idnits 2.17.1 draft-briscoe-intarea-ipv4-id-reuse-00.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 : ---------------------------------------------------------------------------- -- The draft header indicates that this document updates RFC2003, but the abstract doesn't seem to mention this, which it should. -- The draft header indicates that this document updates RFC2780, but the abstract doesn't seem to mention this, which it should. -- The draft header indicates that this document updates RFC4301, but the abstract doesn't seem to mention this, which it should. -- The draft header indicates that this document updates RFC791, but the abstract doesn't seem to mention this, which it should. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 845 has weird spacing: '...ificant bit: ...' (Using the creation date from RFC791, updated by this document, for RFC5378 checks: 1981-09-01) -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (March 7, 2011) is 4798 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) == Unused Reference: 'Fransson04' is defined on line 972, but no explicit reference was found in the text Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 6 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Internet Area Working Group B. Briscoe 3 Internet-Draft BT 4 Updates: 791, 2003, 2780, 4301, March 7, 2011 5 4727, ietf-intarea-ipv4-id-update 6 (if approved) 7 Intended status: Standards Track 8 Expires: September 8, 2011 10 Reusing the IPv4 Identification Field in Atomic Packets 11 draft-briscoe-intarea-ipv4-id-reuse-00 13 Abstract 15 This specification takes a new approach to extensibility that is both 16 principled and a hack. It builds on recent moves to formalise the 17 increasingly common practice where fragmentation in IPv4 more closely 18 matches that of IPv6. The large majority of IPv4 packets are now 19 'atomic', meaning indivisible. In such packets, the 16 bits of the 20 IPv4 Identification (IPv4 ID) field are redundant and could be freed 21 up for the Internet community to put to other uses, at least within 22 the constraints imposed by their original use for reassembly. This 23 specification defines the process for redefining the semantics of 24 these bits. It uses the previously reserved control flag in the IPv4 25 header to indicate that these 16 bits have new semantics. Great care 26 is taken throughout to ease incremental deployment, even in the 27 presence of middleboxes that incorrectly discard or normalise packets 28 that have the reserved control flag set. 30 Status of This Memo 32 This Internet-Draft is submitted in full conformance with the 33 provisions of BCP 78 and BCP 79. 35 Internet-Drafts are working documents of the Internet Engineering 36 Task Force (IETF). Note that other groups may also distribute 37 working documents as Internet-Drafts. The list of current Internet- 38 Drafts is at http://datatracker.ietf.org/drafts/current/. 40 Internet-Drafts are draft documents valid for a maximum of six months 41 and may be updated, replaced, or obsoleted by other documents at any 42 time. It is inappropriate to use Internet-Drafts as reference 43 material or to cite them other than as "work in progress." 45 This Internet-Draft will expire on September 8, 2011. 47 Copyright Notice 48 Copyright (c) 2011 IETF Trust and the persons identified as the 49 document authors. All rights reserved. 51 This document is subject to BCP 78 and the IETF Trust's Legal 52 Provisions Relating to IETF Documents 53 (http://trustee.ietf.org/license-info) in effect on the date of 54 publication of this document. Please review these documents 55 carefully, as they describe your rights and restrictions with respect 56 to this document. Code Components extracted from this document must 57 include Simplified BSD License text as described in Section 4.e of 58 the Trust Legal Provisions and are provided without warranty as 59 described in the Simplified BSD License. 61 Table of Contents 63 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 64 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 5 65 3. IPv4 Wire Protocol Semantics for Reusing the 66 Identification Field . . . . . . . . . . . . . . . . . . . . . 5 67 4. Behaviour of Intermediate Nodes . . . . . . . . . . . . . . . 8 68 4.1. End-to-End Preservation of ID-Reuse Semantics . . . . . . 8 69 4.2. Tunnel Behaviour . . . . . . . . . . . . . . . . . . . . . 8 70 5. Process for Defining Subdivisions of the ID-Reuse Field . . . 9 71 5.1. Constraints on Uses of the ID-Reuse Field . . . . . . . . 10 72 5.2. Process Example . . . . . . . . . . . . . . . . . . . . . 11 73 6. Incremental Deployment of New Uses of the IPv4 ID Field . . . 13 74 6.2. Process for Using the ID-Reuse Field Without Requiring 75 RC=1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 76 7. Updates to Existing RFCs . . . . . . . . . . . . . . . . . . . 17 77 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 18 78 9. Security Considerations . . . . . . . . . . . . . . . . . . . 19 79 10. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 20 80 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 20 81 12. Outstanding Issues (to be removed when all resolved) . . . . . 21 82 13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 21 83 13.1. Normative References . . . . . . . . . . . . . . . . . . . 21 84 13.2. Informative References . . . . . . . . . . . . . . . . . . 21 85 Appendix A. Why More Bits Cannot be Freed (To be Removed by 86 RFC Editor) . . . . . . . . . . . . . . . . . . . . . 22 87 Appendix B. Experimental or Standards Track? (To Be Removed 88 Before Publication) . . . . . . . . . . . . . . . . . 23 90 Intended status: Standards Track? (to be removed before publication) 92 This draft defines a process and a protocol for enabling new 93 protocols, including their progression from experimental track to 94 standards track. A process specification cannot have lesser status 95 than the protocols it enables. So if this specification were to 96 start on the experimental track, it would not initially have 97 sufficient status to enable standards track protocols. 99 In order for the IETF to consider whether this draft itself should be 100 experimental or standards track, it has been written as if it is 101 intended for the standards track. Otherwise the parts of the process 102 for enabling standards track protocols would have had to have been 103 written hypothetically, which would have been highly confusing. If 104 the IETF decides this specification ought to start out on the 105 experimental track, the standards track parts of the process will 106 have to be edited out. 108 Appendix B discusses whether this draft itself would be better to 109 start as experimental or standards track. 