, must be 133 carefully scrutinized with respect to its implementation, 134 deployment, and interoperability costs. 136 This is hardly a recent concern. "TCP Extensions Considered Harmful" 137 [RFC1263] was published in 1991. "Extend" or "extension" occurs in 138 the title of more than 400 existing Request For Comment (RFC) 139 documents. Yet generic extension considerations have not been 140 documented previously. 142 The purpose of this document is to describe the architectural 143 principles of sound extensibility design, in order to minimize such 144 risks. Formal procedures for extending IETF protocols are discussed 145 in "Procedures for Protocol Extensions and Variations" BCP 125 146 [RFC4775]. 148 The rest of this document is organized as follows: Section 2 149 discusses routine and major extensions. Section 3 describes 150 architectural principles for protocol extensibility. Section 4 151 explains how designers of base protocols can take steps to anticipate 152 and facilitate the creation of such subsequent extensions in a safe 153 and reliable manner. 155 Readers are advised to study the whole document, since the 156 considerations are closely linked. 158 1.1. Requirements Language 160 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 161 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 162 document are to be interpreted as described in BCP 14, RFC 2119 163 [RFC2119]. 165 2. Routine and Major Extensions 167 To assist extension designers and reviewers, protocol documents 168 should provide guidelines explaining how extensions should be 169 performed, and guidance on the appropriate use of protocol extension 170 mechanisms should be developed. 172 Protocol components that are designed with the specific intention of 173 allowing extensibility should be clearly identified, with specific 174 and complete instructions on how to extend them. This includes the 175 process for adequate review of extension proposals: do they need 176 community review and if so how much and by whom? 178 The level of review required for protocol extensions will typically 179 vary based on the nature of the extension. Routine extensions may 180 require minimal review, while major extensions may require wide 181 review. Guidance on which extensions may be considered 'routine' and 182 which ones are 'major' are provided in the sections that follow. 184 2.1. When is an Extension Routine? 186 An extension may be considered 'routine' if its handling is opaque to 187 the protocol itself (e.g. does not substantially change the pattern 188 of messages and responses). For this to apply, no changes to the 189 base protocol can be required, nor can changes be required to 190 existing and currently deployed implementations, unless they make use 191 of the extension. Furthermore, existing implementations should not 192 be impacted. This typically requires that implementations be able to 193 ignore 'routine' extensions without ill-effects. 195 Examples of routine extensions include the Dynamic Host Configuration 196 Protocol (DHCP) vendor-specific option [RFC2132], Remote 197 Authentication Dial In User Service (RADIUS) Vendor-Specific 198 Attributes [RFC2865], the enterprise Object IDentifier (OID) tree for 199 Management Information Base (MIB) modules, vendor Multipurpose 200 Internet Mail Extension (MIME) types, and some classes of (non- 201 critical) certification extensions. Such extensions can safely be 202 made with minimal discussion. 204 Processes that allow routine extensions with minimal or no review 205 should be used sparingly (such as the "First Come First Served" 206 (FCFS) allocation policy described in "Guidelines for Writing an IANA 207 Considerations Section in RFCs" [RFC5226]). In particular, they 208 should be limited to cases that are unlikely to result in 209 interoperability problems, or security or operational exposures. 211 Experience has shown that even routine extensions may benefit from 212 review by experts. For example, even though DHCP carries opaque 213 data, defining a new option using completely unstructured data may 214 lead to an option that is unnecessarily hard for clients and servers 215 to process. 217 2.2. What Constitutes a Major Extension? 219 Major extensions may have characteristics leading to a risk of 220 interoperability failure. Where these characteristics are present, 221 it is necessary to pay close attention to backward compatibility with 222 implementations and deployments of the unextended protocol, and to 223 the risk of inadvertent introduction of security or operational 224 exposures. 226 Extension designers should examine their design for the following 227 issues: 229 1. Modifications or extensions to the underlying protocol. This 230 can include specification of additional transports (see Section 231 4.6), changing protocol semantics or defining new message types 232 that may require implementation changes in existing and deployed 233 implementations of the protocol, even if they do not want to make 234 use of the new functions. A base protocol that does not uniformly 235 permit "silent discard" of unknown extensions may automatically 236 enter this category, even for apparently minor extensions. 237 Handling of "unknown" extensions is discussed in more detail in 238 Section 4.7. 240 2. Changes to the basic architectural assumptions. This may 241 include architectural assumptions that are explicitly stated or 242 those that have been assumed by implementers. For example, this 243 would include adding a requirement for session state to a 244 previously stateless protocol. 246 3. New usage scenarios not originally intended or investigated. 247 This can potentially lead to operational difficulties when 248 deployed, even in cases where the "on-the-wire" format has not 249 changed. For example, the level of traffic carried by the 250 protocol may increase substantially, packet sizes may increase, 251 and implementation algorithms that are widely deployed may not 252 scale sufficiently or otherwise be up to the new task at hand. 253 For example, a new DNS Resource Record (RR) type that is too big 254 to fit into a single UDP packet could cause interoperability 255 problems with existing DNS clients and servers. 257 4. Changes to the extension model. Adverse impacts are very 258 likely if the base protocol contains an extension mechanism and 259 the proposed extension does not fit into the model used to create 260 and define that mechanism. Extensions that have the same 261 properties as those that were anticipated when an extension 262 mechanism was devised are much less likely to be disruptive than 263 extensions that don't fit the model. 265 5. Changes to protocol syntax. changes to protocol syntax bring 266 with them the potential for backward compatibility issues. If at 267 all possible, extensions should be designed for compatibility with 268 existing syntax, so as to avoid interoperability failures. 270 3. Architectural Principles 272 This section describes basic principles of protocol extensibility: 274 1. Extensibility features should be limited to what is reasonably 275 anticipated when the protocol is developed. 277 2. Protocol extensions should be designed for global 278 interoperability. 280 3. Protocol extensions should be architecturally compatible with 281 the base protocol. 283 4. Protocol extension mechanisms should not be used to create 284 incompatible protocol variations. 286 5. Extension mechanisms need to be testable. 288 6. Protocol parameter assignments need to be coordinated to avoid 289 potential conflicts. 291 7. Extensions to critical protocols require special care. 293 3.1. Limited Extensibility 295 Designing a protocol for extensibility may have the perverse side 296 effect of making it easy to construct incompatible extensions. 298 Consequently, protocols should not be made more extensible than 299 clearly necessary at inception, and the process for defining new 300 extensibility mechanisms should ensure that adequate review of 301 proposed extensions will take place before widespread adoption. 303 3.2. Design for Global Interoperability 305 The IETF mission [RFC3935] is to create interoperable protocols for 306 the global Internet, not a collection of different incompatible 307 protocols (or "profiles") for use in separate private networks. 308 Experience shows that separate private networks often end up using 309 equipment from the same vendors, or end up having portable equipment 310 like laptop computers move between them, and networks that were 311 originally envisaged as being separate can end up being connected 312 later. 314 As a result, extensions cannot be designed for an isolated 315 environment; instead, extension designers must assume that systems 316 using the extension will need to interoperate with systems on the 317 global Internet. 319 A key requirement for interoperable extension design is that the base 320 protocol must be well designed for interoperability, and that 321 extensions must have unambiguous semantics. Ideally, the protocol 322 mechanisms for extension and versioning should be sufficiently well 323 described that compatibility can be assessed on paper. Otherwise, 324 when two "private" extensions encounter each other on a public 325 network, unexpected interoperability problems may occur. 