idnits 2.17.1 draft-ietf-tcpm-tcp-soft-errors-08.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** It looks like you're using RFC 3978 boilerplate. You should update this to the boilerplate described in the IETF Trust License Policy document (see https://trustee.ietf.org/license-info), which is required now. -- Found old boilerplate from RFC 3978, Section 5.1 on line 16. -- Found old boilerplate from RFC 3978, Section 5.5, updated by RFC 4748 on line 679. -- Found old boilerplate from RFC 3979, Section 5, paragraph 1 on line 690. -- Found old boilerplate from RFC 3979, Section 5, paragraph 2 on line 697. -- Found old boilerplate from RFC 3979, Section 5, paragraph 3 on line 703. 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 Copyright Line does not match the current year -- 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 (April 23, 2008) is 5846 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 793 (Obsoleted by RFC 9293) == Outdated reference: A later version (-08) exists of draft-ietf-behave-tcp-07 == Outdated reference: A later version (-12) exists of draft-ietf-tcpm-icmp-attacks-03 -- Obsolete informational reference (is this intentional?): RFC 816 (Obsoleted by RFC 7805) Summary: 2 errors (**), 0 flaws (~~), 3 warnings (==), 8 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TCP Maintenance and Minor F. Gont 3 Extensions (tcpm) UTN/FRH 4 Internet-Draft April 23, 2008 5 Intended status: Informational 6 Expires: October 25, 2008 8 TCP's Reaction to Soft Errors 9 draft-ietf-tcpm-tcp-soft-errors-08.txt 11 Status of this Memo 13 By submitting this Internet-Draft, each author represents that any 14 applicable patent or other IPR claims of which he or she is aware 15 have been or will be disclosed, and any of which he or she becomes 16 aware will be disclosed, in accordance with Section 6 of BCP 79. 18 Internet-Drafts are working documents of the Internet Engineering 19 Task Force (IETF), its areas, and its working groups. Note that 20 other groups may also distribute working documents as Internet- 21 Drafts. 23 Internet-Drafts are draft documents valid for a maximum of six months 24 and may be updated, replaced, or obsoleted by other documents at any 25 time. It is inappropriate to use Internet-Drafts as reference 26 material or to cite them other than as "work in progress." 28 The list of current Internet-Drafts can be accessed at 29 http://www.ietf.org/ietf/1id-abstracts.txt. 31 The list of Internet-Draft Shadow Directories can be accessed at 32 http://www.ietf.org/shadow.html. 34 This Internet-Draft will expire on October 25, 2008. 36 Abstract 38 This document describes a non-standard, but widely implemented, 39 modification to TCP's handling of ICMP soft error messages, that 40 rejects pending connection-requests when those error messages are 41 received. This behavior reduces the likelihood of long delays 42 between connection establishment attempts that may arise in a number 43 of scenarios, including one in which dual stack nodes that have IPv6 44 enabled by default are deployed in IPv4 or mixed IPv4 and IPv6 45 environments. 47 Table of Contents 49 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 50 2. Error Handling in TCP . . . . . . . . . . . . . . . . . . . . 3 51 2.1. Reaction to ICMP error messages that indicate hard 52 errors . . . . . . . . . . . . . . . . . . . . . . . . . . 4 53 2.2. Reaction to ICMP error messages that indicate soft 54 errors . . . . . . . . . . . . . . . . . . . . . . . . . . 5 55 3. Problems that may arise from TCP's reaction to soft errors . . 5 56 3.1. General Discussion . . . . . . . . . . . . . . . . . . . . 5 57 3.2. Problems that may arise with Dual Stack IPv6 on by 58 Default . . . . . . . . . . . . . . . . . . . . . . . . . 6 59 4. Deployed workarounds for long delays between 60 connection-establishment attempts . . . . . . . . . . . . . . 7 61 4.1. Context-sensitive ICMP/TCP interaction . . . . . . . . . . 7 62 4.2. Context-sensitive ICMP/TCP interaction with repeated 63 confirmation . . . . . . . . . . . . . . . . . . . . . . . 8 64 5. Possible drawbacks of changing ICMP semantics . . . . . . . . 9 65 5.1. Non-deterministic transient network failures . . . . . . . 9 66 5.2. Deterministic transient network failures . . . . . . . . . 9 67 5.3. Non-compliant Network Address Translators (NATs) . . . . . 