111 1. Introduction 113 The Problem: The extensibility provisions in IP (v4 and v6) have 114 turned out not to be usable in practice. Hardware has been optimised 115 for the common case, so packets using extensibility mechanisms (e.g. 116 IPv4 options or IPv6 hop-by-hop options) are very likely to be punted 117 to the software slow-path and consequently likely to be dropped 118 whenever the software processor is busy [Fransson04, Cisco.IPv6Ext]. 120 This specification takes a different approach to extensibility. 121 Rather than flagging protocol extensions as 'extensions', it places 122 extension headers where they will be ignored by pre-existing 123 hardware. As code is added to routers to handle newly added 124 extensions, the code can tell the machine where to look for the 125 relevant header. 127 This approach recognises that extensions added after a protocol suite 128 was first defined are different to options defined as a coherent part 129 of the original protocol suite. Machines that have no code to 130 understand a protocol extension that was added later do not need to 131 punt a packet to the software processor merely to scan through chains 132 of headers that it will not know how to process. 134 Having only settled on this approach long after the TCP/IP suite has 135 been defined, it becomes necessary to find places in the existing 136 protocol headers that are already ignored by existing machines. In 137 an 'atomic' IPv4 packet, the Identification (IPv4 ID) field is one 138 such place that is redundant. This specification defines the process 139 through which the 16 bits in this field can be returned to the IETF 140 for use in future standards actions, at least within the constraints 141 imposed by their original use for reassembly. 143 Background: [ipv4-id-update] proposes to update IPv4 to more closely 144 match the approach taken to fragmentation in IPv6. It recommends 145 that IPv4 sources send 'atomic' packets whenever possible. An atomic 146 packet is one that has not yet been fragmented (MF=0 and fragment 147 offset=0) and for which further fragmentation is inhibited (DF=1) 148 [ipv4-id-update]. If fragmentation is necessary, it is only 149 permitted at devices that control the uniqueness of the IP ID field, 150 e.g., sources, tunnel ingresses (for the outer header), and the 151 public side of NATs. 153 In practical scenarios, the IPv4 ID field is too small to guarantee 154 uniqueness during the lifetime of a packet anyway [RFC4963]. 155 Therefore it has become safer to disable fragmentation altogether and 156 instead use an approach such as packetization layer path MTU 157 discovery [RFC4821]. The large majority of IPv4 packets are now 158 atomic. 160 Approach: This specification defines the IPv4 control flag that was 161 previously reserved [RFC0791] as the Recycled flag (RC). An 162 implementation can set RC=1 in an atomic packet to unambiguously flag 163 that the IPv4 ID field is not to be interpreted as IP Identification, 164 but instead it has the alternative semantics of an ID-Reuse field. 165 By setting RC=1, IPv4 implementations can distinguish a value 166 deliberately written into the ID-Reuse field from the same value that 167 just happened to be written into the IP ID field of an atomic packet 168 by a pre-existing implementation. 170 Thus, this specification effectively uses up the last bit in the IPv4 171 header in order to free up 16 other bits. However, there are some 172 constraints on the use of these 16 bits due to their original use as 173 the IP ID field (enumerated in Section 5.1). Of course the main 174 constraint it that the bits are not available in non-atomic packets. 175 But fragmentation is now used only rarely anyway, so it makes sense 176 to see if the the Internet community can invent ways to use the 16 177 bits in the IPv4 ID field despite the constraints. 179 Frequently Asked Questions: 181 1. There are many cases where a non-compliant machine ignores Don't 182 Fragment (DF=1) and fragments a packet anyway. 184 One answer is that we cannot allow non-complaint behaviour to 185 always block progress. Another answer is that we may be able to 186 detect and circumvent such non-compliant behaviour. For 187 instance, if a non-compliant router fragments packets with DF=1, 188 it may be possible to enhance path maximum transmission unit 189 discovery (PMTUD) to find a lower segment size small enough to 190 prevent the offending box from fragmenting packets. 192 2. {ToDo} 194 Document Roadmap: Section 3 defines the semantics of the updated IPv4 195 wire protocol and Section 4 defines intermediate node behaviour. 196 Section 5 defines the process to be used for reassigning sub-fields 197 of the IPv4 ID-Reuse field. Then Section 6 describes a way to 198 circumvent problems likely to arise when deploying this new protocol. 199 Finally, Section 7 enumerates the updates to pre-existing RFCs, 200 before the tailpiece sections considering IANA, Security and draw 201 conclusions. 203 2. Terminology 205 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 206 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 207 document are to be interpreted as described in RFC 2119 [RFC2119]. 209 Further terminology used within this document: 211 Atomic packet: A packet not yet having been fragmented (MF=0 and 212 fragment offset=0) and for which further fragmentation has been 213 inhibited (DF=1), or in the syntax of the C programming language 214 ((DF==1) && (MF==0) && (Offset==0)) [ipv4-id-update]. 216 Recycled (RC) flag: The control flag that was 'reserved' in 217 [RFC0791] (Figure 1). The flag positioned at bit 48 of the IPv4 218 header (counting from 0). Alternatively, some would call this bit 219 0 (counting from 0) of octet 7 (counting from 1) of the IPv4 220 header. 222 ID-Reuse field: Octets 5 and 6 (counting from 1) of the IPv4 header 223 of an atomic packet (Figure 3). The field that would have been 224 the IP Identification field if the packet were not atomic. 226 3. IPv4 Wire Protocol Semantics for Reusing the Identification Field 228 This specification defines the control flag that was defined as 229 'reserved' in [RFC0791] as the Recycled (RC) flag (Figure 1). 231 0 1 2 232 +---+---+---+ 233 | R | D | M | 234 | C | F | F | 235 +---+---+---+ 237 The Recycled (RC) Flag was previously reserved. 239 Figure 1: The Control Flags at the Start of Byte 7 of the IPv4 Header 241 Figure 2 recaps the definitions of octets 5 to 8 (counting from 1) of 242 the IPv4 header [RFC0791]. 243 0 1 2 3 244 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 245 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 246 | Identification |Flags| Fragment Offset | 247 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 249 Figure 2: Recap of RFC791 Definition of Octets 5 to 8 of the IPv4 250 Header. 