idnits 2.17.1 draft-templin-seal-22.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 15. -- Found old boilerplate from RFC 3978, Section 5.5, updated by RFC 4748 on line 1262. -- Found old boilerplate from RFC 3979, Section 5, paragraph 1 on line 1273. -- Found old boilerplate from RFC 3979, Section 5, paragraph 2 on line 1280. -- Found old boilerplate from RFC 3979, Section 5, paragraph 3 on line 1286. 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 the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). -- 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 (June 26, 2008) is 5783 days in the past. Is this intentional? -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 2460 (Obsoleted by RFC 8200) -- Obsolete informational reference (is this intentional?): RFC 1063 (Obsoleted by RFC 1191) -- Obsolete informational reference (is this intentional?): RFC 1981 (Obsoleted by RFC 8201) Summary: 2 errors (**), 0 flaws (~~), 2 warnings (==), 10 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group F. Templin, Ed. 3 Internet-Draft Boeing Phantom Works 4 Intended status: Informational June 26, 2008 5 Expires: December 28, 2008 7 The Subnetwork Encapsulation and Adaptation Layer (SEAL) 8 draft-templin-seal-22.txt 10 Status of this Memo 12 By submitting this Internet-Draft, each author represents that any 13 applicable patent or other IPR claims of which he or she is aware 14 have been or will be disclosed, and any of which he or she becomes 15 aware will be disclosed, in accordance with Section 6 of BCP 79. 17 Internet-Drafts are working documents of the Internet Engineering 18 Task Force (IETF), its areas, and its working groups. Note that 19 other groups may also distribute working documents as Internet- 20 Drafts. 22 Internet-Drafts are draft documents valid for a maximum of six months 23 and may be updated, replaced, or obsoleted by other documents at any 24 time. It is inappropriate to use Internet-Drafts as reference 25 material or to cite them other than as "work in progress." 27 The list of current Internet-Drafts can be accessed at 28 http://www.ietf.org/ietf/1id-abstracts.txt. 30 The list of Internet-Draft Shadow Directories can be accessed at 31 http://www.ietf.org/shadow.html. 33 This Internet-Draft will expire on December 28, 2008. 35 Abstract 37 For the purpose of this document, subnetworks are defined as virtual 38 topologies that span connected network regions bounded by 39 encapsulated border nodes. These virtual topologies may span 40 multiple IP- and/or sub-IP layer forwarding hops, and can introduce 41 failure modes due to packet duplication and/or links with diverse 42 Maximum Transmission Units (MTUs). This document specifies a 43 Subnetwork Encapsulation and Adaptation Layer (SEAL) that 44 accommodates such virtual topologies over diverse underlying link 45 technologies. 47 Table of Contents 49 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 50 1.1. Motivation . . . . . . . . . . . . . . . . . . . . . . . . 3 51 1.2. Approach . . . . . . . . . . . . . . . . . . . . . . . . . 5 52 2. Terminology and Requirements . . . . . . . . . . . . . . . . . 5 53 3. Applicability Statement . . . . . . . . . . . . . . . . . . . 6 54 4. SEAL Protocol Specification - Tunnel Mode . . . . . . . . . . 7 55 4.1. Model of Operation . . . . . . . . . . . . . . . . . . . . 7 56 4.2. ITE Specification . . . . . . . . . . . . . . . . . . . . 9 57 4.2.1. Tunnel Interface MTU . . . . . . . . . . . . . . . . . 9 58 4.2.2. Accounting for Headers . . . . . . . . . . . . . . . . 10 59 4.2.3. Segmentation and Encapsulation . . . . . . . . . . . . 11 60 4.2.4. Sending Probes . . . . . . . . . . . . . . . . . . . . 13 61 4.2.5. Packet Identification . . . . . . . . . . . . . . . . 14 62 4.2.6. Sending SEAL Protocol Packets . . . . . . . . . . . . 14 63 4.2.7. Processing Raw ICMPv4 Messages . . . . . . . . . . . . 15 64 4.2.8. Processing SEAL-Encapsulated ICMPv4 Messages . . . . . 15 65 4.3. ETE Specification . . . . . . . . . . . . . . . . . . . . 16 66 4.3.1. Reassembly Buffer Requirements . . . . . . . . . . . . 16 67 4.3.2. IPv4-Layer Reassembly . . . . . . . . . . . . . . . . 16 68 4.3.3. Generating SEAL-Encapsulated ICMPv4 Fragmentation 69 Needed Messages . . . . . . . . . . . . . . . . . . . 17 70 4.3.4. SEAL-Layer Reassembly . . . . . . . . . . . . . . . . 18 71 4.3.5. Delivering Packets to Upper Layers . . . . . . . . . . 19 72 5. SEAL Protocol Specification - Transport Mode . . . . . . . . . 20 73 6. Link Requirements . . . . . . . . . . . . . . . . . . . . . . 20 74 7. End System Requirements . . . . . . . . . . . . . . . . . . . 20 75 8. Router Requirements . . . . . . . . . . . . . . . . . . . . . 21 76 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21 77 10. Security Considerations . . . . . . . . . . . . . . . . . . . 21 78 11. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . 21 79 12. SEAL Advantages over Classical Methods . . . . . . . . . . . . 22 80 13. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 23 81 14. References . . . . . . . . . . . . . . . . . . . . . . . . . . 23 82 14.1. Normative References . . . . . . . . . . . . . . . . . . . 23 83 14.2. Informative References . . . . . . . . . . . . . . . . . . 24 84 Appendix A. Historic Evolution of PMTUD . . . . . . . . . . . . . 25 85 Appendix B. Reliability Extensions . . . . . . . . . . . . . . . 27 86 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 27 87 Intellectual Property and Copyright Statements . . . . . . . . . . 28 89 1. Introduction 91 As Internet technology and communication has grown and matured, many 92 techniques have developed that use virtual topologies (frequently 93 tunnels of one form or another) over an actual network that suppors 94 the Internet Protocol (IP) [RFC0791][RFC2460]. Those virtual 95 topologies have elements which appear as one hop in the virtual 96 topology, but are actually multiple IP or sub-IP layer hops. These 97 multiple hops often have quite diverse properties which are often not 98 even visible to the end-points of the virtual hop. This introduces 99 many failure modes that are not dealt with well in current 100 approaches. 102 The use of IP encapsulation has long been considered as the means for 103 creating such virtual topologies. However, the insertion of an outer 104 IP header reduces the effective path MTU as-seen by the IP layer. 105 When IPv4 is used, this reduced MTU can be accommodated through the 106 use of IPv4 fragmentation, but unmitigated in-the-network 107 fragmentation has been found to be harmful through operational 108 experience and studies conducted over the course of many years 109 [FRAG][FOLK][RFC4963]. Additionally, classical path MTU discovery 110 [RFC1191] has known operational issues that are exacerbated by in- 111 the-network tunnels [RFC2923][RFC4459]. In the following 112 subsections, we present further details on the motivation and 113 approach for addressing these issues. 115 1.1. Motivation 117 Before discussing the approach, it is necessary to first understand 118 the problems. In both the Internet and private-use networks today, 119 IPv4 is ubiquitously deployed as the Layer 3 protocol. The two 120 primary functions of IPv4 are to provide for 1) addressing, and 2) a 121 fragmentation and reassembly capability used to accommodate links 122 with diverse MTUs. While it is well known that the addressing 123 properties of IPv4 are limited (hence the larger address space 124 provided by IPv6), there is a lesser-known but growing consensus that 125 other limitations may be unable to sustain continued growth. 127 First, the IPv4 header Identification field is only 16 bits in 128 length, meaning that at most 2^16 packets pertaining to the same 129 (source, destination, protocol, Identification)-tuple may be active 130 in the Internet at a given time. Due to the escalating deployment of 131 high-speed links (e.g., 1Gbps Ethernet), however, this number may 132 soon become too small by several orders of magnitude. Furthermore, 133 there are many well-known limitations pertaining to IPv4 134 fragmentation and reassembly - even to the point that it has been 135 deemed "harmful" in both classic and modern-day studies (cited 136 above). In particular, IPv4 fragmentation raises issues ranging from 137 minor annoyances (e.g., slow-path processing in routers) to the 138 potential for major integrity issues (e.g., mis-association of the 139 fragments of multiple IP packets during reassembly). 