327 Consider a "private" extension installed on a work computer which, 328 being portable, is sometimes connected to a home network or a hotel 329 network. If the "private" extension is incompatible with an 330 unextended version of the same protocol, problems will occur. 332 Similarly, problems can occur if "private" extensions conflict with 333 each other. For example, imagine the situation where one site chose 334 to use DHCP [RFC2132] option code 62 for one meaning, and a different 335 site chose to use DHCP option code 62 for a completely different, 336 incompatible, meaning. It may be impossible for a vendor of portable 337 computing devices to make a device that works correctly in both 338 environments. 340 One approach to solving this problem has been to reserve parts of an 341 identifier namespace for "site-specific" use, such as "X-" headers in 342 email messages [RFC0822]. This problem with this approach is that 343 when a site-specific use turns out to have applicability elsewhere, 344 other vendors will then implement that "X-" header for 345 interoperability, and the "X-" header becomes a de-facto standard, 346 meaning that it is no longer true that any header beginning "X-" is 347 site-specific. The notion of "X-" headers was removed from the 348 Internet Message Format standard when it was updated in 2001 349 [RFC2822]. 351 3.3. Architectural Compatibility 353 Since protocol extension mechanisms may impact interoperability, it 354 is important that they be architecturally compatible with the base 355 protocol. 357 As part of the definition of new extension mechanisms, it is 358 important to address whether the mechanisms make use of features as 359 envisaged by the original protocol designers, or whether a new 360 extension mechanism is being invented. If a new extension mechanism 361 is being invented, then architectural compatibility issues need to be 362 addressed. 364 To assist in the assessment of architectural compatibility, protocol 365 documents should provide guidelines explaining how extensions should 366 be performed, and guidance on the appropriate use of protocol 367 extension mechanisms should be developed. Protocol components that 368 are designed with the specific intention of allowing extensibility 369 should be clearly identified, with specific and complete instructions 370 on how to extend them. This includes the process for adequate review 371 of extension proposals: do they need community review and if so how 372 much and by whom? 374 Documents relying on extension mechanisms need to explicitly identify 375 the mechanisms being relied upon. For example, a document defining 376 new data elements should not implicitly define new data types or 377 protocol operations without explicitly describing those dependencies 378 and discussing their impact. Where extension guidelines are 379 available, mechanisms need to indicate whether they are compliant 380 with those guidelines and if not, why not. 382 Examples of extension guidelines documents include: 384 1. "Guidelines for Extending the Extensible Provisioning Protocol 385 (EPP)" [RFC3735], which provides guidelines for use of EPP's 386 extension mechanisms to define new features and object management 387 capabilities. 389 2. "Guidelines for Authors and Reviewers of MIB Documents" BCP 111 390 [RFC4181], which provides guidance to protocol designers creating 391 new MIB modules. 393 3. "Guidelines for Authors of Extensions to the Session Initiation 394 Protocol (SIP)" [RFC4485], which outlines guidelines for authors 395 of SIP extensions. 397 4. "Considerations for Lightweight Directory Access Protocol 398 (LDAP) Extensions" BCP 118 [RFC4521], which discusses 399 considerations for designers of LDAP extensions. 401 5. "RADIUS Design Guidelines" BCP 158 [RFC6158], which provides 402 guidelines for the design of attributes used by the Remote 403 Authentication Dial In User Service (RADIUS) protocol. 405 3.4. Protocol Variations 407 Protocol variations - specifications that look very similar to the 408 original but don't interoperate with each other or with the original 409 - are even more harmful to interoperability than extensions. In 410 general, such variations should be avoided. Causes of protocol 411 variations include incompatible protocol extensions, uncoordinated 412 protocol development, and poorly designed "profiles". 414 Protocol extension mechanisms should not be used to create 415 incompatible forks in development. An extension may lead to 416 interoperability failures unless the extended protocol correctly 417 supports all mandatory and optional features of the unextended base 418 protocol, and implementations of the base protocol operate correctly 419 in the presence of the extensions. In addition, it is necessary for 420 an extension to interoperate with other extensions. 422 As noted in "Uncoordinated Protocol Development Considered Harmful" 423 [RFC5704] Section 1, incompatible forks in development can result 424 from the uncoordinated adaptation of a protocol, parameter or code- 425 point: 427 In particular, the IAB considers it an essential principle of the 428 protocol development process that only one SDO maintains design 429 authority for a given protocol, with that SDO having ultimate 430 authority over the allocation of protocol parameter code-points 431 and over defining the intended semantics, interpretation, and 432 actions associated with those code-points. 434 3.4.1. Profiles 436 Profiling is a common technique for improving interoperability within 437 a target environment or set of scenarios. Generally speaking, there 438 are two approaches to profiling: 440 a) Removal or downgrading of normative requirements (thereby creating 441 potential interoperability problems); 442 b) Elevation of normative requirement levels (such as from a 443 MAY/SHOULD to a MUST) in order to improve interoperability by 444 narrowing potential implementation choices. 446 While approach a) is potentially harmful, approach b) may be 447 beneficial, but is typically only necessary when the underlying 448 protocol is ill-defined enough to permit non-interoperable yet 449 compliant implementations. 451 In order to avoid creating interoperability problems when profiled 452 implementations interact with others over the Global Internet, 453 profilers need to remain cognizant of the implications of normative 454 requirements. 456 As noted in "Key words for use in RFCs to Indicate Requirement 457 Levels" [RFC2119] Section 6, imperatives are to be used with care, 458 and as a result, their removal within a profile is likely to result 459 in serious consequences: 461 Imperatives of the type defined in this memo must be used with 462 care and sparingly. In particular, they MUST only be used where 463 it is actually required for interoperation or to limit behavior 464 which has potential for causing harm (e.g., limiting 465 retransmissions) For example, they must not be used to try to 466 impose a particular method on implementors where the method is not 467 required for interoperability. 469 As noted in [RFC2119] Sections 3 and 4, recommendations cannot be 470 removed from profiles without serious consideration: 472 there may exist valid reasons in particular circumstances to 473 ignore a particular item, but the full implications must be 474 understood and carefully weighed before choosing a different 475 course. 477 Even the removal of optional features and requirements can have 478 consequences. As noted in [RFC2119] Section 5, implementations which 479 do not support optional features still retain the obligation to 480 ensure interoperation with implementations that do: 482 An implementation which does not include a particular option MUST 483 be prepared to interoperate with another implementation which does 484 include the option, though perhaps with reduced functionality. In 485 the same vein an implementation which does include a particular 486 option MUST be prepared to interoperate with another 487 implementation which does not include the option (except, of 488 course, for the feature the option provides.) 490 3.5. Testability 492 Experience has shown that it is insufficient merely to correctly 493 specify extensibility and backwards compatibility in an RFC. It is 494 also important that implementations respect the compatibility 495 mechanisms; if not, non-interoperable pairs of implementations may 496 arise. The TLS case study (Appendix A.3) shows how important this 497 can be. 499 In order to determine whether protocol extension mechanisms have been 500 properly implemented, testing is required. However, for this to be 501 possible, test cases need to be developed. If a base protocol 502 document specifies extension mechanisms but does not utilize them or 503 provide examples, it may not be possible to develop effective test 504 cases based on the base protocol specification alone. As a result, 505 base protocol implementations may not be properly tested and non- 506 compliant extension behavior may not be detected until these 507 implementations are widely deployed. 509 To encourage correct implementation of extension mechanisms, base 510 protocol specifications should clearly articulate the expected 511 behavior of extension mechanisms and should include examples of 512 correct and incorrect extension behavior. 