252 If an IPv4 implementation sets RC=1 on an atomic packet, octets 5 & 6 253 of the IPv4 header MUST be interpreted with the semantics of the ID- 254 Reuse field, and MUST NOT be interpreted as the Identification field. 255 Figure 3 shows how octets 5 & 6 are redefined as the ID-Reuse field 256 when the packet is atomic, in the case where RC=1. 257 0 1 2 3 258 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 259 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 260 | ID-Reuse |1 1 0|0 0 0 0 0 0 0 0 0 0 0 0 0| 261 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 263 The Identification Field is redefined as the ID-Reuse Field when the 264 Packet is Atomic and specifically when RC=1 266 Figure 3: Octets 5 to 8 of the IPv4 Header. 268 If the Recycled flag is cleared to RC=0 on an atomic packet, some 269 sub-fields of octets 5 & 6 of the IPv4 header MAY be interpreted with 270 the semantics of the ID-Reuse field, but only in the highly 271 constrained circumstances defined in Section 6.2. 273 For the avoidance of doubt, the Recycled flag alone MUST NOT be 274 assumed to indicate that the packet is atomic. Only the combination 275 of ((DF==1) && (MF==0) && (Offset==0)) indicates that a packet is 276 atomic. Then if the Recycled flag is also set, the ID field 277 unambiguously has the semantics of the ID-Reuse field. If the 278 Recycled flag of an atomic packet is cleared, its ID field only has 279 the semantics of the ID-Reuse field in specific limited 280 circumstances. 282 It is expected that proposals to use the ID-Reuse field will each 283 need a few bits, not the whole 16 bit field. Therefore this 284 specification establishes a new IANA registry (Section 8) to record 285 assignments of sub-divisions of the ID-Reuse field. In this way, it 286 will be possible for new uses of different sub-divisions to be 287 orthogonal to each other. The process for incrementally defining new 288 sub-divisions is specified in Section 5. 290 If an IPv4 packet header has RC=1 but it is not atomic ((DF==0) || 291 (MF==1) || (Offset !=0)), then all the fields of the IPv4 header are 292 undefined and reserved for future use. If an implementation receives 293 such a packet, it could imply: 295 o that some currently unknown attack is being attempted 297 o or that some future standards action has defined a meaning for 298 this reserved combination of header values 300 Therefore, if an implementation receives a non-atomic packets with 301 RC=1, it MUST treat the packet as if the Recycled flag were cleared 302 to 0, but it MUST NOT change the Recycled flag to zero. It MAY log 303 the arrival of such packets and/or raise an alarm. It MUST NOT 304 always drop such packets, but it MAY drop them under a policy that 305 can be revoked if it is established that the appearance of such 306 packets is the result of a future standards action. 308 For convenience only, the above rules are summarised in Table 1. The 309 semantics of octets 5 & 6 of the IPv4 header are tabulated for each 310 value of the RC flag (rows) and for whether the packet is atomic or 311 not (columns). 313 +---------+----------------+--------------------+ 314 | RC flag | Non-Atomic | Atomic | 315 +---------+----------------+--------------------+ 316 | 0 | Identification | ID-Reuse (Limited) | 317 | 1 | Undefined | ID-Reuse | 318 +---------+----------------+--------------------+ 320 Table 1: The Dependence of the Semantics of Octets 5 & 6 of the IPv4 321 Header on whether the Packet is Atomic and on the RC Flag 323 4. Behaviour of Intermediate Nodes 325 4.1. End-to-End Preservation of ID-Reuse Semantics 327 If the source sets the RC flag to 1 on an atomic packet, another node 328 MUST NOT clear the RC flag to zero. Otherwise the semantics of the 329 ID-Reuse field would change (see the Security Considerations in 330 Section 9 for discussion of the integrity of the ID-Reuse field). 331 Note that intermediate nodes are already not expected to change an 332 atomic packet to non-atomic, which otherwise would also risk changing 333 the semantics of the ID-Reuse field. 335 If the source zeros the RC flag on an atomic packet, an intermediate 336 node MAY change the RC flag to 1. At this time, no case is envisaged 337 where an intermediate node would need to do this. However, as this 338 behaviour preserves ID-Reuse semantics safely, it is not precluded in 339 case it will prove useful (e.g. for sender proxies). 341 4.2. Tunnel Behaviour 343 This specification does not need to change the following aspects of 344 IPv4-in-IPv4 tunnelling, which already provide the most useful 345 semantics for the ID-Reuse field: 347 o For some time, it has been mandated that an atomic packet "MUST" 348 be encapsulated by an atomic outer header [RFC2003] (although some 349 implementations are broken in this respect). 351 o On decapsulation the outgoing header will naturally propagate the 352 ID-Reuse field of the inner header. 354 However, compliant IPv4 encapsulation implementations SHOULD copy the 355 ID-Reuse field when encapsulating an atomic IPv4 packet in another 356 atomic IPv4 header, irrespective of the setting of the Recycled flag. 357 It would be ideal but impractical to assert 'MUST' in this last 358 clause, given it cannot be assumed that pre-existing IPv4-in-IPv4 359 encapsulators will propagate the ID-Reuse field to the outer header 360 (see Section 5.1). 362 IPv6 packets without a fragmentation extension header are inherently 363 atomic. Therefore, if an IPv4 header encapsulates an IPv6 packet, 364 the encapsulator is already required to set the outer as atomic. 366 There is no direct mapping between the IPv4 ID-Reuse field as a whole 367 and any IPv6 header field (main or extension), because the ID-Reuse 368 field is merely a container for yet-to-be-defined sub-fields. 369 However, sub-fields of the ID-Reuse field might be defined to provide 370 a mapping for IPv6 extension headers that need to be visible in the 371 outer IPv4 header of a tunnel. The present specification cannot say 372 anything in general about any such mappings or any associated tunnel 373 behaviour. Any such behaviour will have to be defined when 374 individual ID-Reuse sub-fields are specified. 376 5. Process for Defining Subdivisions of the ID-Reuse Field 378 When IPv4 was designed, then later IPv6, all the fields in the main 379 IP header were initially defined together in a coordinated fashion. 380 In contrast, the only practical way to define new uses for the bits 381 in the ID-Reuse field will be to adopt a gradual addition approach, 382 in which subsets of the bits or codepoints will have to be assigned 383 on the merits of each request at the time. 385 Each new scheme will need to submit an RFC that requests a 386 subdivision of the ID-Reuse field and assigns behaviours to the 387 codepoints within this subdivision. A specification defining a new 388 use of a subdivision of the ID-Reuse field MUST register this use 389 with the IANA, which will maintain a registry for this purpose 390 (Section 8). 