141 As a result of these perceived limitations, a fragmentation-avoiding 142 technique for discovering the MTU of the forward path from a source 143 to a destination node was devised through the deliberations of the 144 Path MTU Discovery Working Group (PMTUDWG) during the late 1980's 145 through early 1990's (see: Appendix A). In this method, the source 146 node provides explicit instructions to routers in the path to discard 147 the packet and return an ICMP error message if an MTU restriction is 148 encountered. However, this approach has several serious shortcomings 149 that lead to an overall "brittleness". 151 In particular, site border routers in the Internet are more and more 152 being configured to discard ICMP error messages coming from the 153 outside world. This is due in large part to the fact that malicious 154 spoofing of error messages in the Internet is made simple since there 155 is no way to authenticate the source of the messages. Furthermore, 156 when a source node that requires ICMP error message feedback when a 157 packets is dropped due to an MTU restriction does not receive the 158 messages, a path MTU-related black hole occurs. This means that the 159 source will continue to send packets that are too large and never 160 receive an indication from the network that they are being discarded. 162 The issues with both IPv4 fragmentation and this "classical" method 163 of path MTU discovery are exacerbated further when IP-in-IP tunneling 164 is used. For example, site border routers that are configured as 165 ingress tunnel endpoints may be required to forward packets into the 166 subnetwork on behalf of hundreds, thousands, or even more original 167 sources located within the site. If IPv4 fragmentation were used, 168 this would quickly wrap the 16-bit Identification field and could 169 lead to undetected data corruption. If "classical" IPv4 170 fragmentation were used instead, the site border router may be 171 bombarded by ICMP error messages coming from the subnetwork which may 172 be either untrustworthy or insufficiently provisioned to allow 173 translation into error message to be returned to the original 174 sources. 176 The situation is exacerbated further still by IPsec tunnels, since 177 only the first IPv4 fragment of a fragmented packet contains the 178 transport protocol selectors (e.g., the source and destination ports) 179 required for identifying the correct security association rendering 180 fragmentation useless under certain circumstances. Even worse, there 181 may be no way for a site border router the configures an IPsec tunnel 182 to transcribe the encrypted packet fragment contained in an ICMP 183 error message into a suitable ICMP error message to return to the 184 original source. Due to these many limitations, a new approach to 185 accommodate links with diverse MTUs is necessary. 187 1.2. Approach 189 For the purpose of this document, subnetworks are defined as virtual 190 topologies that span connected network regions bounded by 191 encapsulating border nodes. Examples include the global Internet 192 interdomain routing core, Mobile Ad hoc Networks (MANETs) and some 193 enterprise networks. Subnetwork border nodes forward unicast and 194 multicast IP packets over the virtual topology across multiple IP- 195 and/or sub-IP layer forwarding hops which may introduce packet 196 duplication and/or traverse links with diverse Maximum Transmission 197 Units (MTUs) 199 This document introduces a Subnetwork Encapsulation and Adaptation 200 Layer (SEAL) for tunnel-mode operation of IP over subnetworks that 201 connect the Ingress- and Egress Tunnel Endpoints (ITEs/ETEs) of 202 border nodes. Operation in transport mode is also supported when 203 subnetwork border node upper-layer protocols negotiate the use of 204 SEAL during connection establishment. SEAL accommodates links with 205 diverse MTUs and supports efficient duplicate packet detection by 206 introducing a minimal mid-layer encapsulation. 208 The SEAL encapsulation introduces an extended Identification field 209 for packet identification and a mid-layer segmentation and reassembly 210 capability that allows simplified cutting and pasting of packets. 211 Moreover, SEAL senses in-the-network IPv4 fragmentation as a "noise" 212 indication that packet sizing parameters are "out of tune" with 213 respect to the network path. Instead of experiencing this 214 fragmentation as a disasterous event, however, SEAL naturally tunes 215 its packet sizing parameters to eliminate the in-the-network 216 fragmentation and thereby squelch the noise. The SEAL encapsulation 217 layer and protocol is specified in the following sections. 219 2. Terminology and Requirements 221 The terms "inner", "mid-layer" and "outer" respectively refer to the 222 innermost IP {layer, protocol, header, packet, etc.} before any 223 encapsulation, the mid-layer IP {protocol, header, packet, etc.) 224 after any mid-layer '*' encapsulation and the outermost IP {layer, 225 protocol, header, packet etc.} after SEAL/*/IPv4 encapsulation. 227 The term "IP" used throughout the document refers to either Internet 228 Protocol version (IPv4 or IPv6). Additionally, the notation IPvX/*/ 229 SEAL/*/IPvY refers to an inner IPvX packet encapsulated in any mid- 230 layer '*' encapsulations followed by the SEAL header followed by any 231 outer '*' encapsulations followed by an outer IPvY header, where the 232 notation "IPvX" means either IP protocol version (IPv4 or IPv6). 234 The following abbreviations correspond to terms used within this 235 document and elsewhere in common Internetworking nomenclature: 237 ITE - Ingress Tunnel Endpoint 239 ETE - Egress Tunnel Endpoint 241 PTB - an ICMPv6 "Packet Too Big" or an ICMPv4 "Fragmentation 242 Needed" message 244 DF - the IPv4 header "Don't Fragment" flag 246 MHLEN - the length of any mid-layer '*' headers and trailers 248 OHLEN - the length of the outer encapsulating SEAL/*/IPv4 headers 250 S_MRU- the per-ETE SEAL Maximum Reassembly Unit 252 S_MSS - the SEAL Maximum Segment Size 254 SEAL_ID - a 32-bit Identification value; randomly initialized and 255 monotonically incremented for each SEAL protocol packet 257 SEAL_PROTO - an IPv4 protocol number used for SEAL 259 SEAL_PORT - a TCP/UDP service port number used for SEAL 261 SEAL_OPTION - a TCP option number used for (transport-mode) SEAL 263 The keywords MUST, MUST NOT, REQUIRED, SHALL, SHALL NOT, SHOULD, 264 SHOULD NOT, RECOMMENDED, MAY, and OPTIONAL, when they appear in this 265 document, are to be interpreted as described in [RFC2119]. 267 3. Applicability Statement 269 SEAL was motivated by the specific case of subnetwork abstraction for 270 Mobile Ad-hoc Networks (MANETs), however the domain of applicability 271 also extends to subnetwork abstractions of enterprise networks, the 272 interdomain routing core, etc. The domain of application therefore 273 also includes the map-and-encaps architecture proposals in the IRTF 274 Routing Research Group (RRG) (see: http://www3.tools.ietf.org/group/ 275 irtf/trac/wiki/RoutingResearchGroup). 277 SEAL introduces a minimal new sublayer for IPvX in IPvY encapsulation 278 (e.g., as IPv6/SEAL/IPv4), and appears as a subnetwork encapsulation 279 as seen by the inner IP layer. SEAL can also be used as a sublayer 280 for encapsulating inner IP packets within outer UDP/IPv4 header 281 (e.g., as IPv6/SEAL/UDP/IPv4) such as for the Teredo domain of 282 applicability [RFC4380]. When it appears immediately after the outer 283 IPv4 header, the SEAL header is processed exactly as for IPv6 284 extension headers. 286 SEAL can also be used in "transport-mode", e.g., when the inner layer 287 includes upper layer protocol data rather than an encapsulated IP 288 packet. For instance, TCP peers can negotiate the use of SEAL for 289 the carriage of protocol data encapsulated as TCP/SEAL/IPv4. In this 290 sense, the "subnetwork" becomes the entire end-to-end path between 291 the TCP peers and may potentially span the entire Internet. 293 The current document version is specific to the use of IPv4 as the 294 outer encapsulation layer, however the same principles apply when 295 IPv6 is used as the outer layer. 