| 1 |
<pre>Network Working Group M. Degermark |
| 2 |
Request for Comments: 2507 Lulea University of Technology/SICS |
| 3 |
Category: Standards Track B. Nordgren |
| 4 |
Lulea University of Technology/Telia Research AB |
| 5 |
S. Pink |
| 6 |
Lulea University of Technology/SICS |
| 7 |
February 1999 |
| 8 |
|
| 9 |
|
| 10 |
<span class="h1">IP Header Compression</span> |
| 11 |
|
| 12 |
Status of this Memo |
| 13 |
|
| 14 |
This document specifies an Internet standards track protocol for the |
| 15 |
Internet community, and requests discussion and suggestions for |
| 16 |
improvements. Please refer to the current edition of the "Internet |
| 17 |
Official Protocol Standards" (STD 1) for the standardization state |
| 18 |
and status of this protocol. Distribution of this memo is unlimited. |
| 19 |
|
| 20 |
Copyright Notice |
| 21 |
|
| 22 |
Copyright (C) The Internet Society (1999). All Rights Reserved. |
| 23 |
|
| 24 |
Abstract |
| 25 |
|
| 26 |
This document describes how to compress multiple IP headers and TCP |
| 27 |
and UDP headers per hop over point to point links. The methods can be |
| 28 |
applied to of IPv6 base and extension headers, IPv4 headers, TCP and |
| 29 |
UDP headers, and encapsulated IPv6 and IPv4 headers. |
| 30 |
|
| 31 |
Headers of typical UDP or TCP packets can be compressed down to 4-7 |
| 32 |
octets including the 2 octet UDP or TCP checksum. This largely |
| 33 |
removes the negative impact of large IP headers and allows efficient |
| 34 |
use of bandwidth on low and medium speed links. |
| 35 |
|
| 36 |
The compression algorithms are specifically designed to work well |
| 37 |
over links with nontrivial packet-loss rates. Several wireless and |
| 38 |
modem technologies result in such links. |
| 39 |
|
| 40 |
TABLE OF CONTENTS |
| 41 |
|
| 42 |
<a href="#section-1">1</a>. Introduction..............................................<a href="#page-3">3</a> |
| 43 |
<a href="#section-2">2</a>. Terminology...............................................<a href="#page-5">5</a> |
| 44 |
<a href="#section-3">3</a>. Compression method........................................<a href="#page-7">7</a> |
| 45 |
<a href="#section-3.1">3.1</a>. Packet types.......................................<a href="#page-8">8</a> |
| 46 |
<a href="#section-3.2">3.2</a>. Lost packets in TCP packet streams.................<a href="#page-9">9</a> |
| 47 |
<a href="#section-3.3">3.3</a>. Lost packets in UDP and non-TCP packet streams....<a href="#page-10">10</a> |
| 48 |
<a href="#section-4">4</a>. Grouping packets into packet streams.....................<a href="#page-14">14</a> |
| 49 |
|
| 50 |
|
| 51 |
|
| 52 |
<span class="grey">Degermark, et. al. Standards Track [Page 1]</span></pre> |
| 53 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span> |
| 54 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 55 |
|
| 56 |
|
| 57 |
<a href="#section-4.1">4.1</a>. Guidelines for grouping packets...................<a href="#page-15">15</a> |
| 58 |
<a href="#section-5">5</a>. Size Issues..............................................<a href="#page-16">16</a> |
| 59 |
<a href="#section-5.1">5.1</a>. Context identifiers...............................<a href="#page-16">16</a> |
| 60 |
<a href="#section-5.2">5.2</a>. Size of the context...............................<a href="#page-17">17</a> |
| 61 |
<a href="#section-5.3">5.3</a>. Size of full headers..............................<a href="#page-18">18</a> |
| 62 |
<a href="#section-5.3.1">5.3.1</a>. Length fields in full TCP headers............<a href="#page-19">19</a> |
| 63 |
<a href="#section-5.3.2">5.3.2</a>. Length fields in full non-TCP headers........<a href="#page-19">19</a> |
| 64 |
<a href="#section-6">6</a>. Compressed Header Formats................................<a href="#page-20">20</a> |
| 65 |
<a href="#section-7">7</a>. Compression of subheaders................................<a href="#page-22">22</a> |
| 66 |
<a href="#section-7.1">7.1</a>. IPv6 Header.......................................<a href="#page-24">24</a> |
| 67 |
<a href="#section-7.2">7.2</a>. IPv6 Extension Headers............................<a href="#page-25">25</a> |
| 68 |
<a href="#section-7.3">7.3</a>. Options...........................................<a href="#page-25">25</a> |
| 69 |
<a href="#section-7.4">7.4</a>. Hop-by-hop Options Header.........................<a href="#page-26">26</a> |
| 70 |
<a href="#section-7.5">7.5</a>. Routing Header....................................<a href="#page-26">26</a> |
| 71 |
<a href="#section-7.6">7.6</a>. Fragment Header...................................<a href="#page-27">27</a> |
| 72 |
<a href="#section-7.7">7.7</a>. Destination Options Header........................<a href="#page-28">28</a> |
| 73 |