10 68 6. Security Considerations . . . . . . . . . . . . . . . . . . . 10 69 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 11 70 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 11 71 9. Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 11 72 10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 11 73 10.1. Normative References . . . . . . . . . . . . . . . . . . . 11 74 10.2. Informative References . . . . . . . . . . . . . . . . . . 12 75 Appendix A. Change log (to be removed before publication of 76 the document as an RFC) . . . . . . . . . . . . . . . 13 77 A.1. Changes from draft-ietf-tcpm-tcp-soft-errors-07 . . . . . 13 78 A.2. Changes from draft-ietf-tcpm-tcp-soft-errors-06 . . . . . 13 79 A.3. Changes from draft-ietf-tcpm-tcp-soft-errors-05 . . . . . 13 80 A.4. Changes from draft-ietf-tcpm-tcp-soft-errors-04 . . . . . 13 81 A.5. Changes from draft-ietf-tcpm-tcp-soft-errors-03 . . . . . 13 82 A.6. Changes from draft-ietf-tcpm-tcp-soft-errors-02 . . . . . 14 83 A.7. Changes from draft-ietf-tcpm-tcp-soft-errors-01 . . . . . 14 84 A.8. Changes from draft-ietf-tcpm-tcp-soft-errors-00 . . . . . 14 85 A.9. Changes from draft-gont-tcpm-tcp-soft-errors-02 . . . . . 14 86 A.10. Changes from draft-gont-tcpm-tcp-soft-errors-01 . . . . . 14 87 A.11. Changes from draft-gont-tcpm-tcp-soft-errors-00 . . . . . 14 88 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 15 89 Intellectual Property and Copyright Statements . . . . . . . . . . 16 91 1. Introduction 93 The handling of network failures can be separated into two different 94 actions: fault isolation and fault recovery. Fault isolation 95 consists of the actions that hosts and routers take to determine that 96 there is a network failure. Fault recovery, on the other hand, 97 consists of the actions that hosts and routers perform in an attempt 98 to survive a network failure [RFC0816]. 100 In the Internet architecture, the Internet Control Message Protocol 101 (ICMP) [RFC0792] is one fault isolation technique to report network 102 error conditions to the hosts sending datagrams over the network. 104 When a host is notified of a network error, its network stack will 105 attempt to continue communications, if possible, in the presence of 106 the network failure. The fault recovery strategy may depend on the 107 type of network failure taking place, and the time the error 108 condition is detected. 110 This document analyzes the fault recovery strategy of TCP [RFC0793], 111 and the problems that may arise due to TCP's reaction to ICMP soft 112 errors. It analyzes the problems that may arise when a host tries to 113 establish a TCP connection with a multihomed host for which some of 114 its addresses are unreachable. Additionally, it analyzes the 115 problems that may arise in the specific scenario where dual stack 116 nodes that have IPv6 enabled by default are deployed in IPv4 or mixed 117 IPv4 and IPv6 environments. 119 Finally, we document a modification to TCP's reaction to ICMP 120 messages indicating soft errors during connection startup, that has 121 been implemented in a variety of TCP/IP stacks to help overcome the 122 problems outlined below. We stress that this modification runs 123 contrary to the standard behavior and this document unambiguously 124 does not change the standard reaction. 126 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 127 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 128 document are to be interpreted as described in RFC 2119 [RFC2119]. 130 2. Error Handling in TCP 132 Network errors can be divided into soft and hard errors. Soft errors 133 are considered to be transient network failures, which are likely to 134 be solved in the near term. Hard errors, on the other hand, are 135 considered to reflect network error conditions that are unlikely to 136 be solved in the near future. 