297 4. SEAL Protocol Specification - Tunnel Mode 299 4.1. Model of Operation 301 SEAL supports the encapsulation of inner IP packets in mid-layer and 302 outer encapsulating headers/trailers. For example, an inner IPv6 303 packet would appear as IPv6/*/SEAL/*/IPv4 after mid-layer and outer 304 encapsulations, where '*' denotes zero or more additional 305 encapsulation sublayers. Ingres Tunnel Endpoints (ITEs) add mid- 306 layer '*' and outer SEAL/*/IPv4 encapsulations to the inner packets 307 they inject into a subnetwork, where the outermost IPv4 header 308 contains the source and destination addresses of the subnetwork 309 entry/exit points (i.e., the ITE/ETE), respectively. SEAL uses a new 310 Internet Protocol type and a new encapsulation sublayer for both 311 unicast and multicast. The ITE encapsulates an inner IP packet in 312 mid-layer and outer encapsulations as shown in Figure 1: 314 +-------------------------+ 315 | | 316 ~ Outer */IPv4 headers ~ 317 | | 318 I +-------------------------+ 319 n | SEAL Header | 320 n +-------------------------+ +-------------------------+ 321 e ~ Any mid-layer * headers ~ ~ Any mid-layer * headers ~ 322 r +-------------------------+ +-------------------------+ 323 | | | | 324 I --> ~ Inner IP ~ --> ~ Inner IP ~ 325 P --> ~ Packet ~ --> ~ Packet ~ 326 | | | | 327 P +-------------------------+ +-------------------------+ 328 a ~ Any mid-layer trailers ~ ~ Any mid-layer trailers ~ 329 c +-------------------------+ +-------------------------+ 330 k ~ Any outer trailers ~ 331 e +-------------------------+ 332 t 333 (After mid-layer encaps.) (After SEAL/*/IPv4 encaps.) 335 Figure 1: SEAL Encapsulation 337 where the SEAL header is inserted as follows: 339 o For simple IPvX/IPv4 encapsulations (e.g., 340 [RFC2003][RFC2004][RFC4213]), the SEAL header is inserted between 341 the inner IP and outer IPv4 headers as: IPvX/SEAL/IPv4. 343 o For tunnel-mode IPsec encapsulations over IPv4, [RFC4301], the 344 SEAL header is inserted between the {AH,ESP} header and outer IPv4 345 headers as: IPvX/*/{AH,ESP}/SEAL/IPv4. 347 o For IP encapsulations over transports such as UDP, the SEAL header 348 is inserted immediately after the outer transport layer header, 349 e.g., as IPvX/*/SEAL/UDP/IPv4. 351 SEAL-encapsulated packets include a 32-bit SEAL_ID formed from the 352 concatenation of the 16-bit ID Extension field in the SEAL header as 353 the most-significant bits, and with the 16-bit Identification value 354 in the outer IPv4 header as the least-significant bits. (For tunnels 355 that traverse IPv4 Network Address Translators, the SEAL_ID is 356 instead maintained only within the 16-bit ID Extension field in the 357 SEAL header.) Routers within the subnetwork use the SEAL_ID for 358 duplicate packet detection, and ITEs/ETEs use the SEAL_ID for SEAL 359 segmentation and reassembly. 361 SEAL enables a multi-level segmentation and reassembly capability. 363 First, the ITE can use IPv4 fragmentation to fragment inner IPv4 364 packets with DF=0 before SEAL encapsulation to avoid lower-level 365 segmentation and reassembly. Secondly, the SEAL layer itself 366 provides a simple mid-layer cutting-and-pasting of mid-layer packets 367 to avoid IPv4 fragmentation on the outer packet. Finally, ordinary 368 IPv4 fragmentation is permitted on the outer packet after SEAL 369 encapsulation and used to detect and dampen any in-the-network 370 fragmentation as quickly as possible. 372 The following sections specifiy the SEAL-related operations of the 373 ITE and ETE, respectively: 375 4.2. ITE Specification 377 4.2.1. Tunnel Interface MTU 379 The ITE configures a tunnel virtual interface over one or more 380 underlying links that connect the border node to the subnetwork. The 381 tunnel interface must present a fixed MTU to the inner IP layer 382 (i.e., Layer 3) as the size for admission of inner IP packets into 383 the tunnel. Since the tunnel interface may support a potentially 384 large set of ETEs, however, care must be taken in setting a greatest- 385 common-denominator MTU for all ETEs while still upholding end system 386 expectations. 388 Due to the ubiquitous deployment of standard Ethernet and similar 389 networking gear, the nominal Internet cell size has become 1500 390 bytes; this is the de facto size that end systems have come to expect 391 will either be delivered by the network without loss due to an MTU 392 restriction on the path or a suitable PTB message returned. However, 393 the network may not always deliver the necessary PTBs, leading to 394 MTU-related black holes [RFC2923]. The ITE therefore requires a 395 means for conveying 1500 byte (or smaller) packets to the ETE without 396 loss due to MTU restrictions and without dependence on PTB messages 397 from within the subnetwork. 399 In common deployments, there may be many forwarding hops between the 400 original source and the ITE. Within those hops, there may be 401 additional encapsulations (IPSec, L2TP, etc.) such that a 1500 byte 402 packet sent by the original source might grow to a larger size by the 403 time it reaches the ITE for encapsulation as an inner IP packet. 404 Similarly, additional encapsulations on the path from the ITE to the 405 ETE could cause the encapsulated packet to become larger still and 406 trigger in-the-network fragmentation. In order to preserve the end 407 system expectations, the ITE therefore requires a means for conveying 408 these larger packets to the ETE even though there may be links within 409 the subnetwork that configure a smaller MTU. 411 The ITE should therefore set a tunnel virtual interface MTU of 1500 412 bytes plus extra room to accommodate any additional encapsulations 413 that may occur on the path from the original source (i.e., even if 414 the underlying links do not support an MTU of this size). The ITE 415 can set larger MTU values still, but should select a value that is 416 not so large as to cause excessive PTBs coming from within the tunnel 417 interface (see: Sections 4.2.2 and 4.2.6). The ITE can also set 418 smaller MTU values, however care must be taken not to set so small a 419 value that original sources would experience an MTU underflow. In 420 particular, IPv6 sources must see a minimum path MTU of 1280 bytes, 421 and IPv4 sources should see a minimum path MTU of 576 bytes. 423 The inner IP layer consults the tunnel interface MTU when admitting a 424 packet into the interface. For inner IPv4 packets larger than the 425 tunnel interface MTU and with the IPv4 Don't Fragment (DF) bit set to 426 0, the inner IPv4 layer uses IPv4 fragmentation to break the packet 427 into fragments no larger than the tunnel interface MTU then admits 428 each fragment into the tunnel as an independent packet. For all 429 other inner packets (IPv4 or IPv6), the ITE admits the packet if it 430 is no larger than the tunnel interface MTU; otherwise, it drops the 431 packet and sends an ICMP PTB message with an MTU value of the tunnel 432 interface MTU to the source. 434 4.2.2. Accounting for Headers 436 As for any transport layer protocol, ITEs use the MTU of the 437 underlying IPv4 interface, the length of any mid-layer '*' headers 438 and trailers, and the length of the outer SEAL/*/IPv4 headers to 439 determine the maximum-sized upper layer payload. For example, when 440 the underlying IPv4 interface advertises an MTU of 1500 bytes and the 441 ITE inserts a minimum-length (i.e., 20 byte) IPv4 header, the ITE 442 sees a maximum payload size of 1480 bytes. When the ITE inserts IPv4 443 header options, the size is further reduced by as many as 40 444 additional bytes (the maximum length for IPv4 options) such that as 445 few as 1440 bytes may be available for the upper layer payload. When 446 the ITE inserts additional '*' encapsulations, the available MTU for 447 the upper layer payload is reduced further still. 449 The ITE must additionally account for the length of the SEAL header 450 itself as an extra encapsulation that further reduces the size 451 available for the upper layer payload. The length of the SEAL header 452 is not incorporated in the IPv4 header length, therefore the network 453 does not observe the SEAL header as an IPv4 option. In this way, the 454 SEAL header is inserted after the IPv4 options but before the upper 455 layer payload in exactly the same manner as for IPv6 extension 456 headers. 