392 Proposals to reuse the IP ID field could relate to other parts of the 393 IPv4 header in the following different ways {ToDo: this list is not 394 exhaustive}: 396 Orthogonal: Some new protocol proposals will need to apply whatever 397 is in the rest of the packet, e.g. whether unicast or multicast, 398 whatever the Diffserv codepoint and whatever else might have been 399 added in the rest of the IP-Reuse field. Schemes that need to be 400 orthogonal to other elements of the IPv4 protocol will require 401 assignment of a number of bits as a dedicated sub-field of the ID- 402 Reuse field. 404 Mutually exclusive: It might be impossible for two uses of the ID- 405 Reuse field to both apply to the same packet. Such mutually 406 exclusive schemes will only each require a range of codepoints 407 within a sub-field. 409 Conditional: Some protocol proposals might only apply when other 410 parts of the header satisfy certain conditions, e.g. only for 411 multicast packets. The IANA will need to register these 412 conditions so that the bits can still be assigned for other uses 413 when the conditions do not apply. 415 To allow interworking between sub-fields that are being defined 416 incrementally, every new protocol MUST assign the all-zeros codepoint 417 of its sub-field to mean the new protocol is 'turned off'. This 418 means that implementations of the new protocol will treat such 419 packets as they would have been treated before the new protocol was 420 defined. 422 Implementations MUST also clear to zero any bits in the ID-Reuse 423 field that are not defined at the time the implementation is coded. 425 Proposals to use sub-fields of ID-Reuse will have to be assessed in 426 the order they arrive without knowing what future proposals they 427 might preclude. To judge each proposal, at least the following 428 criteria will be used: 430 Constraint satisfaction: Each proposal MUST either satisfy all the 431 constraints in Section 5.1 below, or include measures to 432 circumvent them. 434 General usefulness: Proposals that are not applicable to a broad set 435 of services that can be built over the internetwork protocol 436 SHOULD NOT warrant consuming the newly freed up IPv4 header space. 438 Parsimony: Burning up a large proportion of the remaining bits will 439 count against a proposal. 441 Backward compatibility with prior uses of ID-Reuse: As more sub- 442 fields of the ID-Reuse field become defined, each new proposal 443 SHOULD ensure that it takes into account potential interactions 444 with earlier standards actions or experiments defining other sub- 445 fields. 447 Forward compatibility with potential uses of ID-Reuse: In addition, 448 proposals that demonstrate sensitivity to potential future uses of 449 the remaining sub-fields of the ID-Reuse field will be more likely 450 to progress through the IETF's approval process. 452 Do no harm: Proposals that do no harm to existing uses of the 453 Internet will be favoured over those that do more harm. 455 5.1. Constraints on Uses of the ID-Reuse Field 457 Atomic packets: The IPv4 ID field cannot be reused if the packet is 458 not atomic, because then the IP ID field will need to be used for 459 its original purpose: fragment reassembly. 461 IPsec interaction: The IP Authentication Header (AH) [RFC4302] 462 assumes and requires the IPv4 ID field to be immutable, otherwise 463 verification of authentication and integrity checks will fail. 464 Any new use of bits in the ID-Reuse field MUST ensure the bits are 465 immutable, at least between IPsec endpoints (whether transport or 466 tunnel mode). It cannot be assumed that pre-existing IPsec 467 implementations will check the setting of the Recycled flag. 469 Note that the Recycled flag itself is considered mutable and 470 masked out before calculating an authentication header [RFC4302] 471 (see Section 9). 473 Tunnelling: Any new use of the ID-Reuse field in atomic packets 474 cannot reliably assume that the ID-Reuse field will propagate 475 unchanged into the outer header of an IPv4-in-IPv4 tunnel 476 [RFC2003, RFC4301]. It is likely that an IPv4 tunnel ingress will 477 encapsulate an atomic packet with another atomic outer header, as 478 this behaviour was mandated in [RFC2003]. However it is known 479 that some implementations are broken in this respect. It is 480 possible that an IPv4 encapsulator might copy the IP ID field of 481 an arriving atomic packet into the outer header. However this 482 behaviour has never been required and therefore cannot be 483 guaranteed for pre-existing tunnels. 485 Nonetheless, it can be assumed that the IPv4 ID field will be 486 preserved through the inner header into the outgoing packet at the 487 other end of the tunnel (even though this behaviour would not 488 strictly have been necessary for an atomic packet). 490 Incremental deployment: Each new proposal will need to consider any 491 detrimental effects from pre-existing IPv4 implementations, 492 assuming that they are likely to act on atomic packets without 493 first checking on the setting of the Recycled flag. 495 5.2. Process Example 497 For illustration purposes, imagine two RFCs have been published: an 498 experimental track RFC called Experiment A (ExA) and a standards 499 track RFC called Standard B (StB) and . Imagine they define 500 respectively a use for bits bits 14 to 15 and 11 to 13 of the ID- 501 Reuse field. Figure 4 shows example IANA registry entries for these 502 imaginary sub-fields. 504 Protocol name: StB 505 RFC: BBBB 506 Leftmost bit: 11 507 No. of bits allocated: 3 508 Sub-field defined if: Atomic packet and RC=1 510 Protocol name: ExA 511 RFC: AAAA 512 Leftmost bit: 14 513 No. of bits allocated: 2 514 Sub-field defined if: Atomic packet and RC=1 516 Figure 4: Example IANA Registry of Sub-fields of the ID-Reuse Field 518 Figure 5 shows an example of how incremental specification of 519 subdivisions of ID-Reuse would work. 520 0 1 2 3 521 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 522 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 523 | ID-Reuse _____ ___|1 1 0|0 0 0 0 0 0 0 0 0 0 0 0 0| 524 |0 0 0 0 0 0 0 0 0 0 0| StB |ExA| | | 525 | |1 0 1|0 1| | | 526 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 528 Figure 5: Example of Reuse of Octets 5 & 6 using RC=1 530 The bits shown in each row of Figure 5 define the semantics of the 531 bits shown in the next row down, as follows: 533 o The top row identifies that the packet is atomic and the RC flag 534 is 1. Therefore octets 5 & 6 of the IPv4 header are redefined as 535 the ID-Reuse field. 