138 The Host Requirements RFC [RFC1122] states, in Section 4.2.3.9., that 139 the ICMP messages that indicate soft errors are ICMP "Destination 140 Unreachable" codes 0 (network unreachable), 1 (host unreachable), and 141 5 (source route failed), ICMP "Time Exceeded" codes 0 (time to live 142 exceeded in transit) and 1 (fragment reassembly time exceeded), and 143 ICMP "Parameter Problem". Even though ICMPv6 did not exist when 144 [RFC1122] was written, one could extrapolate the concept of soft 145 errors to ICMPv6 "Destination Unreachable" codes 0 (no route to 146 destination) and 3 (address unreachable), ICMPv6 "Time Exceeded" 147 codes 0 (Hop limit exceeded in transit) and 1 (Fragment reassembly 148 time exceeded), and ICMPv6 "Parameter Problem" codes 0 (Erroneous 149 header field encountered), 1 (Unrecognized Next Header type 150 encountered) and 2 (Unrecognized IPv6 option encountered) [RFC4443]. 152 +----------------------------------+--------------------------------+ 153 | ICMP | ICMPv6 | 154 +----------------------------------+--------------------------------+ 155 | Destination Unreachable (codes | Destination Unreachable (codes | 156 | 0, 1, and 5) | 0 and 3) | 157 +----------------------------------+--------------------------------+ 158 | Time Exceeded (codes 0 and 1) | Time exceeded (codes 0 and 1) | 159 +----------------------------------+--------------------------------+ 160 | Parameter Problem | Parameter Problem (codes 0, 1, | 161 | | and 2) | 162 +----------------------------------+--------------------------------+ 164 Table 1: Extrapolating the concept of soft errors to ICMPv6 166 When there is a network failure that is not signaled to the sending 167 host, such as a gateway corrupting packets, TCP's fault recovery 168 action is to repeatedly retransmit the segment until either it gets 169 acknowledged, or the connection times out. 171 In the case that a host does receive an ICMP error message referring 172 to an ongoing TCP connection, the IP layer will pass this message up 173 to the corresponding TCP instance to raise awareness of the network 174 failure [RFC1122]. TCP's reaction to ICMP messages will depend on 175 the type of error being signaled. 177 2.1. Reaction to ICMP error messages that indicate hard errors 179 When receiving an ICMP error message that indicates a hard error 180 condition, TCP will simply abort the corresponding connection, 181 regardless of the connection state. 183 The Host Requirements RFC [RFC1122] states, in Section 4.2.3.9, that 184 TCP SHOULD abort connections when receiving ICMP error messages that 185 indicate hard errors. This policy is based on the premise that, as 186 hard errors indicate network error conditions that will not change in 187 the near term, it will not be possible for TCP to usefully recover 188 from this type of network failure. 190 2.2. Reaction to ICMP error messages that indicate soft errors 192 If an ICMP error message is received that indicates a soft error, TCP 193 will repeatedly retransmit the segment until it either gets 194 acknowledged or the connection times out. In addition, the TCP 195 sender may record the information for possible later use [Stevens] 196 (pp. 317-319). 198 The Host Requirements RFC [RFC1122] states, in Section 4.2.3.9, that 199 TCP MUST NOT abort connections when receiving ICMP error messages 200 that indicate soft errors. This policy is based on the premise that, 201 as soft errors are transient network failures that will hopefully be 202 solved in the near term, one of the retransmissions will succeed. 204 When the connection timer expires, and an ICMP soft error message has 205 been received before the timeout, TCP can use this information to 206 provide the user with a more specific error message [Stevens] (pp. 207 317-319). 209 This reaction to soft errors exploits the valuable feature of the 210 Internet that for many network failures, the network can be 211 dynamically reconstructed without any disruption of the endpoints. 213 3. Problems that may arise from TCP's reaction to soft errors 215 3.1. General Discussion 217 Even though TCP's fault recovery strategy in the presence of soft 218 errors allows for TCP connections to survive transient network 219 failures, there are scenarios in which this policy may cause 220 undesirable effects. 222 For example, consider a scenario in which an application on a local 223 host is trying to communicate with a destination whose name resolves 224 to several IP addresses. The application on the local host will try 225 to establish a connection with the destination host, usually cycling 226 through the list of IP addresses, until one succeeds [RFC1123]. 