458 4.2.3. Segmentation and Encapsulation 460 For each ETE, the ITE maintains the length of any mid-layer '*' 461 encapsulation headers and trailers (e.g., for '*' = AH, ESP, NULL, 462 etc.) in a variable 'MHLEN' and maintains the length of the outer 463 SEAL/*/IPv4 encapsulation headers in a variable 'OHLEN'. The ITE 464 maintains a SEAL Maximum Reassembly Unit (S_MRU) value for each ETE 465 as soft state within the tunnel interface (e.g., in the IPv4 466 destination cache). The ITE initializes S_MRU to a value no larger 467 than 2KB and uses this value to determine the maximum-sized packet it 468 will require the ETE to reassemble. The ITE additionally maintains a 469 SEAL Maximum Segment Size (S_MSS) value for each ETE. The ITE 470 initializes S_MSS to the maximum of (the underlying IPv4 interface 471 MTU minus OHLEN) and S_MRU/8 bytes, and decreases or increases S_MSS 472 based on any ICMPv4 Fragmentation Needed messages received (see: 473 Section 4.2.6). 475 The ITE performs segmentation and encapsulation on inner packets that 476 have been admitted into the tunnel interface. For inner IPv4 packets 477 with the DF bit set to 0, if the length of the inner packet is larger 478 than (S_MRU - MHLEN) the ITE uses IPv4 fragmentation to break the 479 packet into IPv4 fragments no larger than (S_MRU - MHLEN). For 480 unfragmentable inner packets (e.g., IPv6 packets, IPv4 packets with 481 DF=1, etc.), if the length of the inner packet is larger than 482 (MAX(S_MRU, S_MSS) - MHLEN), the ITE drops the packet and sends an 483 ICMP PTB message with an MTU value of (MAX(S_MRU, S_MSS) - MHLEN) 484 back to the original source. 486 The ITE then encapsulates each inner packet/fragment in the MHLEN 487 bytes of mid-layer '*' headers and trailers. For each such resulting 488 mid-layer packet, if the length of the mid-layer packet is no larger 489 than S_MRU but is larger than S_MSS, the ITE breaks it into N 490 segments (N <= 8) that are no larger than S_MSS bytes each. Each 491 segment except the final one MUST be of equal length, while the final 492 segment MUST be no larger than the initial segment. The first byte 493 of each segment MUST begin immediately after the final byte of the 494 previous segment, i.e., the segments MUST NOT overlap. The ITE 495 should generate the smallest number of segments possible, e.g., it 496 should not generate 6 smaller segments when the packet could be 497 accommodated with 4 larger segments. 499 Note that this SEAL segmentation is used only for mid-layer packets 500 that are no larger than S_MRU; mid-layer packets that are larger than 501 S_MRU are instead encapsulated as a single segment. Note also that 502 this SEAL segmentation ignores the fact that the mid-layer packet may 503 be unfragmentable. This segmentation process is a mid-layer (not an 504 IP layer) operation employed by the ITE to adapt the mid-layer packet 505 to the subnetwork path characteristics, and the ETE will restore the 506 packet to its original form during reassembly. Therefore, the fact 507 that the packet may have been segmented within the subnetwork is not 508 observable after decapsulation. 510 The ITE next encapsulates each segment in a SEAL header formatted as 511 follows: 513 0 1 2 3 514 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 1 515 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 516 | ID Extension |A|R|D|M|E| SEG | Next Header | 517 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 519 Figure 2: SEAL Header Format 521 where the header fields are defined as follows: 523 ID Extension (16) 524 a 16-bit extension of the Identification field in the outer IPv4 525 header; encodes the most-significant 16 bits of a 32 bit SEAL_ID 526 value. 528 A (1) 529 the "Acknowledgement Requested" bit. Set to 1 if the ITE wishes 530 to receive an explicit acknowledgement from the ETE. 532 R (1) 533 the "Report Fragmentation" bit. Set to 1 if the ITE wishes to 534 receive a report from the ETE if any IPv4 fragmentation occurs. 536 D (1) 537 the "Dont Reassemble" bit. Set to 1 if the reassembled SEAL 538 protocol packet is to be discarded by the ETE if any IPv4 539 reassemly is required. 541 M (1) 542 the "More Segments" bit. Set to 1 if this SEAL protocol packet 543 contains a non-final segment of a multi-segment mid-layer packet. 545 E (1) 546 the "Extension" bit; reserved for future use. Must be set to 0 547 for the purpose of this specification. 549 SEG (3) 550 a 3-bit Segment number. Encodes a segment number between 0 - 7. 552 Next Header (8) an 8-bit field that encodes an Internet Protocol 553 number the same as for the IPv4 protocol and IPv6 next header 554 fields. 556 For single-segment mid-layer packets, the ITE encapsulates the 557 segment in a SEAL header with (M=0; Segment=0). For N-segment mid- 558 layer packets (N <= 8), the ITE encapsulates each segment in a SEAL 559 header with (M=1; Segment=0) for the first segment, (M=1; Segment=1) 560 for the second segment, etc., with the final segment setting (M=0; 561 Segment=N-1). For the first segment (i.e., segment 0), the ITE sets 562 D=0 in the SEAL header if the ETE is to accept the packet even if 563 some portion of the packet incurs IPv4 fragmentation; for example, 564 the ITE may set D=0 in the SEAL header of the first segment for all 565 mid-layer packets no larger than S_MRU. The ITE instead sets D=1 in 566 the SEAL header of the first segment if the ETE is to discard the 567 packet if any portion of it incurs IPv4 fragmentation; in particular, 568 the ITE should set D=1 in the SEAL header of all mid-layer packets 569 larger than S_MRU. 571 The ITE next sets E=0 and sets the A and R bits in the SEAL header of 572 the first segment according to whether the packet is to be used as an 573 explicit/implicit probe as specified in Section 4.2.4. The ITE then 574 writes the Internet Protocol number corresponding to the mid-layer 575 packet in the SEAL 'Next Header' field and encapsulates each segment 576 in the requisite */IPv4 outer headers according to the specific 577 encapsulation format (e.g., [RFC2003], [RFC4213], [RFC4380], etc.), 578 except that it writes 'SEAL_PROTO' in the protocol field of the outer 579 IPv4 header (when simple IPv4 encapsualtion is used) or writes 580 'SEAL_PORT' in the outer destination service port field (e.g., when 581 UDP/IPv4 encapsulation is used). The ITE finally sets packet 582 identification values as specified in Section 4.2.5 and sends the 583 packets as specified in Section 4.2.6. 585 4.2.4. Sending Probes 587 When S_MSS is larger than S_MRU/8 bytes, the ITE sends ordinary 588 encapsulated data packets as implicit probes to detect in-the-network 589 IPv4 fragmentation and to determine new values for S_MSS. The ITE 590 sets R=1 in the SEAL header of the first segment of a SEAL-segmented 591 packet to be used as an implicit probe, and will receive ICMPv4 592 Fragmentation Needed messages from the ETE if any IPv4 fragmentation 593 occurs. When the ITE has already reduced S_MSS to the minimum value, 594 it instead sets R=0 in the SEAL header to avoid generating 595 fragmentation reports for unavoidable in-the-network fragmentation. 597 The ITE should send explicit probes periodically to manage a window 598 of SEAL_IDs of outstanding probes as a means to validate any ICMPv4 599 messages it receives. The ITE sets A=1 in the SEAL header of the 600 first segment of a SEAL-segmented packet to be used as an explicit 601 probe, where the probe can be either an ordinary data packet or a 602 NULL packet created by setting the 'Next Header' field in the SEAL 603 header to a value of "No Next Header" (see: [RFC2460], Section 4.7. 605 The ITE should further send explicit probes periodically to detect 606 increases in S_MSS by resetting S_MSS to the maximum of (the 607 underlying IPv4 interface MTU minus OHLEN) and S_MRU/8 bytes, and/or 608 sending explicit probes that are larger than the current S_MSS. 610 4.2.5. Packet Identification 612 For the purpose of packet identification, the ITE maintains a 32-bit 613 SEAL_ID value as per-ETE soft state, e.g. in the IPv4 destination 614 cache. The ITE randomly-initializes SEAL_ID when the soft state is 615 created and monotonically increments it (modulo 2^32) for each 616 successive SEAL protocol packet it sends to the ETE. For each 617 packet, the ITE writes the least-significant 16 bits of the SEAL_ID 618 value in the Identification field in the outer IPv4 header, and 619 writes the most-significant 16 bits in the ID Extension field in the 620 SEAL header. 