514 3.6. Protocol Parameter Registration 516 An extension is often likely to make use of additional values added 517 to an existing IANA registry. To avoid conflicting usage of the same 518 value, as well as to prevent potential difficulties in determining 519 and transferring parameter ownership, it is essential that all new 520 values are properly registered by the applicable procedures. If this 521 is not done, there is nothing to prevent two different extensions 522 picking the same value. When these two extensions "meet" each other 523 on the Internet, failure is inevitable. 525 A surprisingly common case of this is misappropriation of assigned 526 Transmission Control Protocol (TCP) (or User Datagram Protocol (UDP)) 527 registered port numbers. This can lead to a client for one service 528 attempting to communicate with a server for another service. 529 Numerous cases could be cited, but not without embarrassing specific 530 implementers. 532 For general rules see [RFC5226], and for specific rules and 533 registries see the individual protocol specification RFCs and the 534 IANA web site. While in theory a "standards track" or "IETF 535 consensus" parameter allocation policy may be instituted to encourage 536 protocol parameter registration or to improve interoperability, in 537 practice problems can arise if the procedures result in so much delay 538 that requesters give up and "self-allocate" by picking presumably- 539 unused code points. 541 3.6.1. Experimental and Local Use 543 In some cases, it may be appropriate to use values designated as 544 "experimental" or "local use" in early implementations of an 545 extension. For example, "Experimental Values in IPv4, IPv6, ICMPv4, 546 ICMPv6, UDP and TCP Headers" [RFC4727] discusses experimental values 547 for IP and transport headers, and "Definition of the Differentiated 548 Services Field (DS Field) in the IPv4 and IPv6 Headers" [RFC2474] 549 defines experimental/local use ranges for differentiated services 550 code points. 552 Such values should be used with care and only for their stated 553 purpose: experiments and local use. They are unsuitable for 554 Internet-wide use, since they may be used for conflicting purposes 555 and thereby cause interoperability failures. Packets containing 556 experimental or local use values must not be allowed out of the 557 domain in which they are meaningful. 559 As noted in [RFC5226] Section 4.1: 561 For private or local use... No attempt is made to prevent multiple 562 sites from using the same value in different (and incompatible) 563 ways... assignments are not generally useful for broad 564 interoperability. It is the responsibility of the sites making 565 use of the Private Use range to ensure that no conflicts occur 566 (within the intended scope of use). 568 "Assigning Experimental and Testing Numbers Considered Useful" BCP 82 569 [RFC3692] Section 1 provides additional guidance on the use of 570 experimental code points: 572 Numbers in the experimentation range.... are not intended to be 573 used in general deployments or be enabled by default in products 574 or other general releases. In those cases where a product or 575 release makes use of an experimental number, the end user must be 576 required to explicitly enable the experimental feature and 577 likewise have the ability to chose and assign which number from 578 the experimental range will be used for a specific purpose (i.e., 579 so the end user can ensure that use of a particular number doesn't 580 conflict with other on-going uses). Shipping a product with a 581 specific value pre-enabled would be inappropriate and can lead to 582 interoperability problems when the chosen value collides with a 583 different usage, as it someday surely will. 585 From the above, it follows that it would be inappropriate for a 586 group of vendors, a consortia, or another Standards Development 587 Organization to agree among themselves to use a particular value 588 for a specific purpose and then agree to deploy devices using 589 those values. By definition, experimental numbers are not 590 guaranteed to be unique in any environment other than one where 591 the local system administrator has chosen to use a particular 592 number for a particular purpose and can ensure that a particular 593 value is not already in use for some other purpose. 595 3.7. Extensions to Critical Protocols 597 Some protocols (such as Domain Name Service (DNS) and Border Gateway 598 Protocol (BGP)) have become critical components of the Internet 599 infrastructure. When such protocols are extended, the potential 600 exists for negatively impacting the reliability and security of the 601 global Internet. 603 As a result, special care needs to be taken with these extensions, 604 such as taking explicit steps to isolate existing uses from new ones. 605 For example, this can be accomplished by requiring the extension to 606 utilize a different port or multicast address, or by implementing the 607 extension within a separate process, without access to the data and 608 control structures of the base protocol. 610 Experience has shown that even when a mechanism has proven benign in 611 other uses, unforseen issues may result when adding it to a critical 612 protocol. For example, both ISIS and OSPF support opaque Link State 613 Attributes (LSAs) which are propagated by intermediate nodes that 614 don't understand the LSA. Within Interior Gateway Protocols (IGPs), 615 support for opaque LSAs has proven useful without introducing 616 instability. 618 However, within BGP, 'attribute tunneling' has resulted in large 619 scale routing instabilities, since remote nodes may reset the LOCAL 620 session if the tunneled attributes are malformed or aren't 621 understood. This has required modification to BGP error handling, as 622 noted in "Error Handling for Optional Transitive Attribute BGP 623 Attributes" [Transitive]. 625 In general, when extending protocols with local failure conditions, 626 tunneling of attributes that may trigger failures in non-adjacent 627 nodes should be avoided. This is particularly problematic when the 628 originating node receives no indicators of remote failures it may 629 have triggered. 631 4. Considerations for the Base Protocol 633 Good extension design depends on a well designed base protocol. 634 Interoperability stems from a number of factors, including: 636 1. A well-written base protocol specification. Does the base 637 protocol specification make clear what an implementor needs to 638 support and does it define the impact that individual operations 639 (e.g. a message sent to a peer) will have when invoked? 641 2. Design for deployability. This includes understanding what 642 current implementations do and how a proposed extension will 643 interact with deployed systems. Is it clear when a proposed 644 extension (or its proposed usage) will operationally stress 645 existing implementations or the underlying protocol itself if 646 widely deployed? If this is not explained in the base protocol 647 specification, is this covered in an extension design guidelines 648 document? 650 3. Design for backward compatibility. Does the base protocol 651 specification describe how to determine the capabilities of a 652 peer, and negotiate the use of extensions? Does it indicate how 653 implementations handle extensions that they do not understand? Is 654 it possible for an extended implementation to negotiate with an 655 unextended peer to find a common subset of useful functions? 657 4. Respecting underlying architectural or security assumptions. 658 Is there a document describing the underlying architectural 659 assumptions, as well as considerations that have arisen in 660 operational experience? Or are there undocumented considerations 661 that have arisen as the result of operational experience, after 662 the original protocol was published? 664 For example, will backward compatibility issues arise if 665 extensions reverse the flow of data, allow formerly static 666 parameters to be changed on the fly, or change assumptions 667 relating to the frequency of reads/writes? 669 5. Minimizing impact on critical infrastructure. For a protocol 670 that represents a critical element of Internet infrastructure, it 671 is important to explain when it is appropriate to isolate new uses 672 of the protocol from existing ones. 674 For example, is it explained when a proposed extension (or usage) 675 has the potential for negatively impacting critical infrastructure 676 to the point where explicit steps would be appropriate to isolate 677 existing uses from new ones? 678 6. Data model extensions. Is there a document that explains when 679 a protocol extension is routine and when it represents a major 680 change? 682 For example, is it clear when a data model extension represents a 683 major versus a routine change? Are there guidelines describing 684 when an extension (such as a new data type) is likely to require a 685 code change within existing implementations? in -0.3i 687 4.1. Version Numbers 689 Any mechanism for extension by versioning must include provisions to 690 ensure interoperability, or at least clean failure modes. Imagine 691 someone creating a protocol and using a "version" field and 692 populating it with a value (1, let's say), but giving no information 693 about what would happen when a new version number appears in it. 694 That's bad protocol design and description; it should be clear what 695 the expectation is and how you test it. For example, stating that 696 1.X must be compatible with any version 1 code, but version 2 or 697 greater is not expected to be compatible, has different implications 698 than stating that version 1 must be a proper subset of version 2. 