227 Suppose that some (but not all) of the addresses in the returned list 228 are permanently unreachable. If such a permanently unreachable 229 address is the first in the list, the application will likely try to 230 use the permanently unreachable address first and block waiting for a 231 timeout before trying an alternate address. 233 As discussed in Section 2, this unreachability condition may or may 234 not be signaled to the sending host. If the local TCP is not 235 signaled concerning the error condition, there is very little that 236 can be done other than repeatedly retransmit the SYN segment, and 237 wait for the existing timeout mechanism in TCP, or an application 238 timeout, to be triggered. However, even if unreachability is 239 signaled by some intermediate router to the local TCP by means of an 240 ICMP soft error message, the local TCP will still repeatedly 241 retransmit the SYN segment until the connection timer expires (in the 242 hopes that the error is transient). The Host Requirements RFC 243 [RFC1122] states that this timer MUST be large enough to provide 244 retransmission of the SYN segment for at least 3 minutes. This would 245 mean that the application on the local host would spend several 246 minutes for each unreachable address it uses for trying to establish 247 the TCP connection. These long delays between connection 248 establishment attempts would be inappropriate for many interactive 249 applications such as the web. [Shneiderman] and [Thadani] offer some 250 insight into interactive systems (e.g., how the response time affects 251 the usability of an application). This highlights that there is no 252 one definition of a "transient error" and that the level of 253 persistence in the face of failure represents a tradeoff. 255 It is worth noting that while most applications try the addresses 256 returned by the name-to-address function in serial, this is certainly 257 not the only possible approach. For example, applications could try 258 multiple addresses in parallel until one succeeds, possibly avoiding 259 the problem of long delays between connection establishment attempts 260 described in this document. 262 3.2. Problems that may arise with Dual Stack IPv6 on by Default 264 A particular scenario in which the above sketched type of problem may 265 occur regularly is that where dual stack nodes that have IPv6 enabled 266 by default are deployed in IPv4 or mixed IPv4 and IPv6 environments, 267 and the IPv6 connectivity is non-existent 268 [I-D.ietf-v6ops-v6onbydefault]. 270 As discussed in [I-D.ietf-v6ops-v6onbydefault], there are two 271 possible variants of this scenario, which differ in whether the lack 272 of connectivity is signaled to the sending node, or not. 274 In those scenarios in which packets sent to a destination are 275 silently dropped and no ICMPv6 [RFC4443] errors are generated, there 276 is little that can be done other than waiting for the existing 277 connection timeout mechanism in TCP, or an application timeout, to be 278 triggered. 280 In scenarios where a node has no default routers and Neighbor 281 Unreachability Detection (NUD) [RFC4861] fails for destinations 282 assumed to be on-link, or where firewalls or other systems that 283 enforce scope boundaries send ICMPv6 errors, the sending node will be 284 signaled of the unreachability problem. However, as discussed in 285 Section 2.2, standard TCP implementations will not abort connections 286 when receiving ICMP error messages that indicate soft errors. 288 4. Deployed workarounds for long delays between connection- 289 establishment attempts 291 The following subsections describe a number of workarounds for the 292 problem of long delays between connection-establishment attempts that 293 have been implemented in a variety of TCP/IP stacks. We note that 294 treating soft errors as hard errors during connection establishment, 295 while widespread, is not part of standard TCP behavior and this 296 document does not change that state of affairs. The TCPM WG 297 consensus was to document this widespread implementation of 298 nonstandard TCP behavior, but to not change the TCP standard. 