622 For SEAL encapsulations specifically designed for the traversal of 623 IPv4 Network Address Translators (NATs), e.g., for encapsulations 624 that insert a UDP header between the SEAL header and outer IPv4 625 header such as IPv6/SEAL/UDP/IPv4, the ITE instead maintains SEAL_ID 626 as a 16-bit value that it randomly-initializes when the soft state is 627 created and monotonically increments (modulo 2^16) for each 628 successive packet. For each packet, the ITE writes SEAL_ID in the ID 629 extension field of the SEAL header and writes a random 16-bit value 630 in the Identification field in the outer IPv4 header. This is due to 631 the fact that the ITE has no way to control IPv4 NATs in the path 632 that coud rewrite the Identification value in the outer IPv4 header. 634 4.2.6. Sending SEAL Protocol Packets 636 Following SEAL segmentation and encapsulation, the ITE sets DF=0 in 637 the outer IPv4 header of every outer packet it sends. For 638 "expendable" packets (e.g., for NULL packets used as probes - see: 639 Section 4.2.6), the ITE may optionally set DF=1. 641 The ITE then sends each outer packet that encapsulates a segment of 642 the same mid-layer packet into the tunnel in canonical order, i.e., 643 Segment 0 first, followed by Segment 1, etc. and finally Segment N-1. 645 4.2.7. Processing Raw ICMPv4 Messages 647 The ITE may receive "raw" ICMPv4 error messages from either the ETE 648 or routers within the subnetwork that comprise an outer IPv4 header 649 followed by an ICMPv4 header followed by a portion of the SEAL packet 650 that generated the error (also known as the "packet-in-error"). For 651 such messages, the ITE can use the 32-bit SEAL ID encoded in the 652 packet-in-error as a nonce to confirm that the ICMP message came from 653 either the ETE or an on-path router. The ITE MAY process raw ICMPv4 654 messages as soft errors indicating that the path to the ETE may be 655 failing, but it discards any raw ICMPv4 Fragmentation Needed messages 656 for which the IPv4 header of the packet-in-error has DF=0. 658 The ITE should specifically process raw ICMPv4 Protocol Unreachable 659 messages as a hint that the ETE does not implement the SEAL protocol. 661 4.2.8. Processing SEAL-Encapsulated ICMPv4 Messages 663 In addition to any raw ICMPv4 messages, the ITE may receive SEAL- 664 encapsulated ICMPv4 messages from the ETE that comprise outer ICMPv4/ 665 */SEAL/*/IPv4 headers followed by a portion of the SEAL-encapsulated 666 packet-in-error. The ITE can use the 32-bit SEAL ID encoded in the 667 packet-in-error as well as information in the outer IPv4 and SEAL 668 headers as nonces to confirm that the ICMP message came from a 669 legitimate ETE. The ITE then verifies that the SEAL_ID encoded in 670 the packet-in-error is within the current window of transmitted 671 SEAL_IDs for this ETE. If the SEAL_ID is outside of the window, the 672 ITE discards the message; otherwise, it advances the window and 673 processes the message. 675 The ITE processes SEAL-encapsulated ICMPv4 messages other than ICMPv4 676 Fragmentation Needed exactly as specified in [RFC0792]. For SEAL- 677 encapsulated ICMPv4 Fragmentation Needed messages, if the IPv4 length 678 of the packet-in-error minus OHLEN is larger than S_MSS the ITE sets 679 S_MSS to this new value. Otherwise, if the packet-in-error is an 680 IPv4 first-fragment (i.e., with MF=1; Offset=0) the ITE sets S_MSS to 681 this new value if the value is no smaller than (576 - OHLEN) and sets 682 S_MSS to MAX(S_MSS/2, S_MRU/8) if the value is smaller than (576 - 683 OHLEN). 685 Note that in the above, 576 accounts for the nominal minimum MTU for 686 common IPv4 links. When an ETE returns an IPv4 first-fragment 687 packet-in-error with length smaller than 576, the ITE performs a 688 "limited halving" of S_MSS to account for IPv4 links with unusually 689 small MTUs or cases in which the ETE otherwise receives an undersized 690 IPv4 first-fragment. This limited halving may require multiple 691 iterations of sending probes and receiving ICMPv4 Fragmentation 692 Needed messages, but will soon converge to a stable S_MSS value. 694 When performing this limited having, it is important that the ITE 695 adjust its S_MSS size based on the first ICMPv4 Fragmentation Needed 696 message and refrain from reducing S_MSS until ICMPv4 Fragmentation 697 Needed messages pertaining to packets sent under the new S_MSS are 698 received. For example, the ITE should not repeatedly halve the S_MSS 699 based on a burst of ICMPv4 Fragmentation Needed messages all 700 pertaining to packets sent under the same S_MSS. 702 After deterimining a new value for S_MSS, if the packet-in-error is 703 an IPv4 first fragment and its SEAL header has D=1 the ITE MAY 704 transcribe the message into an ICMP PTB message to send back to the 705 original source. To do so, the ITE discards the SEAL/*/IPv4 headers 706 plus any mid-layer '*' headers/trailers of the packet-in-error then 707 encapsulates the remaining inner IP packet portion in a PTB message 708 with the MTU field set to MAX((S_MRU, S_MSS) - MHLEN). Note that 709 this may not be possible when the inner IP packet portion was 710 encrypted (e.g. via IPsec/ESP), and is otherwise not entirely 711 necessary since the ITE will discard subsequent large packets and 712 send back an ICMP PTB *before* encapsulating them and sending to the 713 ETE. Transcribing ICMPv4 Fragmentation Needed messages into ICMP 714 PTBs is therefore offered only as an optional optimization. 716 4.3. ETE Specification 718 4.3.1. Reassembly Buffer Requirements 720 ETEs MUST be capable of using IPv4-layer reassembly to reassemble 721 SEAL protocol outer IPv4 packets of (2KB + OHELN) and MUST also be 722 capable of using SEAL-layer reassembly to reassemble mid-layer 723 packets of (2KB + OHLEN). The term OHLEN is included to account for 724 the length of the SEAL/*/IPv4 header, which must be retained during 725 reassembly for the purpose of associating the fragments/segments of 726 the same packet. Note that the term S_MRU used in section 4.2 omits 727 OHLEN for the purpose of specification clarity. 729 4.3.2. IPv4-Layer Reassembly 731 The ETE performs IPv4 reassembly as-normal, and should maintain a 732 conservative high- and low-water mark for the number of outstanding 733 reassemblies pending for each ITE. When the size of the reassembly 734 buffer exceeds this high-water mark, the ETE actively discards 735 incomplete reassemblies (e.g., using an Active Queue Management (AQM) 736 strategy) until the size falls below the low-water mark. The ETE 737 should also use a reduced IPv4 maximum segment lifetime value (e.g., 738 15 seconds), i.e., the time after which it will discard an incomplete 739 IPv4 reassembly for a SEAL protocol packet. 741 After reassembly, the ETE either accepts or discards the reassembled 742 packet based on the current status of the IPv4 reassembly cache 743 (congested vs uncongested). The SEAL_ID included in the IPv4 first- 744 fragment provides an additional level of reassembly assurance, since 745 it can record a distinct arrival timestamp useful for associating the 746 first-fragment with its corresponding non-initial fragments. The 747 choice of accepting/discarding a reassembly may also depend on the 748 strength of the upper-layer integrity check if known (e.g., IPSec/ESP 749 provides a strong upper-layer integrity check) and/or the corruption 750 tolerance of the data (e.g., multicast streaming audio/video may be 751 more corruption-tolerant than file transfer, etc.). In the limiting 752 case, the ETE may choose to discard all IPv4 reassemblies and process 753 only the IPv4 first-fragment for SEAL-encapsulated error generation 754 purposes (see the following sections). 