537 o The middle row shows the bits assigned to Standard B and 538 Experiment A by IANA. An implementer has to ensure that all the 539 bits of the ID-Reuse field that are yet to be defined (bits 0-10) 540 are cleared to zero. 542 o The bottom row shows that an implementation of ExA has set its 543 2-bit sub-field to codepoint 01 and an implementation of StB has 544 set its 3-bit sub-field to codepoint 101. The meaning of each 545 would be defined in the RFCs for ExA and StB respectively. 547 Imagine now that Experiment C (ExC) is defined later to use bits 0-7 548 of the ID-Reuse field. If the packet in Figure 5 is received by an 549 implementation of ExC, then it will see only zeros in the ExC sub- 550 field. Therefore the implementation of ExC will treat the packet as 551 if ExC is turned off (as mandated in Section 5). 553 Similarly, the implementation of protocol StB can rely on being able 554 to turn off Experiment A by setting bits 14 & 15 to zero. 556 6. Incremental Deployment of New Uses of the IPv4 ID Field 558 When implementations first set the Recycled flag to 1, they are 559 likely to be blocked by certain middleboxes, either deliberately 560 (e.g. firewalls that assume anomalies are attacks) or erroneously 561 (e.g. having misunderstood the phrase "reserved, must be zero" in 562 RFC791). It is also possible that broken 'normalisers' might clear 563 RC to zero if it is 1, although so far no tests have found such 564 broken behaviour. 566 To address this problem, Section 6.2 introduces a way to use a sub- 567 field of ID-Reuse without having to set RC=1. In this approach, 568 packet headers using the new protocol will be indistinguishable from 569 an IPv4 header not using the new protocol. Therefore it will be 570 possible to guarantee that middleboxes will not treat packets using 571 the new protocol any differently from other IPv4 packets. 573 Many pre-existing IPv4 hosts cycle through all the values in the IP 574 ID field even when sending atomic packets in which the IP ID field 575 has no function. Therefore, these pre-existing IPv4 hosts will 576 occasionally issue a packet that happens to look as if it is using a 577 codepoint of a new protocol using the IP ID field. Without RC=1, 578 there will be no way to distinguish the two. 580 +------+---------------------+--------------------------+ 581 | | middlebox traversal | new protocol recognition | 582 +------+---------------------+--------------------------+ 583 | RC=0 | Assured | Uncertain | 584 | RC=1 | Uncertain | Assured | 585 +------+---------------------+--------------------------+ 587 Table 2: Tradeoff between deterministic middlebox traversal and 588 deterministic protocol recognition 590 Table 2 shows the tradeoff between using RC=0 or RC=1: 592 RC=0: If an implementation of a new protocol uses RC=0, its packets 593 will traverse middleboxes, but it will suffer a small fraction of 594 false positives when recognising which packets using the new 595 protocol -- occasionally it will mistakenly assume a packet is 596 using the new protocol when it is actually just random noise in 597 the IP ID field from a pre-existing implementation. 599 RC=1: If an implementation of a new protocol uses RC=1, its packets 600 may be black-holed by some middleboxes, but it will be certain 601 which packets use the new protocol and which don't. 603 Nonetheless, a probabilistic protocol that can be deployed may be 604 more useful than a deterministic protocol that cannot. 606 6.1. Process Example with RC=0 608 Figure 6 shows an example of how this approach would work with RC=0. 609 For illustration purposes imagine, as in the previous example in 610 Section 5.2, that an experimental track RFC has been published called 611 Experiment A (ExA) that defines bits 14 to 15 of the ID-Reuse field 612 for atomic packets with RC=1. Now imagine another experimental track 613 RFC has been published called Experiment B (ExB) that defines a use 614 for bits 11 to 13 of the ID-Reuse field, but does not require RC=1. 615 In fact a packet is defined as complying with ExB whether RC=1 or 616 RC=0 (i.e., RC=X, where 'X' means don't care). Figure 7 shows the 617 IANA registry entries for these imaginary sub-fields. 618 0 1 2 3 619 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 620 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 621 | ID-Reuse _____ |X 1 0|0 0 0 0 0 0 0 0 0 0 0 0 0| 622 |0 0 0 0 0 0 0 0 0 0 0| ExB |0 0|0 | | 623 | |1 0 1| | | | 624 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 626 Figure 6: Example of Experimental Reuse of Octets 5 & 6 Without 627 Requiring RC=1 629 The bits shown in each row of Figure 6 define the semantics of the 630 bits shown in the next row down, as follows: 632 o The top row identifies that the packet is atomic. The RC flag is 633 don't care ('X'), so RC does not have to be 1. Implementations 634 can clear RC to 0 to traverse awkward middleboxes, but RC can be 635 set to 1 otherwise. 637 o The middle row shows that an implementation of Experiment B (ExB) 638 has set RC=0. It is also using the ID-Reuse field, so it clears 639 all the bits to zero except those in its own sub-field (bits 640 11-13). It will have registered this experimental use with the 641 IANA as shown in the top example of Figure 7. 643 o The bottom row shows that an implementation of ExB has set its 644 3-bit sub-field to codepoint 101, the meaning of which will have 645 been defined in the RFC specifying the ExB protocol. 647 Note that, the process for using protocol ExB without RC=1 648 (Section 6.2) precludes an implementation from using the ExA protocol 649 in the same packet -- any one packet can only be part of one RC=0 650 protocol at a time. 652 6.2. Process for Using the ID-Reuse Field Without Requiring RC=1 654 This approach SHOULD NOT be used unless the preferred approach 655 (Section 5) is impractical due to middleboxes blocking packets with 656 RC set to 1. 658 To follow this non-preferred approach, the registration with the IANA 659 MUST specify that the sub-field of ID-Reuse is defined for 'RC=X', 660 meaning "don't care", that is RC may be either set or cleared (for an 661 example, see the final bullet of the imaginary registration details 662 in Section 8). The RFC defining the relevant ID-Reuse sub-field MUST 663 also make it clear that the sub-field is defined for either value of 664 the Recycled flag (RC=X) in an atomic IPv4 packet. 666 This approach will not be feasible for all protocols; only those that 667 satisfy the severe constraints laid down below. Otherwise, for 668 protocols that cannot satisfy these prerequisite constraints, the 669 preferred approach in Section 5 wth RC=1 will be the only option. 671 Once a sub-field of the ID-Reuse field has been registered with the 672 IANA, implementations of the protocol can use any of the available 673 codepoints within that sub-field in atomic packets without having to 674 set RC=1, if and only if the following constraints can be satisfied: 676 1. New protocol implementations MUST NOT use RC=0 unless the 677 treatments associated with all the new codepoints are generally 678 benign to packets not taking part in the protocol. 