700 An example of an under-specified versioning mechanism is provided by 701 the MIME-Version header, originally defined in "MIME (Multipurpose 702 Internet Mail Extensions)" [RFC1341]. As noted in [RFC1341] Section 703 1: 705 A MIME-Version header field....uses a version number to declare 706 a message to be conformant with this specification and 707 allows mail processing agents to distinguish between such 708 messages and those generated by older or non-conformant software, 709 which is presumed to lack such a field. 711 Beyond this, [RFC1341] provided little guidance on versioning 712 behavior, or even the format of the MIME-Version header, which was 713 specified to contain "text". [RFC1521] which obsoleted [RFC1341], 714 better defined the format of the version field, but still did not 715 clarify the versioning behavior: 717 Thus, future format specifiers, which might replace or extend 718 "1.0", are constrained to be two integer fields, separated by a 719 period. If a message is received with a MIME-version value other 720 than "1.0", it cannot be assumed to conform with this 721 specification... 723 It is not possible to fully specify how a mail reader that 724 conforms with MIME as defined in this document should treat a 725 message that might arrive in the future with some value of MIME- 726 Version other than "1.0". However, conformant software is 727 encouraged to check the version number and at least warn the user 728 if an unrecognized MIME- version is encountered. 730 Thus, even though [RFC1521] defined a MIME-Version header with a 731 syntax suggestive of a "Major/Minor" versioning scheme, in practice 732 the MIME-Version header was little more than a decoration. 734 A better example is ROHC (Robust Header Compression). ROHCv1 735 [RFC3095] supports a certain set of profiles for compression 736 algorithms. But experience had shown that these profiles had 737 limitations, so the ROHC WG developed ROHCv2 [RFC5225]. A ROHCv1 738 implementation does not contain code for the ROHCv2 profiles. As the 739 ROHC WG charter said during the development of ROHCv2: 741 It should be noted that the v2 profiles will thus not be 742 compatible with the original (ROHCv1) profiles, which means less 743 complex ROHC implementations can be realized by not providing 744 support for ROHCv1 (over links not yet supporting ROHC, or by 745 shifting out support for ROHCv1 in the long run). Profile support 746 is agreed through the ROHC channel negotiation, which is part of 747 the ROHC framework and thus not changed by ROHCv2. 749 Thus in this case both backwards-compatible and backwards- 750 incompatible deployments are possible. The important point is a 751 clearly thought out approach to the question of operational 752 compatibility. In the past, protocols have utilized a variety of 753 strategies for versioning, many of which have proven problematic. 754 These include: 756 1. No versioning support. This approach is exemplified by 757 Extensible Authentication Protocol (EAP) [RFC3748] as well as 758 Remote Authentication Dial In User Service (RADIUS) [RFC2865], 759 both of which provide no support for versioning. While lack of 760 versioning support protects against the proliferation of 761 incompatible dialects, the need for extensibility is likely to 762 assert itself in other ways, so that ignoring versioning entirely 763 may not be the most forward thinking approach. 765 2. Highest mutually supported version (HMSV). In this approach, 766 implementations exchange the version numbers of the highest 767 version each supports, with the negotiation agreeing on the 768 highest mutually supported protocol version. This approach 769 implicitly assumes that later versions provide improved 770 functionality, and that advertisement of a particular version 771 number implies support for all lower version numbers. Where these 772 assumptions are invalid, this approach breaks down, potentially 773 resulting in interoperability problems. An example of this issue 774 occurs in Protected Extensible Authentication Protocol [PEAP] 775 where implementations of higher versions may not necessarily 776 provide support for lower versions. 778 3. Assumed backward compatibility. In this approach, 779 implementations may send packets with higher version numbers to 780 legacy implementations supporting lower versions, but with the 781 assumption that the legacy implementations will interpret packets 782 with higher version numbers using the semantics and syntax defined 783 for lower versions. This is the approach taken by Port-Based 784 Access Control [IEEE-802.1X]. For this approach to work, legacy 785 implementations need to be able to accept packets of known types 786 with higher protocol versions without discarding them; protocol 787 enhancements need to permit silent discard of unsupported 788 extensions; implementations supporting higher versions need to 789 refrain from mandating new features when encountering legacy 790 implementations. 792 4. Major/minor versioning. In this approach, implementations with 793 the same major version but a different minor version are assumed 794 to be backward compatible, but implementations are required to 795 negotiate a mutually supported major version number. This 796 approach assumes that implementations with a lower minor version 797 number but the same major version can safely ignore unsupported 798 protocol messages. 800 5. Min/max versioning. This approach is similar to HMSV, but 801 without the implied obligation for clients and servers to support 802 all versions back to version 1, in perpetuity. It allows clients 803 and servers to cleanly drop support for early versions when those 804 versions become so old that they are no longer relevant and no 805 longer required. In this approach, the client initiating the 806 connection reports the highest and lowest protocol versions it 807 understands. The server reports back the chosen protocol version: 809 a. If the server understands one or more versions in the client's 810 range, it reports back the highest mutually understood version. 812 b. If there is no mutual version, then the server reports back 813 some version that it does understand (selected as described 814 below). The connection is then typically dropped by client or 815 server, but reporting this version number first helps facilitate 816 useful error messages at the client end: 818 * If there is no mutual version, and the server speaks any 819 version higher than client max, it reports the lowest version it 820 speaks which is greater than the client max. The client can 821 then report to the user, "You need to upgrade to at least 822 version ." 824 * Else, the server reports the highest version it speaks. The 825 client can then report to the user, "You need to request the 826 server operator to upgrade to at least version ." 828 Protocols generally do not need any version-negotiation mechanism 829 more complicated than the mechanisms described here. The nature of 830 protocol version-negotiation mechanisms is that, by definition, they 831 don't get widespread real-world testing until *after* the base 832 protocol has been deployed for a while, and its deficiencies have 833 become evident. This means that, to be useful, a protocol version 834 negotiation mechanism should be simple enough that it can reasonably 835 be assumed that all the implementers of the first protocol version at 836 least managed to implement the version-negotiation mechanism 837 correctly. 839 4.2. Reserved Fields 841 Protocols commonly include one or more "reserved" fields, clearly 842 intended for future extensions. It is good practice to specify the 843 value to be inserted in such a field by the sender (typically zero) 844 and the action to be taken by the receiver when seeing some other 845 value (typically no action). In packet format diagrams, such fields 846 are typically labeled "MBZ", to be read as, "Must Be Zero on 847 transmission, Must Be Ignored on reception." 849 A common mistake of inexperienced protocol implementers is to think 850 that "MBZ" means that it's their software's job to verify that the 851 value of the field is zero on reception, and reject the packet if 852 not. This is a mistake, and such software will fail when it 853 encounters future versions of the protocol where these previously 854 reserved fields are given new defined meanings. Similarly, protocols 855 should carefully specify how receivers should react to unknown 856 extensions (headers, TLVs etc.), such that failures occur only when 857 that is truly the intended outcome. 859 4.3. Encoding Formats 861 Using widely-supported encoding formats leads to better 862 interoperability and easier extensibility. 