756 4.3.3. Generating SEAL-Encapsulated ICMPv4 Fragmentation Needed 757 Messages 759 During IPv4-layer reassembly, the ETE determines whether the packet 760 belongs to the SEAL protocol by checking for SEAL_PROTO in the outer 761 IPv4 header (i.e., for simple IPv4 encapsulation) or for SEAL_PORT in 762 the outer */IPv4 header (e.g., for '*'=UDP). When the ETE processes 763 the IPv4 first-fragment (i.e, one with DF=1 and Offset =0 in the IPv4 764 header) of a SEAL protocol IPv4 packet with (R=1; E=0; Segment=0) in 765 the SEAL header, it sends a SEAL-encapsulated ICMPv4 Fragmentation 766 Needed message back to the ITE with the MTU field set to 0. (Note 767 that setting a non-zero value in the MTU field of the ICMPv4 768 Fragmentation Needed message would be redundant with the length value 769 in the IPv4 header of the first fragment, since this value is set to 770 the correct path MTU through in-the-network fragmentation. Setting 771 the MTU field to 0 therefore avoids the ambiguous case in which the 772 MTU field and the IPv4 length field of the first fragment would 773 record different non-zero values.) 775 When the ETE processes a SEAL protocol IPv4 packet with (A=1; E=0; 776 Segment=0) for which no IPv4 reassembly was required, or for which 777 IPv4 reassembly was successful and the R bit was not set, it sends a 778 SEAL-encapsulated ICMPv4 Fragmentation Needed message back to the ITE 779 with the MTU value set to 0. Note therefore that when both the A and 780 R bits are set and fragmentation occurs, the ETE only sends a single 781 ICMPv4 Fragmentation Needed message, i.e., it does not send two 782 separate messages (one for the first fragment and a second for the 783 reassembled whole SEAL packet). 785 The ETE prepares the ICMPv4 Fragmentation Needed message by 786 encapsulating as much of the first fragment (or the whole IPv4 787 packet) as possible in outer */SEAL/*/IPv4 headers without the length 788 of the message exceeding 576 bytes as shown in Figure 3: 790 +-------------------------+ - 791 | | \ 792 ~ Outer */SEAL/*/IPv4 hdrs~ | 793 | | | 794 +-------------------------+ | 795 | ICMPv4 Header | | 796 |(Dest Unreach; Frag Need)| | 797 +-------------------------+ | 798 | | > Up to 576 bytes 799 ~ IP/*/SEAL/*/IPv4 ~ | 800 ~ hdrs of packet/fragment ~ | 801 | | | 802 +-------------------------+ | 803 | | | 804 ~ Data of packet/fragment ~ | 805 | | / 806 +-------------------------+ - 808 Figure 3: SEAL-encapsulated ICMPv4 Fragmentation Needed Message 810 The ETE next sets D=0, A=0, R=0, E=0 and Segment=0 in the outer SEAL 811 header, sets the SEAL_ID the same as for any SEAL packet, then sets 812 the SEAL Next Header field and the fields of the outer */IPv4 headers 813 the same as for ordinay SEAL encapsulation. The ETE then sets outer 814 IPv4 destination address to the source address of the first-fragment 815 and sets the outer IPv4 source address to the destination address of 816 the first-fragment. If the destination address in the first-fragment 817 was multicast, the ETE instead sets the outer IPv4 source address to 818 an address assigned to the underlying IPv4 interface. The ETE 819 finally sends the SEAL-encapsulated ICMPv4 message to the ITE the 820 same as specified in Section 4.2.5, except that when the A bit in the 821 packet-in-error is not set the ETE sends the messages subject to rate 822 limiting since it is not entirely critical that all fragmentation be 823 reported to the ITE. 825 4.3.4. SEAL-Layer Reassembly 827 Following IPv4 reassembly, for SEAL packets with (E=1; Segment=0) in 828 the SEAL header (other than SEAL-encapsulated ICMPv4 messages), the 829 ETE discards the packet and sends a SEAL-encapsulated ICMPv4 830 Parameter Problem message with pointer set to the flags field in the 831 SEAL header back to the ITE subject to rate limiting exactly as for 832 SEAL-encapsulated ICMPv4 Fragmentation Needed messages (see: Section 833 4.3.3). For all other SEAL packets, the ETE adds the packet to a 834 SEAL-Layer pending-reassembly queue if either the M bit or the 835 Segment field in the SEAL header is non-zero. If the packet arrived 836 as multiple IPv4 fragments and with D=1 in the SEAL header, or if any 837 of the SEAL segments had E=1 in the SEAL header, the ETE marks the 838 packet and/or pending reassembly queue as "discard following 839 reassembly". 841 The ETE performs SEAL-layer reassembly through simple in-order 842 concatenation of the encapsulated segments from N consecutive SEAL 843 protocol packets from the same mid-layer packet. SEAL-layer 844 reassembly requires the ETE to maintain a cache of recently received 845 SEAL packet segments for a hold time that would allow for reasonable 846 inter-segment delays. The ETE uses a SEAL maximum segment lifetime 847 of 15 seconds for this purpose, i.e., the time after which it will 848 discard an incomplete reassembly. However, the ETE should also 849 actively discard any pending reassemblies that clearly have no 850 opportunity for completion, e.g., when a considerable number of new 851 SEAL packets have been received before a packet that completes a 852 pending reassembly has arrived. 854 The ETE reassembles the mid-layer packet segments in SEAL protocol 855 packets that contain Segment numbers 0 through N-1, with M=1/0 in 856 non-final/final segments, respectively, and with consecutive SEAL_ID 857 values. That is, for an N-segment mid-layer packet, reassembly 858 entails the concatenation of the SEAL-encapsulated segments with 859 (Segment 0, SEAL_ID i), followed by (Segment 1, SEAL_ID ((i + 1) mod 860 2^32)), etc. up to (Segment N-1, SEAL_ID ((i + N-1) mod 2^32)). (For 861 SEAL encapsulations specifically designed for traversal of IPv4 NATs, 862 the ETE instead uses only a 16-bit SEAL_ID value, and uses mod 2^16 863 arithmetic to associate the segments of the same packet.) 865 4.3.5. Delivering Packets to Upper Layers 867 Following SEAL-layer reassembly, if the packet had the value "No Next 868 Header" in the SEAL header's Next Header field, or if the packet was 869 marked "discard following reassembly" and IPv4 fragmentation was 870 experienced, the ETE silently discards the reassembled mid-layer 871 packet. In the same manner, the ETE silently discards any 872 reassembled mid-layer packet larger than 2KB that did not arrive as a 873 single packet that did not undergo SEAL segmentation nor IPv4 874 fragmentation. 876 Next, if the ETE determines that the inner packet cannot be processed 877 further it prepares an appropriate SEAL-encapsulated ICMPv4 error 878 message (if the inner packet itself was not an ICMP message), sends 879 the error message back to the ITE subject to rate limiting and drops 880 the packet. Otherwise, the ETE delivers the inner packet to the 881 upper layer protocol indicated in the Next Header field. 883 5. SEAL Protocol Specification - Transport Mode 885 Section 4 specifies the operation of SEAL in "tunnel mode", i.e., 886 when there is both an inner and outer IP layer and with a SEAL 887 encapsulation layer between. However, the SEAL protocol can also be 888 used in a "transport mode" of operation in which the inner layer 889 corresponds to a transport layer protocol (e.g., UDP, TCP, etc.) 890 instead of an inner IP layer. 892 For example, two TCP endpoints connected to the same subnetwork 893 region can negotiate the use of transport-mode SEAL for a connection 894 by inserting a 'SEAL_OPTION' TCP option during the connection 895 establishment phase. If both TCPs agree on the use of SEAL, their 896 protocol messages will be carriaged as TCP/SEAL/IPv4 and the 897 connection will be serviced by the SEAL protocol using TCP (nstead of 898 an encapsulating tunnel endpoint) as the transport layer protocol. 899 The SEAL protocol for transport mode otherwise observes the same 900 specifications as for Section 4. 902 6. Link Requirements 904 Subnetwork designers are strongly encouraged to follow the 905 recommendations in [RFC3819] when configuring link MTUs, where all 906 IPv4 links SHOULD configure a minimum MTU of 576 bytes. Links that 907 cannot configure an MTU of at least 576 bytes (e.g., due to 908 performance characteristics) SHOULD implement transparent link-layer 909 segmentation and reassembly such that an MTU of at least 576 can 910 still be presented to the IPv4 layer. 912 In the case that a fast IPv4 link within the subnetwork configures an 913 unusually small MTU, the ITE can sense this smaller value through 914 implicit probing and should reduce S_MRU to a size that would avoid 915 IPv4 fragmentation. However, the ITE must take care not to reduce 916 S_MRU to such small a value that original sources experience an MTU 917 underflow and hence an unusable path. (For example, IPv6 sources 918 must see a minimum path MTU of 1280 bytes.) 