'Benign' 679 means the new protocol SHOULD do no more harm to other packets 680 than previous implementations did. Using the term 'SHOULD' 681 rather than 'MUST' does not completely rule out new protocol 682 proposals that might sometimes introduce slightly more harm, but 683 such proposals will need to give strong justifications 685 2. Implementations MUST clear all the other bits of the ID-Reuse 686 field (except those in the new protocol's sub-field) to zero. 687 Note that this is different to the approach with RC=1, where more 688 than one sub-field at once can be non-zero 690 3. In addition the constraints in Section 5.1 must also be 691 satisfied. 693 Constraint #1 is severe but necessary in order to ensure that a new 694 protocol (e.g. ExB) does not harm atomic packets from pre-existing 695 IPv4 implementations. For example, a receiving implementation of ExB 696 can assume that most packets with all zeros in bits 0-10 and 14-15 697 were deliberately set by another implementation of ExB. But many 698 pre-existing implementations of IPv4 will be cycling (sequentially or 699 randomly) through all the IPID values as they send out packets. 700 Occasionally they will send out a packet that happens to look like it 701 complies with protocol ExB. For the case of ExB with a 3-bit sub- 702 field, such false positives will occur with probability 1 in 2^13 703 (~0.01%). We term this the misrecognition probability. 705 If the new protocol were designed to do harm (e.g. to deprioritise 706 certain packets against others) that would be fine for those packets 707 intended to take part in the new protocol. But it would not be 708 acceptable to harm even a small proportion of packets misrecognised 709 as using the new protocol. This is why the RC=0 approach can only be 710 allowed for a new protocol that is generally benign. 712 Constraint #2 is necessary in order to ensure the misrecognition 713 probability remains low. If only one sub-field is allowed at one 714 time, all the other bits in the ID-Reuse field will have to be zero. 715 This ensures that a pre-existing IPv4 implementation cycling through 716 all the IP ID values will collide less frequently with values used 717 for each new protocol. 719 As already stated (Section 5), each new protocol MUST define the all- 720 zeros codepoint of its sub-field to mean that the new protocol is 721 'turned off'. 723 This arrangement ensures that packets with an IPv4 ID of zero will 724 never collide with a codepoint used by any ID-Reuse scheme, whether 725 RC=0 or RC=1. All zeros was deliberately chosen as the common 726 'turned off' codepoint because some pre-existing implementations have 727 used zero as the default IP ID for atomic packets. 729 In either case, whether the Recycled flag is set or not, a sub-field 730 of the ID-Reuse field MUST be registered with the IANA, initially for 731 experimental use, by referencing the relevant experimental track RFC. 732 This will ensure that experiments with different sub-fields of the 733 ID-Reuse field can proceed in parallel on the public Internet without 734 colliding with each other. The referenced RFC MUST define a coherent 735 process for returning the bits for other uses if the experimental 736 approach does not progress to the standards track. 738 The same sub-field can be used with the same semantics as the 739 experiment progresses, initially with the Recycled flag cleared to 0 740 and later set to 1. And the same protocol semantics can be used 741 whether the proposal is experimental or standards track. Thus, the 742 whole process is designed to: 744 1. allow initial experiments to use RC=0 to traverse non-compliant 745 middleboxes (Section 6); 747 2. then, once sufficient middleboxes forward RC=1 packets, the 748 experiment can either be continued with RC=1 (Section 5); 750 3. or the experiment can progress cleanly to the standards track, 751 while still using the same sub-field but with RC=1; 753 4. or the experiment can be terminated without having wasted any 754 header bits. 756 (Step 1 is only feasible if the extra constraints in Section 6.2 can 757 be satisfied. If not, Step 2 will be the only feasible first step.) 759 For the avoidance of doubt, any use of ID-Reuse, whether experimental 760 or not, is also subject to the general constraints already enumerated 761 in Section 5.1. 763 7. Updates to Existing RFCs 765 Great care has been taken to ensure all the updates defined in this 766 specifications are incrementally deployable. 768 The definition of the RC flag in Section 3 updates the status of this 769 flag that was "reserved, must be zero" in [RFC0791]. The 770 redefinition of the IP Identification field as the ID-Reuse field 771 when an IPv4 packet is atomic also updates RFC791. 773 Updates to existing RFC791 implementations are only REQUIRED if they 774 discard IPv4 packets with RC=1, or change RC from 1 to 0, both of 775 which are misinterpretations of RFC791 anyway. Otherwise, there will 776 be no need to update an RFC791-compliant IPv4 stack until new use(s) 777 for the ID-Reuse field are also specified. 779 The recommendation in Section 4.2 to copy the ID-Reuse field when 780 encapsulating an atomic IPv4 packet with another atomic IPv4 header 781 updates IPv4-in-IPv4 encapsulation specifications [RFC2003] 782 [RFC4301]. These updates to tunnels are likely to be recommended 783 rather than essential for interworking, so they can be implemented as 784 part of routine code maintenance. 786 The ability to redefine the IPv4 ID field of an atomic packet updates 787 [ipv4-id-update], which states "The IP ID is not defined if the 788 packet (datagram) is atomic". Nonetheless, octets 5 & 6 of an atomic 789 packet still MUST NOT be interpreted with the semantics of the 790 Identification field. 792 [RFC2780] provides the IANA with guidelines on allocating values in 793 IP and related headers. The process defined in Section 5 and 794 Section 6 update RFC2780, given ID-Reuse is effectively a new field 795 in the IPv4 header. 797 [RFC4727] defines the processes for experimental use of values in 798 fields in the IP header that are managed by the IANA. The processes 799 defined in Section 5 and Section 6 update RFC4727 to include the new 800 alternative use of the IPv4 ID field as an ID-Reuse field. 