864 As described in "IAB Thoughts on Encodings for International Domain 865 Names" [RFC6055], the number of encodings should be minimized and 866 complex encodings are generally a bad idea. As soon as one moves 867 outside the ASCII repertoire, issues relating to collation and/or 868 comparison arise that extensions must handle with care. 870 An example is the Simple Network Management Protocol (SNMP) Structure 871 of Managed Information (SMI). Guidelines exist for defining the 872 Management Information Base (MIB) objects that SNMP carries 873 [RFC4181]. Also, multiple textual conventions have been published, 874 so that MIB designers do not have to reinvent the wheel when they 875 need a commonly encountered construct. For example, the "Textual 876 Conventions for Internet Network Addresses" [RFC4001] can be used by 877 any MIB designer needing to define objects containing IP addresses, 878 thus ensuring consistency as the body of MIBs is extended. 880 4.4. Parameter Space Design 882 In some protocols the parameter space is either infinite (e.g. Header 883 field names) or sufficiently large that it is unlikely to be 884 exhausted. In other protocols, the parameter space is finite, and in 885 some cases, has proven inadequate to accommodate demand. Common 886 mistakes include: 888 a. A version field that is too small (e.g. two bits or less). When 889 designing a version field, existing as well as potential versions of 890 a protocol need to be taken into account. For example, if a protocol 891 is being standardized for which there are existing implementations 892 with known interoperability issues, more than one version for "pre- 893 standard" implementations may be required. If two "pre-standard" 894 versions are required in addition to a version for an IETF standard, 895 then a two-bit version field would only leave one additional version 896 code-point for a future update, which could be insufficient. This 897 problem was encountered during the development of the PEAPv2 protocol 898 [PEAP]. 900 b. A small parameter space (e.g. 8-bits or less) along with a First 901 Come, First Served (FCFS) allocation policy. In general, an FCFS 902 allocation policy is only appropriate in situations where parameter 903 exhaustion is highly unlikely. In situations where substantial 904 demand is anticipated within a parameter space, the space should 905 either be designed to be sufficient to handle that demand, or vendor 906 extensibility should be provided to enable vendors to self-allocate. 907 The combination of a small parameter space, an FCFS allocation 908 policy, and no support for vendor extensibility is particularly 909 likely to prove ill-advised. An example of such a combination was 910 the design of the original 8-bit EAP Method Type space [RFC2284]. 912 Once the potential for parameter exhaustion becomes apparent, it is 913 important that it be addressed as quickly as possible. Protocol 914 changes can take years to appear in implementations and by then the 915 exhaustion problem could become acute. 917 Options for addressing a protocol parameter exhaustion problem 918 include: 920 Rethinking the allocation regime 921 Where it becomes apparent that the size of a parameter space is 922 insufficient to meet demand, it may be necessary to rethink the 923 allocation mechanism, in order to prevent or delay parameter space 924 exhaustion. In revising parameter allocation mechanisms, it is 925 important to consider both supply and demand aspects so as to avoid 926 unintended consequences such as self-allocation or the development 927 of black markets for the re-sale of protocol parameters. 929 For example, a few years after approval of RFC 2284 [RFC2284], it 930 became clear that the combination of a FCFS allocation policy and 931 lack of support for vendor-extensions had created the potential for 932 exhaustion of the EAP Method Type space within a few years. To 933 address the issue, [RFC3748] Section 6.2 changed the allocation 934 policy for EAP Method Types from FCFS to Expert Review, with 935 Specification Required. Since this allocation policy revision did 936 not change the demand for EAP Method Types, it would have been 937 likely to result in self-allocation within the standards space, had 938 mechanisms not been provided to expand the method type space 939 (including support for vendor-specific method types). 941 Support for vendor-specific parameters 942 If the demand that cannot be accommodated is being generated by 943 vendors, merely making allocation harder could make things worse if 944 this encourages vendors to self-allocate, creating interoperability 945 problems. In such a situation, support for vendor-specific 946 parameters should be considered, allowing each vendor to self- 947 allocate within their own vendor-specific space based on a vendor's 948 Private Enterprise Code (PEC). For example, in the case of the EAP 949 Method Type space, [RFC3748] Section 6.2 also provided for an 950 Expanded Type space for "functions specific only to one vendor's 951 implementation". 953 Extensions to the parameter space 954 If the goal is to stave off exhaustion in the face of high demand, 955 a larger parameter space may be helpful. Where vendor-specific 956 parameter support is available, this may be achieved by allocating 957 an PEC for IETF use. Otherwise it may be necessary to try to extend 958 the size of the parameter fields, which could require a new 959 protocol version or other substantial protocol changes. 961 Parameter reclamation 962 In order to gain time, it may be necessary to reclaim unused 963 parameters. However, it may not be easy to determine whether a 964 parameter that has been allocated is in use or not, particularly if 965 the entity that obtained the allocation no longer exists or has 966 been acquired (possibly multiple times). 968 Parameter Transfer 969 When all the above mechanisms have proved infeasible and parameter 970 exhaustion looms in the near future, enabling the transfer of 971 ownership of protocol parameters can be considered as a means for 972 improving allocation efficiency. However, enabling transfer of 973 parameter ownership can be far from simple if the parameter 974 allocation process was not originally designed to enable title 975 searches and ownership transfers. 977 A parameter allocation process designed to uniquely allocate code- 978 points is fundamentally different from one designed to enable title 979 search and transfer. If the only goal is to ensure that a 980 parameter is not allocated more than once, the parameter registry 981 will only need to record the initial allocation. On the other 982 hand, if the goal is to enable transfer of ownership of a protocol 983 parameter, then it is important not only to record the initial 984 allocation, but also to track subsequent ownership changes, so as 985 to make it possible to determine and transfer title. Given the 986 difficulty of converting from a unique allocation regime to one 987 requiring support for title search and ownership transfer, it is 988 best for the desired capabilities to be carefully thought through 989 at the time of registry establishment. 991 4.5. Cryptographic Agility 993 Extensibility with respect to cryptographic algorithms is desirable 994 in order to provide resilience against the compromise of any 995 particular algorithm. "Guidance for Authentication, Authorization, 996 and Accounting (AAA) Key Management" BCP 132 [RFC4962] Section 3 997 provides some basic advice: 999 The ability to negotiate the use of a particular cryptographic 1000 algorithm provides resilience against compromise of a particular 1001 cryptographic algorithm... This is usually accomplished by 1002 including an algorithm identifier and parameters in the protocol, 1003 and by specifying the algorithm requirements in the protocol 1004 specification. While highly desirable, the ability to negotiate 1005 key derivation functions (KDFs) is not required. For 1006 interoperability, at least one suite of mandatory-to-implement 1007 algorithms MUST be selected... 1009 This requirement does not mean that a protocol must support both 1010 public-key and symmetric-key cryptographic algorithms. It means 1011 that the protocol needs to be structured in such a way that 1012 multiple public-key algorithms can be used whenever a public-key 1013 algorithm is employed. Likewise, it means that the protocol needs 1014 to be structured in such a way that multiple symmetric-key 1015 algorithms can be used whenever a symmetric-key algorithm is 1016 employed. 1018 In practice, the most difficult challenge in providing cryptographic 1019 agility is providing for a smooth transition in the event that a 1020 mandatory-to-implement algorithm is compromised. Since it may take 1021 significant time to provide for widespread implementation of a 1022 previously undeployed alternative, it is often advisable to recommend 1023 implementation of alternative algorithms of distinct lineage in 1024 addition to those made mandatory-to-implement, so that an alternative 1025 algorithm is readily available. If such a recommended alternative is 1026 not in place, then it would be wise to issue such a recommendation as 1027 soon as indications of a potential weakness surface. This is 1028 particularly important in the case of potential weakness in 1029 algorithms used to authenticate and integrity-protect the 1030 cryptographic negotiation itself, such as KDFs or message integrity 1031 checks (MICs). Without secure alternatives to compromised KDF or MIC 1032 algorithms, it may not be possible to secure the cryptographic 1033 negotiation while retaining backward compatibility. 1035 4.6. Transport 1037 In the past, IETF protocols have been specified to operate over 1038 multiple transports. Often the protocol was originally specified to 1039 utilize a single transport, but limitations were discovered in 1040 subsequent deployment, so that additional transports were 1041 subsequently specified. 