920 7. End System Requirements 922 SEAL provides robust mechanisms for returning PTB messages to the 923 original source, however end systems that send unfragmentable IP 924 packets larger than 1500 bytes are strongly encouraged to use 925 Packetization Layer Path MTU Discovery per [RFC4821]. 927 8. Router Requirements 929 IPv4 routers within the subnetwork are strongly encouraged to 930 implement IPv4 fragmentation such that the first fragment is the 931 largest and approximately the size of the underlying link MTU. 933 9. IANA Considerations 935 SEAL_PROTO, SEAL_PORT and SEAL_OPTION are taken from their respective 936 range of experimental values documented in [RFC3692][RFC4727]. These 937 values are for experimentation purposes only, and not to be used for 938 any kind of deployments (i.e., they are not to be shipped in any 939 products). This document therefore has no actions for IANA. 941 10. Security Considerations 943 Unlike IPv4 fragmentation, overlapping fragment attacks are not 944 possible due to the requirement that SEAL segments be non- 945 overlapping. 947 An amplification/reflection attack is possible when an attacker sends 948 IPv4 first-fragments with spoofed source addresses to an ETE, 949 resulting in a stream of ICMPv4 Fragmentation Needed messages 950 returned to a victim ITE. The encapsulated segment of the spoofed 951 IPv4 first-fragment provides mitigation for the ITE to detect and 952 discard spurious ICMPv4 Fragmentation Needed messages. 954 The SEAL header is sent in-the-clear (outside of any IPsec/ESP 955 encapsulations) the same as for the IPv4 header. As for IPv6 956 extension headers, the SEAL header is protected only by L2 integrity 957 checks and is not covered under any L3 integrity checks. 959 11. Related Work 961 Section 3.1.7 of [RFC2764] provides a high-level sketch for 962 supporting large tunnel MTUs via a tunnel-level segmentation and 963 reassembly capability to avoid IP level fragmentation, which is in 964 part the same approach used by tunnel-mode SEAL. SEAL could 965 therefore be considered as a fully-functioned manifestation of the 966 method postulated by that informational reference, however SEAL also 967 supports other modes of operation including transport-mode and 968 duplicate packet detection. 970 Section 3 of[RFC4459] describes inner and outer fragmentation at the 971 tunnel endpoints as alternatives for accommodating the tunnel MTU, 972 however the SEAL protocol specifies a mid-layer segmentation and 973 reassembly capability that is distinct from both inner and outer 974 fragmentation. 976 Section 4 of [RFC2460] specifies a method for inserting and 977 processing extension headers between the base IPv6 header and 978 transport layer protocol data. The SEAL header is in fact inserted 979 and processed in exactly the same manner. 981 The concepts of path MTU determination through the report of 982 fragmentation and extending the IP Identification field were first 983 proposed in deliberations of the TCP-IP mailing list and the Path MTU 984 Discovery Working Group (MTUDWG) during the late 1980's and early 985 1990's. SEAL supports a report fragmentation capability using bits 986 in an extension header (the original proposal used a spare bit in the 987 IP header) and supports ID extension through a 16 bit field in an 988 extension header (the original proposal used a new IP option). An 989 historical analysis of the evolution of these concepts as well as the 990 development of the eventual path MTU discovery mechanism for IP 991 appears in Appendix A of this document. 993 12. SEAL Advantages over Classical Methods 995 The SEAL approach offers a number of distinct advantages over the 996 classical path MTU discovery methods[RFC1191] [RFC1981]: 998 1. Classical path MTU discovery *always* results in packet loss when 999 an MTU restriction is encountered. Using SEAL, IPv4 1000 fragmentation provides a short-term interim mechanism for 1001 ensuring that packets are delivered while SEAL adjusts its packet 1002 sizing parameters. 1004 2. Classical path MTU discovery requires that routers generate an 1005 ICMP PTB message for *all* packets lost due to an MTU 1006 restriction; this situation is exacerbated at high data rates and 1007 becomes severe for in-the-network tunnels that service many 1008 communicating end systems. Since SEAL ensures that packets no 1009 larger than S_MRU are delivered, however, it is sufficient for 1010 the ETE to return ICMPv4 Fragmentation Needed messages subject to 1011 rate limiting and not for every packet-in-error. 1013 3. Classical path MTU may require several iterations of dropping 1014 packets and returning ICMP PTB messasges until an acceptable path 1015 MTU value is determined. Under normal circumstances, SEAL 1016 determines the correct packet sizing parameters in a single 1017 iteration. 1019 4. Using SEAL, ordinary packets serve as implicit probes without 1020 exposing data to unnecessary loss. SEAL also provides an 1021 explicit probing mode not available in the classic methods. 1023 5. Using SEAL, ICMP error messages are encapsulated in an outer SEAL 1024 header such that packet-filtering network middleboxes can 1025 distinguish them from "raw" ICMP messages that may be generated 1026 by an attacker. 1028 6. Most importantly, all SEAL packets have a 32 bit Identification 1029 value that can be used for duplicate packet detection purposes 1030 and to match ICMP error messages with actual packets sent without 1031 requiring per-packet state. Moreover, the SEAL ITE can be 1032 configured to accept ICMP feedback only from the legitimate ETE, 1033 hence the packet spoofing-related denial of service attack 1034 vectors open to the classical methods are eliminated. 1036 In summary, the SEAL approach represents an architecturally superior 1037 method for ensuring that packets of various sizes are either 1038 delivered or deterministically dropped. When end systems use their 1039 own end-to-end MTU determination mechanisms [RFC4821], the SEAL 1040 advantages are further enhanced. 1042 13. Acknowledgments 1044 The following individuals are acknowledged for helpful comments and 1045 suggestions: Jari Arkko, Fred Baker, Iljitsch van Beijnum, Teco Boot, 1046 Bob Braden, Brian Carpenter, Steve Casner, Ian Chakeres, Remi Denis- 1047 Courmont, Aurnaud Ebalard, Gorry Fairhurst, Joel Halpern, John 1048 Heffner, Thomas Henderson, Bob Hinden, Christian Huitema, Joe Macker, 1049 Matt Mathis, Dan Romascanu, Dave Thaler, Joe Touch, Magnus 1050 Westerlund, Robin Whittle, James Woodyatt and members of the Boeing 1051 PhantomWorks DC&NT group. 1053 Path MTU determination through the report of fragmentation was first 1054 proposed by Charles Lynn on the TCP-IP mailing list in 1987. 1055 Extending the IP identification field was first proposed by Steve 1056 Deering on the MTUDWG mailing list in 1989. 1058 14. References 1060 14.1. Normative References 1062 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, 1063 September 1981. 1065 [RFC0792] Postel, J., "Internet Control Message Protocol", STD 5, 1066 RFC 792, September 1981. 1068 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1069 Requirement Levels", BCP 14, RFC 2119, March 1997. 1071 [RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1072 (IPv6) Specification", RFC 2460, December 1998. 1074 14.2. Informative References 1076 [FOLK] C, C., D, D., and k. k, "Beyond Folklore: Observations on 1077 Fragmented Traffic", December 2002. 1079 [FRAG] Kent, C. and J. Mogul, "Fragmentation Considered Harmful", 1080 October 1987. 1082 [MTUDWG] "IETF MTU Discovery Working Group mailing list, 1083 gatekeeper.dec.com/pub/DEC/WRL/mogul/mtudwg-log, November 1084 1989 - February 1995.". 1086 [RFC1063] Mogul, J., Kent, C., Partridge, C., and K. McCloghrie, "IP 1087 MTU discovery options", RFC 1063, July 1988. 1089 [RFC1191] Mogul, J. and S. Deering, "Path MTU discovery", RFC 1191, 1090 November 1990. 1092 [RFC1981] McCann, J., Deering, S., and J. Mogul, "Path MTU Discovery 1093 for IP version 6", RFC 1981, August 1996. 1095 [RFC2003] Perkins, C., "IP Encapsulation within IP", RFC 2003, 1096 October 1996. 1098 [RFC2004] Perkins, C., "Minimal Encapsulation within IP", RFC 2004, 1099 October 1996. 1101 [RFC2764] Gleeson, B., Heinanen, J., Lin, A., Armitage, G., and A. 1102 Malis, "A Framework for IP Based Virtual Private 1103 Networks", RFC 2764, February 2000. 