802 8. IANA Considerations 804 The IANA is requested to establish a new registry to record 805 allocation of sub-divisions of the ID-Reuse field and to avoid 806 duplicate allocations. The ID-Reuse field is an alternative use of 807 the Identification field of the IPv4 header in atomic packets 808 (Section 3). All 16 bits are available for assignment, either as 809 sub-fields of bits or as sets of codepoints within a sub-field of 810 bits. Each sub-division of the ID-Reuse field MUST be allocated 811 through an IETF Consensus action. The registry MUST then record: 813 Protocol name: the name for the protocol, as used in the RFC 814 defining it 816 RFC: the RFC that defines the semantics of the codepoints used by 817 the protocol 819 Leftmost bit: the leftmost bit allocated, counting from bit 0 at the 820 most significant bit (which is bit 32 of the IPv4 header, counting 821 from 0) 823 No. of bits allocated: the width in bits of the allocated sub-field 825 Codepoint range (optional): The range of codepoints within the 826 assigned sub-field of bits that the protocol uses 828 Sub-field defined if: the precondition for the sub-field to be 829 defined (Section 5). Valid entries MUST include the condition 830 that the packet is atomic and MUST specifiy valid values of the 831 Recycled (RC) flag, either 'RC=1' or 'RC=X', where 'X' means don't 832 care (Section 6). 834 Two example registrations are shown in Figure 7. 836 Protocol name: ExB 837 RFC: BBBB 838 Most significant bit: 11 839 No. of bits allocated: 3 840 Codepoint range: all 841 Sub-field defined if: Atomic packet and RC=X 843 Protocol name: ExA 844 RFC: AAAA 845 Most significant bit: 14 846 No. of bits allocated: 2 847 Codepoint range: all 848 Sub-field defined if: Atomic packet and RC=1 850 Figure 7: Example IANA Registry of Sub-fields of the ID-Reuse Field 852 9. Security Considerations 854 Integrity Checking: This specification make the semantics of octets 855 5 & 6 of the IPv4 header (IP ID or ID-Reuse) depend on the setting 856 of octets 7 & 8 (all the Control Flags and the Fragment Offset 857 field). The IP Authentication Header (AH) [RFC4302] covers octets 858 5 & 6 but not octets 7 & 8. Therefore AH can assure the integrity 859 of the bits in the ID-Reuse field, but it cannot verify whether or 860 not the sender intended those bits to have the semantics of an ID- 861 Reuse field. 863 Any security-sensitive application of the ID-Reuse field will 864 therefore need to provide its own integrity checking of the status 865 of the Control Flags and Fragment Offset. Such a facility would 866 need to take into account that the present specification allows an 867 intermediate node to set the Recycled flag, but not to clear it 868 (Section 4.1). 870 Covert channels: It has always been possible to use bit 48 of the 871 IPv4 header for a 1 bit per packet covert channel, for instance 872 between a network protected by IPsec and an unprotected network. 873 Bit 48 could be covertly toggled to pass messages because it had 874 no function (so no-one would notice any affect on the main 875 communication channel) and it was not covered by IPsec 876 authentication. On the other hand, once alerted to the 877 vulnerability, it has always been easy for an IPsec gateway to 878 spot bit 48 being used as a covert channel, given bit 48 was meant 879 to always be zero. 881 Now that bit 48 has been given a function, it will often no longer 882 be possible for an attacker to toggle it without affecting the 883 main data communication. However, whenever the main communication 884 does not depend on bit 48, it will be easier to for an attacker to 885 toggle it covertly given it will no longer stand out as anomalous 886 behaviour. 888 10. Conclusions 890 This specification builds on recent moves to make the approach to 891 fragmentation in IPv4 more closely match that of IPv6. Already the 892 fields that support fragmentation in the IPv4 header are usually 893 redundant, but unfortunately they are non-optional. 895 This specification makes it possible to reuse the 16 bits of the IPv4 896 ID field when they are not needed for reassembly. The last unused 897 bit in the IPv4 header is used in order to unambiguously flag that 898 the IP ID field has new semantics. The bit is called the Recycled 899 flag, because it allows the IP ID field to be recycled for new 900 purposes when it would otherwise be redundant. Whenever the IP ID 901 field has new semantics, it is termed the ID-Reuse field. 903 The process for redefining the semantics of sub-fields of this ID- 904 Reuse field has been laid down, both for experimental and standards 905 actions. Great care has been taken throughout to ease incremental 906 deployment. The same sub-field can be used with the same semantics 907 as an experiment evolves into a standards action. Initially it is 908 even possible for certain experiments to leave the Recycled flag 909 cleared to zero, in order to traverse any awkward middleboxes that 910 incorrectly discard or normalise packets if the Recycled flag is set. 912 11. Acknowledgements 914 Rob Hancock originally pointed out that code to handle new protocols 915 can tell the machine where to look for the relevant header. Dan Wing 916 pointed out that codepoints, not just whole bits, could be assigned 917 for protocols that are mutually exclusive. 919 Bob Briscoe is partly funded by Trilogy, a research project (ICT- 920 216372) supported by the European Community under its Seventh 921 Framework Programme. 923 Comments Solicited (to be removed by the RFC Editor): 925 Comments and questions are encouraged and very welcome. They can be 926 addressed to the IETF Internet Area working group mailing list 927 , and/or to the author(s). 929 12. Outstanding Issues (to be removed when all resolved) 931 1. ... 933 13. References 935 13.1. Normative References 937 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, 938 September 1981. 940 [RFC2003] Perkins, C., "IP Encapsulation within IP", 941 RFC 2003, October 1996. 943 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 944 Requirement Levels", BCP 14, RFC 2119, March 1997. 946 [RFC2780] Bradner, S. and V. Paxson, "IANA Allocation 947 Guidelines For Values In the Internet Protocol and 948 Related Headers", BCP 37, RFC 2780, March 2000. 950 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 951 Internet Protocol", RFC 4301, December 2005. 953 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, 954 December 2005. 956 [RFC4727] Fenner, B., "Experimental Values In IPv4, IPv6, 957 ICMPv4, ICMPv6, UDP, and TCP Headers", RFC 4727, 958 November 2006. 960 [ipv4-id-update] Touch, J., "Updated Specification of the IPv4 ID 961 Field", draft-ietf-intarea-ipv4-id-update-01 (work 962 in progress), October 2010. 964 13.2. Informative References 966 [Cisco.IPv6Ext] Cisco, "IPv6 Extension Headers Review and 967 Considerations", Cisco Technology White Paper , 968 October 2006, . 972 [Fransson04] Fransson, P. and A. Jonsson, "End-to-end 973 measurements on performance penalties of IPv4 974 options", Luleae University of Technology, 975 Technical Report 2004:03, 2004, . 