300 4.1. Context-sensitive ICMP/TCP interaction 302 As discussed in Section 1, it may make sense for the fault recovery 303 action to depend not only on the type of error being reported, but 304 also on the state of the connection against which the error is 305 reported. For example, one could infer that when an error arrives in 306 response to opening a new connection, it is probably caused by 307 opening the connection improperly, rather than by a transient network 308 failure [RFC0816]. 310 A number of TCP implementations have modified their reaction to soft 311 errors, to treat the errors as hard errors in the SYN-SENT or SYN- 312 RECEIVED states. For example, this workaround has been implemented, 313 for example, in the Linux kernel since version 2.0.0 (released in 314 1996) [Linux]. However, it should be noted that this change violates 315 section 4.2.3.9 of [RFC1122], which states that these Unreachable 316 messages indicate soft error conditions and therefore TCP MUST NOT 317 abort the corresponding connection. 319 [RFC3168] states that a host that receives a RST in response to the 320 transmission of an ECN-setup SYN packet MAY resend a SYN with CWR and 321 ECE cleared. This is meant to deal with faulty middle-boxes that 322 reject connections when a SYN segment has the ECE and CWR bits set. 323 Given that this section describes a modification that processes ICMP 324 error messages as hard errors when they are received for a connection 325 in any of the non-synchronized states, systems implementing this 326 behavior could resend the SYN segment with the ECE and CWR bits 327 cleared when an ICMP error message is received in response to a SYN 328 segment that had these bits set. 330 Section 4.2 discusses a more conservative approach than that sketched 331 above, that is implemented in FreeBSD. 333 4.2. Context-sensitive ICMP/TCP interaction with repeated confirmation 335 A more conservative approach than simply treating soft errors as hard 336 errors as described above would be to abort a connection in the SYN- 337 SENT or SYN-RECEIVED states only after an ICMP Destination 338 Unreachable has been received a specified number of times, and the 339 SYN segment has been retransmitted more than some specified number of 340 times. 342 Two new parameters would have to be introduced to TCP, to be used 343 only during the connection-establishment phase: MAXSYNREXMIT and 344 MAXSOFTERROR. MAXSYNREXMIT would specify the number of times the SYN 345 segment would have to be retransmitted before a connection is 346 aborted. MAXSOFTERROR would specify the number of ICMP messages 347 indicating soft errors that would have to be received before a 348 connection is aborted. 350 Two additional state variables would need to be introduced to store 351 additional state information during the connection-establishment 352 phase: "nsynrexmit" and "nsofterror". Both would be initialized to 353 zero when a connection attempt is initiated, with "nsynrexmit" being 354 incremented by one every time the SYN segment is retransmitted and 355 "nsofterror" being incremented by one every time an ICMP message that 356 indicates a soft error is received. 358 A connection in the SYN-SENT or SYN-RECEIVED states would be aborted 359 if "nsynrexmit" was greater than MAXSYNREXMIT and "nsofterror" was 360 simultaneously greater than MAXSOFTERROR. 362 This approach would give the network more time to solve the 363 connectivity problem than simply aborting a connection attempt upon 364 reception of the first soft error. However, it should be noted that 365 depending on the values chosen for the MAXSYNREXMIT and MAXSOFTERROR 366 parameters, this approach could still lead to long delays between 367 connection establishment attempts, thus not solving the problem. For 368 example, BSD systems abort connections in the SYN-SENT or the SYN- 369 RECEIVED state when a second ICMP error is received, and the SYN 370 segment has been retransmitted more than three times. They also set 371 up a "connection-establishment timer" that imposes an upper limit on 372 the time the connection establishment attempt has to succeed, which 373 expires after 75 seconds [Stevens2] (pp. 828-829). Even when this 374 policy may be better than the three-minutes timeout policy specified 375 in [RFC1122], it may still be inappropriate for handling the 376 potential problems described in this document. This more 377 conservative approach has been implemented in BSD systems for more 378 than ten years [Stevens2]. 