1105 [RFC2923] Lahey, K., "TCP Problems with Path MTU Discovery", 1106 RFC 2923, September 2000. 1108 [RFC3692] Narten, T., "Assigning Experimental and Testing Numbers 1109 Considered Useful", BCP 82, RFC 3692, January 2004. 1111 [RFC3819] Karn, P., Bormann, C., Fairhurst, G., Grossman, D., 1112 Ludwig, R., Mahdavi, J., Montenegro, G., Touch, J., and L. 1114 Wood, "Advice for Internet Subnetwork Designers", BCP 89, 1115 RFC 3819, July 2004. 1117 [RFC4213] Nordmark, E. and R. Gilligan, "Basic Transition Mechanisms 1118 for IPv6 Hosts and Routers", RFC 4213, October 2005. 1120 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 1121 Internet Protocol", RFC 4301, December 2005. 1123 [RFC4380] Huitema, C., "Teredo: Tunneling IPv6 over UDP through 1124 Network Address Translations (NATs)", RFC 4380, 1125 February 2006. 1127 [RFC4459] Savola, P., "MTU and Fragmentation Issues with In-the- 1128 Network Tunneling", RFC 4459, April 2006. 1130 [RFC4727] Fenner, B., "Experimental Values In IPv4, IPv6, ICMPv4, 1131 ICMPv6, UDP, and TCP Headers", RFC 4727, November 2006. 1133 [RFC4821] Mathis, M. and J. Heffner, "Packetization Layer Path MTU 1134 Discovery", RFC 4821, March 2007. 1136 [RFC4963] Heffner, J., Mathis, M., and B. Chandler, "IPv4 Reassembly 1137 Errors at High Data Rates", RFC 4963, July 2007. 1139 [TCP-IP] "TCP-IP mailing list archives, 1140 http://www-mice.cs.ucl.ac.uk/multimedia/mist/tcpip, May 1141 1987 - May 1990.". 1143 Appendix A. Historic Evolution of PMTUD 1145 (Taken from 'draft-templin-v6v4-ndisc-01.txt'; written 10/30/2002): 1147 The topic of Path MTU discovery (PMTUD) saw a flurry of discussion 1148 and numerous proposals in the late 1980's through early 1990. The 1149 initial problem was posed by Art Berggreen on May 22, 1987 in a 1150 message to the TCP-IP discussion group [TCP-IP]. The discussion that 1151 followed provided significant reference material for [FRAG]. An IETF 1152 Path MTU Discovery Working Group [MTUDWG] was formed in late 1989 1153 with charter to produce an RFC. Several variations on a very few 1154 basic proposals were entertained, including: 1156 1. Routers record the PMTUD estimate in ICMP-like path probe 1157 messages (proposed in [FRAG] and later [RFC1063]) 1159 2. The destination reports any fragmentation that occurs for packets 1160 received with the "RF" (Report Fragmentation) bit set (Steve 1161 Deering's 1989 adaptation of Charles Lynn's Nov. 1987 proposal) 1163 3. A hybrid combination of 1) and Charles Lynn's Nov. 1987 proposal 1164 (straw RFC draft by McCloughrie, Fox and Mogul on Jan 12, 1990) 1166 4. Combination of the Lynn proposal with TCP (Fred Bohle, Jan 30, 1167 1990) 1169 5. Fragmentation avoidance by setting "IP_DF" flag on all packets 1170 and retransmitting if ICMPv4 "fragmentation needed" messages 1171 occur (Geof Cooper's 1987 proposal; later adapted into [RFC1191] 1172 by Mogul and Deering). 1174 Option 1) seemed attractive to the group at the time, since it was 1175 believed that routers would migrate more quickly than hosts. Option 1176 2) was a strong contender, but repeated attempts to secure an "RF" 1177 bit in the IPv4 header from the IESG failed and the proponents became 1178 discouraged. 3) was abandoned because it was perceived as too 1179 complicated, and 4) never received any apparent serious 1180 consideration. Proposal 5) was a late entry into the discussion from 1181 Steve Deering on Feb. 24th, 1990. The discussion group soon 1182 thereafter seemingly lost track of all other proposals and adopted 1183 5), which eventually evolved into [RFC1191] and later [RFC1981]. 1185 In retrospect, the "RF" bit postulated in 2) is not needed if a 1186 "contract" is first established between the peers, as in proposal 4) 1187 and a message to the MTUDWG mailing list from jrd@PTT.LCS.MIT.EDU on 1188 Feb 19. 1990. These proposals saw little discussion or rebuttal, and 1189 were dismissed based on the following the assertions: 1191 o routers upgrade their software faster than hosts 1193 o PCs could not reassemble fragmented packets 1195 o Proteon and Wellfleet routers did not reproduce the "RF" bit 1196 properly in fragmented packets 1198 o Ethernet-FDDI bridges would need to perform fragmentation (i.e., 1199 "translucent" not "transparent" bridging) 1201 o the 16-bit IP_ID field could wrap around and disrupt reassembly at 1202 high packet arrival rates 1204 The first four assertions, although perhaps valid at the time, have 1205 been overcome by historical events leaving only the final to 1206 consider. But, [FOLK] has shown that IP_ID wraparound simply does 1207 not occur within several orders of magnitude the reassembly timeout 1208 window on high-bandwidth networks. 1210 (Authors 2/11/08 note: this final point was based on a loose 1211 interpretation of [FOLK], and is more accurately addressed in 1212 [RFC4963].) 1214 Appendix B. Reliability Extensions 1216 The SEAL header includes an Extension (E) bit that is set to zero for 1217 the purpose of this specification. This bit may be used by future 1218 updates to this specification for the purpose of improved reliability 1219 in the face of loss due to congestion, singal intermittence, etc. 1220 Automatic Repeat-ReQuest (ARQ) mechanisms are used to ensure relaible 1221 delivery between the endpoints of physical links (e.g., on-link 1222 neighbors in an IEEE 802.11 network) as well as between the endpoints 1223 of an end-to-end transport (e.g., the endpoints of a TCP connection). 1224 However, ARQ mechanisms are poorly suited to in-the-network elements 1225 such as the SEAL ITE and ETE, since retransmission of lost segments 1226 would require unacceptable state maintenance at the ITE and would 1227 result in packet reordering within the subnetwork. 1229 Instead, Forward Error Correction (FEC) mechanisms may be specified 1230 in future updates to this specification for the purpose of improved 1231 reliability. Such mechanisms may entail the ITE performing proactive 1232 transmissions of redundant data, e.g., by sending multiple copies of 1233 the same data. Signaling from the ETE (e.g., by sending SEAL- 1234 encapsulated ICMPv4 Source Quench messages) may be specified in a 1235 future document as a means for the ETE to dynamically update the ITE 1236 of changing FEC conditions. 1238 Author's Address 1240 Fred L. Templin (editor) 1241 Boeing Phantom Works 1242 P.O. Box 3707 1243 Seattle, WA 98124 1244 USA 1246 Email: fltemplin@acm.org 1248 Full Copyright Statement 1250 Copyright (C) The IETF Trust (2008). 1252 This document is subject to the rights, licenses and restrictions 1253 contained in BCP 78, and except as set forth therein, the authors 1254 retain all their rights. 1256 This document and the information contained herein are provided on an 1257 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS 1258 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND 1259 THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS 1260 OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF 1261 THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED 1262 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. 1264 Intellectual Property 1266 The IETF takes no position regarding the validity or scope of any 1267 Intellectual Property Rights or other rights that might be claimed to 1268 pertain to the implementation or use of the technology described in 1269 this document or the extent to which any license under such rights 1270 might or might not be available; nor does it represent that it has 1271 made any independent effort to identify any such rights. Information 1272 on the procedures with respect to rights in RFC documents can be 1273 found in BCP 78 and BCP 79. 1275 Copies of IPR disclosures made to the IETF Secretariat and any 1276 assurances of licenses to be made available, or the result of an 1277 attempt made to obtain a general license or permission for the use of 1278 such proprietary rights by implementers or users of this 1279 specification can be obtained from the IETF on-line IPR repository at 1280 http://www.ietf.org/ipr. 1282 The IETF invites any interested party to bring to its attention any 1283 copyrights, patents or patent applications, or other proprietary 1284 rights that may cover technology that may be required to implement 1285 this standard. Please address the information to the IETF at 1286 ietf-ipr@ietf.org.