979 [RFC4821] Mathis, M. and J. Heffner, "Packetization Layer 980 Path MTU Discovery", RFC 4821, March 2007. 982 [RFC4963] Heffner, J., Mathis, M., and B. Chandler, "IPv4 983 Reassembly Errors at High Data Rates", RFC 4963, 984 July 2007. 986 Appendix A. Why More Bits Cannot be Freed (To be Removed by RFC Editor) 988 Given this specification uses the last unassigned bit in the IPv4 989 header, it is worth checking whether it can be used to flag a new use 990 for more than the 16 bits in the IP ID field of atomic packets. 992 IHL: Ideally, the Internet header length field (4 bits) could be 993 made redundant if the length of those IPv4 headers with bit 48 set 994 were redefined to be fixed at 20 octets. Then a similar approach 995 to IPv6 could be taken with the Protocol field redefined as a Next 996 Header field and each extension header specifying its own length. 998 Unfortunately, although IPv4 options are rarely used and generally 999 ignored, this idea would not be incrementally deployable. There 1000 are probably billions of pre-existing implementations of the IPv4 1001 stack that will use the IHL field to find the transport protocol 1002 header, without ever looking at bit 48. If the IHL field were 1003 given any other semantics conditional on bit 48 being set, all 1004 these pre-existing stacks would break. 1006 Header Checksum: Ideally, the Header Checksum (16 bits) could be 1007 made redundant in those IPv4 headers with bit 48 set. Then a 1008 similar approach to IPv6 could be taken where the integrity of the 1009 IP header relies on the end-to-end checksum of the transport 1010 protocol, which includes the main fields in the IP header. 1012 Unfortunately, again, this idea would not be incrementally 1013 deployable. Pre-existing implementations of the IPv4 stack might 1014 verify the header checksum without ever looking at bit 48. And 1015 anyway IPv4 stacks on probably every pre-existing router 1016 implementation would update the checksum field without knowing to 1017 check whether bit 48 was set. Therefore if the field were used 1018 for any other purpose than a checksum, it would be impossible to 1019 predict how its value might be changed by a combination of pre- 1020 existing and new stacks. 1022 It is clear that reusing fields other than the IPv4 ID would be 1023 fraught with incremental deployment problems. The reason the IPv4 ID 1024 field can be reused, is that an atomic packet already does not need 1025 an Identification field, whether bit 48 is set or not. Setting bit 1026 48 merely allows new implementations that understand ID-Reuse 1027 semantics to be certain the value in the ID-Reuse field was not 1028 written by an implementation that intended it to have Identification 1029 semantics. 1031 Appendix B. Experimental or Standards Track? (To Be Removed Before 1032 Publication) 1034 This document defines a protocol (using the Recycled flag) to enable 1035 other protocols (using the ID-Reuse field). The Recycled flag 1036 protocol is currently written as if it is on the IETF standards 1037 track. Nonetheless it might be feasible to write it for the 1038 experimental track. This appendix discusses the pros and cons of 1039 each. 1041 The Recycled flag uses up the last unused bit in the IPv4 header. 1042 The present specification defines a use for this last bit in the 1043 expectation that the Internet community will find ingeneous new 1044 use(s) for sub-fields of the ID-Reuse field, because then the 1045 Recycled flag will be needed to unambiguously indicate the new 1046 semantics. However, there is a risk that the last IPv4 header bit 1047 could be wasted, if no new uses for the IP ID field can be found 1048 within the constraints of its previous use for fragment reassembly, 1049 or if new experimental uses are proposed but none successfully 1050 proceed through to standards actions. 1052 The risk of wasting the last bit would be mitigated if the definition 1053 of the Recycled flag itself was initially on the experimental track. 1054 Then, if some experimental use(s) of the ID-Reuse field did see 1055 widespread adoption, the RC flag protocol could progress to the 1056 standards track. On the other hand, if no ID-Reuse experiments 1057 happened, the RC flag could possibly be reclaimed for another use in 1058 the future. This would require all experiments with the RC flag to 1059 be confined in time, so that stray implementations of old experiments 1060 would not conflict with future uses of the flag. 1062 Eventually, each specification for each sub-field of ID-Reuse might 1063 either progress on the experimental track or standards track. 1064 However, an enabler for standards track specifications cannot itself 1065 only be experimental. Therefore the RC flag protocol would have to 1066 be on the standards track, to enable standards track protocols as 1067 well as experimental. Figure 8 illustrates this need for the RC flag 1068 protocol to have sufficient rank for any protocols it enables. 1070 +---------+---------------------------+ 1071 | RC flag | ID-Reuse sub-field track | 1072 | track +-------------+-------------+ 1073 | | Expt | Stds | 1074 +---------+-------------+-------------+ 1075 | Expt | Expt | INVALID | 1076 | Stds | Expt | Stds | 1077 +---------+-------------+-------------+ 1079 The IETF track of the RC flag protocol in the present document (rows) 1080 and of any particular RFC specifying a sub-field of the ID-Reuse 1081 field (columns). The combination determines the status of any 1082 particular sub-field as shown at the intersection of the relevant row 1083 and column. 1085 Figure 8: Validity of Combinations of IETF tracks for the RC flag and 1086 an ID-Reuse Subfield 1088 One purpose of the present draft is to outline how new uses of ID- 1089 Reuse sub-fields can progress seamlessly from experimental track to 1090 standards track. Therefore, this draft is written as if it were on 1091 the standards track. Otherwise the processes for enabling standards 1092 track documents would have had to be written hypothetically, which 1093 would have been highly confusing. Nonetheless, no intent to prejudge 1094 that this document should be or will be on the standards track is 1095 implied. 1097 If it were decided that the present draft should start on the 1098 experimental track, all the text about enabling standards track 1099 protocols would have to be edited out, or perhaps moved to a non- 1100 normative appendix. 1102 Alternatively, the IETF might see some obvious new uses for sub- 1103 fields of the ID-Reuse field that would make it reasonable to fast- 1104 track the RC flag straight onto the standards track. 1106 Author's Address 1108 Bob Briscoe 1109 BT 1110 B54/77, Adastral Park 1111 Martlesham Heath 1112 Ipswich IP5 3RE 1113 UK 1115 Phone: +44 1473 645196 1116 EMail: bob.briscoe@bt.com 1117 URI: http://bobbriscoe.net/