380 We also note that the approach given in this section is a generalized 381 version of the approach sketched in the previous section. In 382 particular, with MAXSOFTERROR set to 1 and MAXSYNREXMIT set to zero 383 the schemes are identical. 385 5. Possible drawbacks of changing ICMP semantics 387 The following subsections discuss some of the possible drawbacks 388 arising from the use of the non-standard modifications to TCP's 389 reaction to soft errors described in Section 4.1 and Section 4.2. 391 5.1. Non-deterministic transient network failures 393 In scenarios where a transient network failure affects all of the 394 addresses returned by the name-to-address translation function, all 395 destinations could be unreachable for some short period of time. For 396 example, a mobile system consisting of a cell and a repeater may pass 397 through a tunnel, leading to a loss of connectivity at the repeater, 398 with the repeater sending ICMP soft errors back to the cell. In such 399 scenarios, the application could quickly cycle through all the IP 400 addresses in the list and return an error, when it could have let TCP 401 retry a destination a few seconds later, when the transient problem 402 could have disappeared. In this case, the modifications described 403 here make TCP less robust than a standards-compliant implementation. 405 Additionally, in many cases a domain name maps to a single IP 406 address. In such a case, it might be better to try that address 407 persistently according to normal TCP rules, instead of just aborting 408 the pending connection upon receipt of an ICMP soft error. 410 5.2. Deterministic transient network failures 412 There are some scenarios in which transient network failures could be 413 deterministic. For example, consider a scenario in which upstream 414 network connectivity is triggered by network use. That is, network 415 connectivity is instantiated only on an "as needed" basis. In this 416 scenario, the connection triggering the upstream connectivity could 417 deterministically receive ICMP Destination Unreachables while the 418 upstream connectivity is being activated, and thus would be aborted. 419 Again, in this case, the modifications described here make TCP less 420 robust than a standards-compliant implementation. 422 5.3. Non-compliant Network Address Translators (NATs) 424 Some NATs respond to an unsolicited inbound SYN segment with an ICMP 425 soft error message. If the system sending the unsolicited SYN 426 segment implements the workaround described in this document, it will 427 abort the connection upon receipt of the ICMP error message, thus 428 probably preventing TCP's simultaneous open through the NAT from 429 succeeding. However, it must be stressed that those NATs described 430 in this section are not BEHAVE-compliant, and therefore should 431 implement REQ-4 of [I-D.ietf-behave-tcp] instead. 433 6. Security Considerations 435 This document describes a non-standard modification to TCP's reaction 436 to soft errors that has been implemented in a variety of TCP 437 implementations. This modification makes TCP abort a connection in 438 the SYN-SENT or the SYN-RECEIVED states when it receives an ICMP 439 "Destination Unreachable" message that indicates a soft error. 440 Therefore, the modification could be exploited to reset valid 441 connections during the connection-establishment phase. 443 The non-standard workaround described in this document makes TCP more 444 vulnerable to attack, even if only slightly. However, we note that 445 an attacker wishing to reset ongoing TCP connections could send any 446 of the ICMP hard error messages in any connection state. 448 Generally, TCP backs off its retransmission timer each time it 449 retransmits the SYN segment for the same connection. If a TCP 450 implements the modification described in this document, that is, 451 tries the next address in the list upon receipt of an ICMP error 452 message, it might end up injecting more packets into the network than 453 if it had simply retried the same address a number of times. 454 However, compliant TCP implementations might already incur into this 455 behaviour (e.g., as a result of cycling through the list of IP 456 addressses in response to RST segments) as there are currently no 457 recommendations on methods for limiting the rate at which SYN 458 segments are sent for connecting to a specific destination. 