1043 In a number of cases, the protocol was originally specified to 1044 operate over UDP, but subsequent operation disclosed one or more of 1045 the following issues, leading to the specification of alternative 1046 transports: 1048 a. Payload fragmentation (often due to the introduction of 1049 extensions or additional usage scenarios); 1051 b. Problems with congestion control, transport reliability or 1052 efficiency; 1054 c. Lack of deployment in multicast scenarios, which had been a 1055 motivator for UDP transport. 1057 On the other hand, there are also protocols that were originally 1058 specified to operate over reliable transport that have subsequently 1059 defined transport over UDP, due to one or more of the following 1060 issues: 1062 d. NAT traversal concerns that were more easily addressed with UDP 1063 transport; 1065 e. Scalability problems, which could be improved by UDP transport. 1067 Since specification of a single transport offers the highest 1068 potential for interoperability, protocol designers should carefully 1069 consider not only initial but potential future requirements in the 1070 selection of a transport protocol. Where UDP transport is selected, 1071 the guidance provided in "Unicast UDP Usage Guidelines for 1072 Application Designers" [RFC5405]. should be taken into account. 1074 After significant deployment has occurred, there are few satisfactory 1075 options for addressing problems with the originally selected 1076 transport. While specification of additional transports is possible, 1077 removal of a widely implemented transport protocol is likely to 1078 result in interoperability problems and should be avoided. 1080 Mandating support for the initially selected transport, while 1081 designating additional transports as optional may have limitations. 1082 Since optional transport protocols are typically introduced due to 1083 the advantages they afford in certain scenarios, in those situations 1084 implementations not supporting optional transport protocols may 1085 exhibit degraded performance or may even fail. 1087 While mandating support for multiple transport protocols may appear 1088 attractive, designers need to realistically evaluate the likelihood 1089 that implementers will conform to the requirements. For example, 1090 where resources are limited (such as in embedded systems), 1091 implementers may choose to only support a subset of the mandated 1092 transport protocols, resulting in non-interoperable protocol 1093 variants. 1095 4.7. Handling of Unknown Extensions 1097 IETF protocols have utilized several techniques for handling of 1098 unknown extensions. One technique (often used for vendor-specific 1099 extensions) is to specify that unknown extensions be "silently 1100 discarded". 1102 While this approach can deliver a high level of interoperability, 1103 there are situations in which it is problematic. For example, where 1104 security functionality is involved, "silent discard" may not be 1105 satisfactory, particularly if the recipient does not provide feedback 1106 as to whether it supports the extension or not. This can lead to 1107 operational security issues that are difficult to detect and correct, 1108 as noted in Appendix A.2 and "common RADIUS Implementation Issues and 1109 Suggested Fixes" [RFC5080] Section 2.5. 1111 In order to ensure that a recipient supports an extension, a 1112 recipient encountering an unknown extension may be required to 1113 explicitly reject it and to return an error, rather than proceeding. 1114 This can be accomplished via a "Mandatory" bit in a TLV-based 1115 protocol such as L2TP [RFC2661], or a "Require" or "Proxy-Require" 1116 header in a text-based protocol such as SIP [RFC3261] or HTTP 1117 [RFC2616]. 1119 Since a mandatory extension can result in an interoperability failure 1120 when communicating with a party that does not support the extension, 1121 this designation may not be permitted for vendor-specific extensions, 1122 and may only be allowed for standards-track extensions. To enable 1123 fallback operation with degraded functionality, it is good practice 1124 for the recipient to indicate the reason for the failure, including a 1125 list of unsupported extensions. The initiator can then retry without 1126 the offending extensions. 1128 Typically only the recipient will find itself in the position of 1129 rejecting a mandatory extension, since the initiator can explicitly 1130 indicate which extensions are supported, with the recipient choosing 1131 from among the supported extensions. This can be accomplished via an 1132 exchange of TLVs, such as in IKEv2 [RFC5996] or Diameter [RFC3588], 1133 or via use of "Accept", "Accept-Encoding", "Accept-Language", "Allow" 1134 and "Supported" headers in a text-based protocol such as SIP 1135 [RFC3261] or HTTP [RFC2616]. 1137 5. Security Considerations 1139 An extension must not introduce new security risks without also 1140 providing adequate counter-measures, and in particular it must not 1141 inadvertently defeat security measures in the unextended protocol. 1142 Thus, the security analysis for an extension needs to be as thorough 1143 as for the original protocol - effectively it needs to be a 1144 regression analysis to check that the extension doesn't inadvertently 1145 invalidate the original security model. 1147 This analysis may be simple (e.g. adding an extra opaque data element 1148 is unlikely to create a new risk) or quite complex (e.g. adding a 1149 handshake to a previously stateless protocol may create a completely 1150 new opportunity for an attacker). 1152 When the extensibility of a design includes allowing for new and 1153 presumably more powerful cryptographic algorithms to be added, 1154 particular care is needed to ensure that the result is in fact 1155 increased security. For example, it may be undesirable from a 1156 security viewpoint to allow negotiation down to an older, less secure 1157 algorithm. 1159 6. IANA Considerations 1161 [RFC Editor: please remove this section prior to publication.] 1163 This document has no IANA Actions. 1165 7. References 1167 7.1. Normative References 1169 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1170 Requirement Levels", BCP 14, RFC 2119, March 1997. 1172 [RFC4775] Bradner, S., Carpenter, B., and T. Narten, "Procedures for 1173 Protocol Extensions and Variations", BCP 125, RFC 4775, 1174 December 2006. 1176 [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA 1177 Considerations Section in RFCs", BCP 26, RFC 5226, May 2008. 1179 7.2. Informative References 1181 [IEEE-802.1X] 1182 Institute of Electrical and Electronics Engineers, "Local and 1183 Metropolitan Area Networks: Port-Based Network Access 1184 Control", IEEE Standard 802.1X-2004, December 2004. 1186 [PEAP] Palekar, A., Simon, D., Salowey, J., Zhou, H., Zorn, G. and 1187 S. Josefsson, "Protected EAP Protocol (PEAP) Version 2", 1188 draft-josefsson-pppext-eap-tls-eap-10.txt, Expired Internet 1189 draft (work in progress), October 2004. 1191 [RFC0822] Crocker, D., "Standard for the format of ARPA Internet text 1192 messages", STD 11, RFC 822, August 1982. 1194 [RFC1263] O'Malley, S. and L. Peterson, "TCP Extensions Considered 1195 Harmful", RFC 1263, October 1991. 1197 [RFC1341] Freed, N. and N. Borenstein, "MIME (Multipurpose Internet 1198 Mail Extensions): Mechanisms for Specifying and Describing 1199 the Format of Internet Message Bodies", RFC 1341, June 1992. 1201 [RFC1521] Borenstein, N. and N. Freed, "MIME (Multipurpose Internet 1202 Mail Extensions) Part One: Mechanisms for Specifying and 1203 Describing the Format of Internet Message Bodies", RFC 1521, 1204 September 1993. 1206 [RFC2132] Alexander, S. and R. Droms, "DHCP Options and BOOTP Vendor 1207 Extensions", RFC 2132, March 1997. 1209 [RFC2246] Dierks, T. and C. Allen, "The TLS Protocol Version 1.0", RFC 1210 2246, January 1999. 1212 [RFC2284] Blunk, L. and J. Vollbrecht, "PPP Extensible Authentication 1213 Protocol (EAP)", RFC 2284, March 1998. 1215 [RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black, "Definition 1216 of the Differentiated Services Field (DS Field) in the IPv4 1217 and IPv6 Headers", RFC 2474, December 1998. 1219 [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, 1220 L., Leach, P., and T. Berners-Lee, "Hypertext Transfer 1221 Protocol -- HTTP/1.1", RFC 2616, June 1999. 1223 [RFC2661] Townsley, W., Valencia, A., Rubens, A., Pall, G., Zorn, G., 1224 and B. Palter, "Layer Two Tunneling Protocol "L2TP"", RFC 1225 2661, August 1999. 1227 [RFC2671] Vixie, P., "Extension Mechanisms for DNS (EDNS0)",RFC 2671, 1228 August 1999. 1230 [RFC2822] Resnick, P., "Internet Message Format", RFC 2822, April 2001. 1232 [RFC2865] Rigney, C., Willens, S., Rubens, A., and W. Simpson, "Remote 1233 Authentication Dial In User Service (RADIUS)", RFC 2865, June 1234 2000. 1236 [RFC3095] Bormann, C., Burmeister, C., Degermark, M., Fukushima, H., 1237 Hannu, H., Jonsson, L-E., Hakenberg, R., Koren, T., Le, K., 1238 Liu, Z., Martensson, A., Miyazaki, A., Svanbro, K., Wiebke, 1239 T., Yoshimura, T., and H. Zheng, "RObust Header Compression 1240 (ROHC): Framework and four profiles: RTP, UDP, ESP, and 1241 uncompressed", RFC 3095, July 2001. 