460 A discussion of the use of ICMP to perform a variety of attacks 461 against TCP, and a number of counter-measures that minimize the 462 impact of these attacks can be found in [I-D.ietf-tcpm-icmp-attacks]. 464 A discussion of the security issues arising from the use of ICMPv6 465 can be found in [RFC4443]. 467 7. IANA Considerations 469 This document has no actions for IANA. 471 8. Acknowledgements 473 The author wishes to thank Mark Allman, Ron Bonica, Ted Faber, Gorry 474 Fairhurst, Sally Floyd, Tomohiro Fujisaki, Guillermo Gont, Saikat 475 Guha, Alfred Hoenes, Michael Kerrisk, Eddie Kohler, Mika Liljeberg, 476 Arifumi Matsumoto, Carlos Pignataro, Pasi Sarolahti, Pekka Savola, 477 Pyda Srisuresh, and Joe Touch, for contributing many valuable 478 comments on earlier versions of this document. 480 The author wishes to express deep and heartfelt gratitude to Jorge 481 Oscar Gont and Nelida Garcia, for their precious motivation and 482 guidance. 484 9. Contributors 486 Mika Liljeberg was the first to describe how their implementation 487 treated soft errors. Based on that, the solution discussed in 488 Section 4.1 was documented in [I-D.ietf-v6ops-v6onbydefault] by 489 Sebastien Roy, Alain Durand and James Paugh. 491 10. References 493 10.1. Normative References 495 [RFC0792] Postel, J., "Internet Control Message Protocol", STD 5, 496 RFC 792, September 1981. 498 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, 499 RFC 793, September 1981. 501 [RFC1122] Braden, R., "Requirements for Internet Hosts - 502 Communication Layers", STD 3, RFC 1122, October 1989. 504 [RFC1123] Braden, R., "Requirements for Internet Hosts - Application 505 and Support", STD 3, RFC 1123, October 1989. 507 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 508 Requirement Levels", BCP 14, RFC 2119, March 1997. 510 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 511 of Explicit Congestion Notification (ECN) to IP", 512 RFC 3168, September 2001. 514 [RFC4443] Conta, A., Deering, S., and M. Gupta, "Internet Control 515 Message Protocol (ICMPv6) for the Internet Protocol 516 Version 6 (IPv6) Specification", RFC 4443, March 2006. 518 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 519 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 520 September 2007. 522 10.2. Informative References 524 [I-D.ietf-behave-tcp] 525 Guha, S., "NAT Behavioral Requirements for TCP", 526 draft-ietf-behave-tcp-07 (work in progress), April 2007. 528 [I-D.ietf-tcpm-icmp-attacks] 529 Gont, F., "ICMP attacks against TCP", 530 draft-ietf-tcpm-icmp-attacks-03 (work in progress), 531 March 2008. 533 [I-D.ietf-v6ops-v6onbydefault] 534 Roy, S., Durand, A., and J. Paugh, "Issues with Dual Stack 535 IPv6 on by Default", draft-ietf-v6ops-v6onbydefault-03 536 (work in progress), July 2004. 538 [Linux] The Linux Project, "http://www.kernel.org". 540 [RFC0816] Clark, D., "Fault isolation and recovery", RFC 816, 541 July 1982. 543 [Shneiderman] 544 Shneiderman, B., "Response Time and Display Rate in Human 545 Performance with Computers", ACM Computing Surveys , 1984. 547 [Stevens] Stevens, W., "TCP/IP Illustrated, Volume 1: The 548 Protocols", Addison-Wesley , 1994. 550 [Stevens2] 551 Wright, G. and W. Stevens, "TCP/IP Illustrated, Volume 2: 552 The Implementation", Addison-Wesley , 1994. 554 [Thadani] Thadani, A., "Interactive User Productivity", IBM Systems 555 Journal No. 1, 1981. 557 Appendix A. Change log (to be removed before publication of the 558 document as an RFC) 560 A.1. Changes from draft-ietf-tcpm-tcp-soft-errors-07 562 o Fixes id nits. 564 A.2. Changes from draft-ietf-tcpm-tcp-soft-errors-06 566 o Added a paragraph (in Section 4.1) about the interaction of the 567 described modification with ECN-enabled connections 569 o Added a paragraph (in Section 6) about the possible scenario in 570 which a host injects SYN segments into the network at a high rate, 571 in response to ICMP soft errors. 573 o Miscellaneous editorial changes 575 A.3. Changes from draft-ietf-tcpm-tcp-soft-errors-05 577 o Miscellaneous edits, clarifications, and reorganization of both 578 workarounds into a single top-level section, as suggested by Pasi 579 Sarolahti. 581 o Added note on non-compliant NATs, as suggested by Ted Faber and 582 Saikat Guha 584 o Miscellaneous edits suggested by Gorry Fairhurst 586 o Added a table to clarify how to extrapolate the concept of ICMPv4 587 "soft errors" to ICMPv6 (as suggested by Arifumi Matsumoto and 588 Gorry Fairhurst). 590 o Miscellaneous edits, clarification on alternative approach by 591 sending connection requests in parallel, example of mobile system 592 (for non-deterministic errors), and note on the possible impact of 593 the workarounds on TCP's robusteness (as suggested by Joe Touch) 595 A.4. Changes from draft-ietf-tcpm-tcp-soft-errors-04 597 o Addresses feedback sent by Carlos Pignataro (adds missing error 598 codes in Section 2, and fixes a number of typos/writeos). 600 A.5. Changes from draft-ietf-tcpm-tcp-soft-errors-03 602 o Addresses feedback sent by Ted Faber and Gorry Fairhurst 603 (miscellaneous editorial changes). 605 A.6. Changes from draft-ietf-tcpm-tcp-soft-errors-02 607 o Moved appendix on FreeBSD's approach to the body of the draft. 609 o Removed rest of the appendix, as suggested by Ron Bonica and Mark 610 Allman. 612 o Reworded some parts of the document to make the text more neutral. 614 o Miscellaneous editorial changes. 616 A.7. Changes from draft-ietf-tcpm-tcp-soft-errors-01 618 o Addressed feedback posted by Sally Floyd (remove sentence in 619 Section 2.1 regarding processing of RST segments) 621 A.8. Changes from draft-ietf-tcpm-tcp-soft-errors-00 623 o Miscellaneous editorial changes 625 A.9. Changes from draft-gont-tcpm-tcp-soft-errors-02 627 o Draft resubmitted as draft-ietf. 629 o Miscellaneous editorial changes 631 A.10. Changes from draft-gont-tcpm-tcp-soft-errors-01 633 o Changed wording to describe the mechanism, rather than proposing 634 it 636 o Miscellaneous editorial changes 638 A.11. Changes from draft-gont-tcpm-tcp-soft-errors-00 640 o Added reference to the Linux implementation in Section 4.1 642 o Added Section 5 644 o Added section on Higher-Level API 646 o Added Section 4.2 648 o Moved section "Asynchronous Application Notification" to Appendix 650 o Added section on parallel connection requests 651 o Miscellaneous editorial changes 653 Author's Address 655 Fernando Gont 656 Universidad Tecnologica Nacional / Facultad Regional Haedo 657 Evaristo Carriego 2644 658 Haedo, Provincia de Buenos Aires 1706 659 Argentina 661 Phone: +54 11 4650 8472 662 Email: fernando@gont.com.ar 663 URI: http://www.gont.com.ar 665 Full Copyright Statement 667 Copyright (C) The IETF Trust (2008). 669 This document is subject to the rights, licenses and restrictions 670 contained in BCP 78, and except as set forth therein, the authors 671 retain all their rights. 673 This document and the information contained herein are provided on an 674 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS 675 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND 676 THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS 677 OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF 678 THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED 679 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. 681 Intellectual Property 683 The IETF takes no position regarding the validity or scope of any 684 Intellectual Property Rights or other rights that might be claimed to 685 pertain to the implementation or use of the technology described in 686 this document or the extent to which any license under such rights 687 might or might not be available; nor does it represent that it has 688 made any independent effort to identify any such rights. Information 689 on the procedures with respect to rights in RFC documents can be 690 found in BCP 78 and BCP 79. 692 Copies of IPR disclosures made to the IETF Secretariat and any 693 assurances of licenses to be made available, or the result of an 694 attempt made to obtain a general license or permission for the use of 695 such proprietary rights by implementers or users of this 696 specification can be obtained from the IETF on-line IPR repository at 697 http://www.ietf.org/ipr. 699 The IETF invites any interested party to bring to its attention any 700 copyrights, patents or patent applications, or other proprietary 701 rights that may cover technology that may be required to implement 702 this standard. Please address the information to the IETF at 703 ietf-ipr@ietf.org.