1243 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., 1244 Peterson, J., Sparks, R., Handley, M. and E. Schooler, "SIP: 1245 Session Initiation Protocol", RFC 3261, June 2002. 1247 [RFC3427] Mankin, A., Bradner, S., Mahy, R., Willis, D., Ott, J., and 1248 B. Rosen, "Change Process for the Session Initiation Protocol 1249 (SIP)", BCP 67, RFC 3427, December 2002. 1251 [RFC3575] Aboba, B., "IANA Considerations for RADIUS (Remote 1252 Authentication Dial In User Service)", RFC 3575, July 2003. 1254 [RFC3588] Calhoun, P., Loughney, J., Guttman, E., Zorn, G. and J. 1255 Arkko, "Diameter Base Protocol", RFC 3588, September 2003. 1257 [RFC3597] Gustafsson, A., "Handling of Unknown DNS Resource Record (RR) 1258 Types", RFC 3597, September 2003. 1260 [RFC3692] Narten, T., "Assigning Experimental and Testing Numbers 1261 Considered Useful", BCP 82, RFC 3692, January 2004. 1263 [RFC3735] Hollenbeck, S., "Guidelines for Extending the Extensible 1264 Provisioning Protocol (EPP)", RFC 3735, March 2004. 1266 [RFC3748] Aboba, B., Blunk, L., Vollbrecht, J., Carlson, J. and H. 1267 Lefkowetz, "Extensible Authentication Protocol (EAP)", RFC 1268 3748, June 2004. 1270 [RFC3935] Alvestrand, H., "A Mission Statement for the IETF", RFC 3935, 1271 October 2004. 1273 [RFC4001] Daniele, M., Haberman, B., Routhier, S., and J. 1274 Schoenwaelder, "Textual Conventions for Internet Network 1275 Addresses", RFC 4001, February 2005. 1277 [RFC4181] Heard, C., "Guidelines for Authors and Reviewers of MIB 1278 Documents", BCP 111, RFC 4181, September 2005. 1280 [RFC4366] Blake-Wilson, S., Nystrom, M., Hopwood, D., Mikkelsen, J., 1281 and T. Wright, "Transport Layer Security (TLS) Extensions", 1282 RFC 4366, April 2006. 1284 [RFC4485] Rosenberg, J. and H. Schulzrinne, "Guidelines for Authors of 1285 Extensions to the Session Initiation Protocol (SIP)", RFC 1286 4485, May 2006. 1288 [RFC4521] Zeilenga, K., "Considerations for Lightweight Directory 1289 Access Protocol (LDAP) Extensions", BCP 118, RFC 4521, June 1290 2006. 1292 [RFC4727] Fenner, B., "Experimental Values In IPv4, IPv6, ICMPv4, 1293 ICMPv6, UDP, and TCP Headers", RFC 4727, November 2006. 1295 [RFC4929] Andersson, L. and A. Farrel, "Change Process for 1296 Multiprotocol Label Switching (MPLS) and Generalized MPLS 1297 (GMPLS) Protocols and Procedures", BCP 129, RFC 4929, June 1298 2007. 1300 [RFC4962] Housley, R. and B. Aboba, "Guidance for Authentication, 1301 Authorization, and Accounting (AAA) Key Management", BCP 132, 1302 RFC 4962, July 2007. 1304 [RFC5080] Nelson, D. and A. DeKok, "Common Remote Authentication Dial 1305 In User Service (RADIUS) Implementation Issues and Suggested 1306 Fixes", RFC 5080, December 2007. 1308 [RFC5218] Thaler, D., and B. Aboba, "What Makes for a Successful 1309 Protocol?", RFC 5218, July 2008. 1311 [RFC5225] Pelletier, G. and K. Sandlund, "RObust Header Compression 1312 Version 2 (ROHCv2): Profiles for RTP, UDP, IP, ESP and UDP- 1313 Lite", RFC 5225, April 2008. 1315 [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security 1316 (TLS) Protocol Version 1.2", RFC 5246, August 2008. 1318 [RFC5321] Klensin, J., "Simple Mail Transfer Protocol", RFC 5321, 1319 October 2008. 1321 [RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines 1322 for Application Designers", RFC 5405 (BCP 145), November 1323 2008. 1325 [RFC5704] Bryant, S. and M. Morrow, "Uncoordinated Protocol Development 1326 Considered Harmful", RFC 5704, November 2009. 1328 [RFC5727] Peterson, J., Jennings, C. and R. Sparks, "Change Process for 1329 the Session Initiation Protocol (SIP) and the Real-time 1330 Applications and Infrastructure Area", BCP 67, RFC 5727, 1331 March 2010. 1333 [RFC5996] Kaufman, C., Hoffman, P., Nir, Y. and P. Eronen, "Internet 1334 Key Exchange Protocol Version 2 (IKEv2)", RFC 5996, September 1335 2010. 1337 [RFC6055] Thaler, D., Klensin, J. and S. Cheshire, "IAB Thoughts on 1338 Encodings for Internationalized Domain Names", RFC 6055, 1339 February 2011. 1341 [RFC6158] DeKok, A. and G. Weber, "RADIUS Design Guidelines", BCP 158, 1342 RFC 6158, March 2011. 1344 [Transitive] 1345 Scudder, J. and E. Chen, "Error Handling for Optional 1346 Transitive BGP Attributes", Internet draft (work in 1347 progress), draft-ietf-idr-optional-transitive-03, September, 1348 2010. 1350 Acknowledgments 1352 This document is heavily based on an earlier draft under a different 1353 title by Scott Bradner and Thomas Narten. 1355 That draft stated: The initial version of this document was put 1356 together by the IESG in 2002. Since then, it has been reworked in 1357 response to feedback from John Loughney, Henrik Levkowetz, Mark 1358 Townsley, Randy Bush and others. 1360 Valuable comments and suggestions on the current form of the document 1361 were made by Loa Andersson, Jari Arkko, Leslie Daigle, Phillip 1362 Hallam-Baker, Ted Hardie, Alfred Hoenes, John Klensin, Danny 1363 McPherson, Eric Rescorla, Adam Roach and Pekka Savola. 1365 The text on TLS experience was contributed by Yngve Pettersen. 1367 IAB Members at the Time of Approval 1369 Bernard Aboba 1370 Ross Callon 1371 Alissa Cooper 1372 Spencer Dawkins 1373 Joel Halpern 1374 Russ Housley 1375 David Kessens 1376 Olaf Kolkman 1377 Danny McPherson 1378 Jon Peterson 1379 Andrei Robachevsky 1380 Dave Thaler 1381 Hannes Tschofenig 1383 Appendix A. Examples 1385 This section discusses some specific examples, as case studies. 1387 A.1. Already documented cases 1389 There are certain documents that specify a change process or describe 1390 extension considerations for specific IETF protocols: 1392 The SIP change process [RFC3427], [RFC4485], [RFC5727] 1393 The (G)MPLS change process (mainly procedural) [RFC4929] 1394 LDAP extensions [RFC4521] 1395 EPP extensions [RFC3735] 1396 DNS extensions [RFC2671][RFC3597] 1397 SMTP extensions [RFC5321] 1399 It is relatively common for MIBs, which are all in effect extensions 1400 of the SMI data model, to be defined or extended outside the IETF. 1401 BCP 111 [RFC4181] offers detailed guidance for authors and reviewers. 1403 A.2. RADIUS Extensions 1405 The RADIUS [RFC2865] protocol was designed to be extensible via 1406 addition of Attributes to a Data Dictionary on the server, without 1407 requiring code changes. However, this extensibility model assumed 1408 that Attributes would conform to a limited set of data types and that 1409 vendor extensions would be limited to use by vendors, in situations 1410 in which interoperability was not required. Subsequent developments 1411 have stretched those assumptions. 1413 Section 6.2 of the RADIUS specification [RFC2865] defines a mechanism 1414 for Vendor-Specific extensions (Attribute 26), and states that use of 1415 Vendor-Specific extensions: 1417 should be encouraged instead of allocation of global attribute 1418 types, for functions specific only to one vendor's implementation 1419 of RADIUS, where no interoperability is deemed useful. 1421 However, in practice usage of Vendor-Specific Attributes (VSAs) has 1422 been considerably broader than this. In particular, VSAs have been 1423 used by Standards Development Organizations (SDOs) to define their 1424 own extensions to the RADIUS protocol. 1426 This has caused a number of problems. Since the VSA mechanism was 1427 not designed for interoperability, VSAs do not contain a "mandatory" 1428 bit. As a result, RADIUS clients and servers may not know whether it 1429 is safe to ignore unknown attributes. For example, Section 5 of the 1430 RADIUS specification [RFC2865] states: 1432 A RADIUS server MAY ignore Attributes with an unknown Type. A 1433 RADIUS client MAY ignore Attributes with an unknown Type. 1435 However, in the case where the VSAs pertain to security (e.g. 1436 Filters) it may not be safe to ignore them, since the RADIUS 1437 specification [RFC2865] also states: 1439 A NAS that does not implement a given service MUST NOT implement 1440 the RADIUS attributes for that service. For example, a NAS that 1441 is unable to offer ARAP service MUST NOT implement the RADIUS 1442 attributes for ARAP. A NAS MUST treat a RADIUS access-accept 1443 authorizing an unavailable service as an access-reject instead." 1445 Detailed discussion of the issues arising from this can be found in 1446 "Common Remote Authentication Dial In User Service (RADIUS) 1447 Implementation Issues and Suggested Fixes" [RFC5080] Section 2.5. 1449 Since it was not envisaged that multi-vendor VSA implementations 1450 would need to interoperate, the RADIUS specification [RFC2865] does 1451 not define the data model for VSAs, and allows multiple sub- 1452 attributes to be included within a single Attribute of type 26. 1453 However, this enables VSAs to be defined which would not be 1454 supportable by current implementations if placed within the standard 1455 RADIUS attribute space. This has caused problems in standardizing 1456 widely deployed VSAs, as discussed in "RADIUS Design Guidelines" 1457 BCP 158 [RFC6158]. 1459 In addition to extending RADIUS by use of VSAs, SDOs have also 1460 defined new values of the Service-Type attribute in order to create 1461 new RADIUS commands. Since the RADIUS specification [RFC2865] 1462 defined Service-Type values as being allocated First Come, First 1463 Served (FCFS), this essentially enabled new RADIUS commands to be 1464 allocated without IETF review. This oversight has since been fixed 1465 in "IANA Considerations for RADIUS" [RFC3575]. 1467 A.3. TLS Extensions 1469 The Secure Sockets Layer (SSL) v2 protocol was developed by Netscape 1470 to be used to secure online transactions on the Internet. It was 1471 later replaced by SSL v3, also developed by Netscape. SSL v3 was 1472 then further developed by the IETF as the Transport Layer Security 1473 (TLS) 1.0 [RFC2246]. 1475 The SSL v3 protocol was not explicitly specified to be extended. 1476 Even TLS 1.0 did not define an extension mechanism explicitly. 1477 However, extension "loopholes" were available. Extension mechanisms 1478 were finally defined in "Transport Layer Security (TLS) Extensions" 1479 [RFC4366]: 1481 o New versions 1482 o New cipher suites 1483 o Compression 1484 o Expanded handshake messages 1485 o New record types 1486 o New handshake messages 1488 The protocol also defines how implementations should handle unknown 1489 extensions. 1491 Of the above extension methods, new versions and expanded handshake 1492 messages have caused the most interoperability problems. 1493 Implementations are supposed to ignore unknown record types but to 1494 reject unknown handshake messages. 1496 The new version support in SSL/TLS includes a capability to define 1497 new versions of the protocol, while allowing newer implementations to 1498 communicate with older implementations. As part of this 1499 functionality some Key Exchange methods include functionality to 1500 prevent version rollback attacks. 1502 The experience with this upgrade functionality in SSL and TLS is 1503 decidedly mixed: 1505 o SSL v2 and SSL v3/TLS are not compatible. It is possible to use 1506 SSL v2 protocol messages to initiate a SSL v3/TLS connection, but 1507 it is not possible to communicate with a SSL v2 implementation 1508 using SSL v3/TLS protocol messages. 1509 o There are implementations that refuse to accept handshakes using 1510 newer versions of the protocol than they support. 1511 o There are other implementations that accept newer versions, but 1512 have implemented the version rollback protection clumsily. 1514 The SSL v2 problem has forced SSL v3 and TLS clients to continue to 1515 use SSL v2 Client Hellos for their initial handshake with almost all 1516 servers until 2006, much longer than would have been desirable, in 1517 order to interoperate with old servers. 1519 The problem with incorrect handling of newer versions has also forced 1520 many clients to actually disable the newer protocol versions, either 1521 by default, or by automatically disabling the functionality, to be 1522 able to connect to such servers. Effectively, this means that the 1523 version rollback protection in SSL and TLS is non-existent if talking 1524 to a fatally compromised older version. 1526 SSL v3 and TLS also permitted expansion of the Client Hello and 1527 Server Hello handshake messages. This functionality was fully 1528 defined by the introduction of TLS Extensions, which makes it 1529 possible to add new functionality to the handshake, such as the name 1530 of the server the client is connecting to, request certificate status 1531 information, indicate Certificate Authority support, maximum record 1532 length, etc. Several of these extensions also introduce new 1533 handshake messages. 1535 It has turned out that many SSL v3 and TLS implementations that do 1536 not support TLS Extensions, did not, as required by the protocol 1537 specifications, ignore the unknown extensions, but instead failed to 1538 establish connections. Several of the implementations behaving in 1539 this manner are used by high profile Internet sites, such as online 1540 banking sites, and this has caused a significant delay in the 1541 deployment of clients supporting TLS Extensions, and several of the 1542 clients that have enabled support are using heuristics that allow 1543 them to disable the functionality when they detect a problem. 1545 Looking forward, the protocol version problem, in particular, can 1546 cause future security problems for the TLS protocol. The strength of 1547 the digest algorithms (MD5 and SHA-1) used by SSL and TLS is 1548 weakening. If MD5 and SHA-1 weaken to the point where it is feasible 1549 to mount successful attacks against older SSL and TLS versions, the 1550 current error recovery used by clients would become a security 1551 vulnerability (among many other serious problems for the Internet). 1553 To address this issue, TLS 1.2 [RFC5246] makes use of a newer 1554 cryptographic hash algorithm (SHA-256) during the TLS handshake by 1555 default. Legacy ciphersuites can still be used to protect 1556 application data, but new ciphersuites are specified for data 1557 protection as well as for authentication within the TLS handshake. 1558 The hashing method can also be negotiated via a Hello extension. 1559 Implementations are encouraged to implement new ciphersuites, and to 1560 enable the negotiation of the ciphersuite used during a TLS session 1561 to be governed by policy, thus enabling a more rapid transition away 1562 from weakened ciphersuites. 1564 The lesson to be drawn from this experience is that it isn't 1565 sufficient to design extensibility carefully; it must also be 1566 implemented carefully by every implementer, without exception. Test 1567 suites and certification programs can help provide incentives for 1568 implementers to pay attention to implementing extensibility 1569 mechanisms correctly. 1571 A.4. L2TP Extensions 1573 Layer Two Tunneling Protocol (L2TP) [RFC2661] carries Attribute-Value 1574 Pairs (AVPs), with most AVPs having no semantics to the L2TP protocol 1575 itself. However, it should be noted that L2TP message types are 1576 identified by a Message Type AVP (Attribute Type 0) with specific AVP 1577 values indicating the actual message type. Thus, extensions relating 1578 to Message Type AVPs would likely be considered major extensions. 1580 L2TP also provides for Vendor-Specific AVPs. Because everything in 1581 L2TP is encoded using AVPs, it would be easy to define vendor- 1582 specific AVPs that would be considered major extensions. 1584 L2TP also provides for a "mandatory" bit in AVPs. Recipients of L2TP 1585 messages containing AVPs they do not understand but that have the 1586 mandatory bit set, are expected to reject the message and terminate 1587 the tunnel or session the message refers to. This leads to 1588 interesting interoperability issues, because a sender can include a 1589 vendor-specific AVP with the M-bit set, which then causes the 1590 recipient to not interoperate with the sender. This sort of behavior 1591 is counter to the IETF ideals, as implementations of the IETF 1592 standard should interoperate successfully with other implementations 1593 and not require the implementation of non-IETF extensions in order to 1594 interoperate successfully. Section 4.2 of the L2TP specification 1595 [RFC2661] includes specific wording on this point, though there was 1596 significant debate at the time as to whether such language was by 1597 itself sufficient. 1599 Fortunately, it does not appear that the potential problems described 1600 above have yet become a problem in practice. At the time of this 1601 writing, the authors are not aware of the existence of any vendor- 1602 specific AVPs that also set the M-bit. 1604 Change log [RFC Editor: please remove this section] 1606 draft-iab-extension-recs-08: 2011-10-22. Incorporated additional 1607 resolutions to issues raised during Call for Comment. 1609 draft-iab-extension-recs-07: 2011-7-24. Incorporated resolutions to 1610 issues raised during Call for Comment. 1612 draft-iab-extension-recs-06: 2011-3-1. Incorporated additional 1613 corrections and organizational improvements. 1615 draft-iab-extension-recs-05: 2011-2-4. Added to the security 1616 considerations section. 1618 draft-iab-extension-recs-04: 2011-2-1. Added a section on 1619 cryptographic agility. Additional reorganization. 1621 draft-iab-extension-recs-03: 2011-1-25. Updates and reorganization 1622 to reflect comments from the IETF community. 1624 draft-iab-extension-recs-02: 2010-7-12. Updates by Bernard Aboba 1625 draft-iab-extension-recs-01: 2010-4-7. Updates by Stuart 1626 Cheshire. 1628 draft-iab-extension-recs-00: 2009-4-24. Updated boilerplate, 1629 author list. 1631 draft-carpenter-extension-recs-04: 2008-10-24. Updated author 1632 addresses, fixed editorial issues. 1634 draft-carpenter-extension-recs-03: 2008-10-17. Updated references, 1635 added material relating to versioning. 1637 draft-carpenter-extension-recs-02: 2007-06-15. Reorganized Sections 1638 2 and 3. 1640 draft-carpenter-extension-recs-01: 2007-03-04. Updated according to 1641 comments, especially the wording about TLS, added various specific 1642 examples. 1644 draft-carpenter-extension-recs-00: original version, 2006-10-12. 1645 Derived from draft-iesg-vendor-extensions-02.txt dated 2004-06-04 by 1646 focusing on architectural issues; the more procedural issues in that 1647 draft were moved to RFC 4775. 1649 Authors' Addresses 1651 Brian Carpenter 1652 Department of Computer Science 1653 University of Auckland 1654 PB 92019 1655 Auckland, 1142 1656 New Zealand 1658 Email: brian.e.carpenter@gmail.com 1660 Bernard Aboba 1661 Microsoft Corporation 1662 One Microsoft Way 1663 Redmond, WA 98052 1665 EMail: bernard_aboba@hotmail.com 1667 Stuart Cheshire 1668 Apple Computer, Inc. 1669 1 Infinite Loop 1670 Cupertino, CA 95014 1672 EMail: cheshire@apple.com