<a href="#section-7.8">7.8</a>. No Next Header....................................<a href="#page-29">29</a> |
| 74 |
<a href="#section-7.9">7.9</a>. Authentication Header.............................<a href="#page-29">29</a> |
| 75 |
<a href="#section-7.10">7.10</a>. Encapsulating Security Payload Header.............<a href="#page-29">29</a> |
| 76 |
<a href="#section-7.11">7.11</a>. UDP Header........................................<a href="#page-30">30</a> |
| 77 |
<a href="#section-7.12">7.12</a>. TCP Header........................................<a href="#page-30">30</a> |
| 78 |
<a href="#section-7.13">7.13</a>. IPv4 Header.......................................<a href="#page-33">33</a> |
| 79 |
<a href="#section-7.14">7.14</a> Minimal Encapsulation header......................<a href="#page-34">34</a> |
| 80 |
<a href="#section-8">8</a>. Changing context identifiers.............................<a href="#page-35">35</a> |
| 81 |
<a href="#section-9">9</a>. Rules for dropping or temporarily storing packets........<a href="#page-35">35</a> |
| 82 |
<a href="#section-10">10</a>. Low-loss header compression for TCP .....................<a href="#page-36">36</a> |
| 83 |
<a href="#section-10.1">10.1</a>. The "twice" algorithm............................<a href="#page-37">37</a> |
| 84 |
<a href="#section-10.2">10.2</a>. Header Requests..................................<a href="#page-37">37</a> |
| 85 |
<a href="#section-11">11</a>. Links that reorder packets...............................<a href="#page-38">38</a> |
| 86 |
<a href="#section-11.1">11.1</a>. Reordering in non-TCP packet streams.............<a href="#page-39">39</a> |
| 87 |
<a href="#section-11.2">11.2</a>. Reordering in TCP packet streams.................<a href="#page-39">39</a> |
| 88 |
<a href="#section-12">12</a>. Hooks for additional header compression..................<a href="#page-40">40</a> |
| 89 |
<a href="#section-13">13</a>. Demultiplexing...........................................<a href="#page-41">41</a> |
| 90 |
<a href="#section-14">14</a>. Configuration Parameters.................................<a href="#page-42">42</a> |
| 91 |
<a href="#section-15">15</a>. Implementation Status....................................<a href="#page-43">43</a> |
| 92 |
<a href="#section-16">16</a>. Acknowledgments..........................................<a href="#page-44">44</a> |
| 93 |
<a href="#section-17">17</a>. Security Considerations..................................<a href="#page-44">44</a> |
| 94 |
<a href="#section-18">18</a>. Authors' Addresses.......................................<a href="#page-45">45</a> |
| 95 |
<a href="#section-19">19</a>. References...............................................<a href="#page-46">46</a> |
| 96 |
<a href="#section-20">20</a>. Full Copyright Statement.................................<a href="#page-47">47</a> |
| 97 |
|
| 98 |
|
| 99 |
|
| 100 |
|
| 101 |
|
| 102 |
|
| 103 |
|
| 104 |
|
| 105 |
|
| 106 |
|
| 107 |
|
| 108 |
<span class="grey">Degermark, et. al. Standards Track [Page 2]</span></pre> |
| 109 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span> |
| 110 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 111 |
|
| 112 |
|
| 113 |
<span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>. Introduction</span> |
| 114 |
|
| 115 |
There are several reasons to do header compression on low- or |
| 116 |
medium-speed links. Header compression can |
| 117 |
|
| 118 |
* Improve interactive response time |
| 119 |
|
| 120 |
For very low-speed links, echoing of characters may take longer |
| 121 |
than 100-200 ms because of the time required to transmit large |
| 122 |
headers. 100-200 ms is the maximum time people can tolerate |
| 123 |
without feeling that the system is sluggish. |
| 124 |
|
| 125 |
* Allow using small packets for bulk data with good line efficiency |
| 126 |
|
| 127 |
This is important when interactive (for example Telnet) and bulk |
| 128 |
traffic (for example FTP) is mixed because the bulk data should be |
| 129 |
carried in small packets to decrease the waiting time when a |
| 130 |
packet with interactive data is caught behind a bulk data packet. |
| 131 |
|
| 132 |
Using small packet sizes for the FTP traffic in this case is a |
| 133 |
global solution to a local problem. It will increase the load on |
| 134 |
the network as it has to deal with many small packets. A better |
| 135 |
solution might be to locally fragment the large packets over the |
| 136 |
slow link. |
| 137 |
|
| 138 |
* Allow using small packets for delay sensitive low data-rate traffic |
| 139 |
|
| 140 |
For such applications, for example voice, the time to fill a |
| 141 |
packet with data is significant if packets are large. To get low |
| 142 |
end-to-end delay small packets are preferred. Without header |
| 143 |
compression, the smallest possible IPv6/UDP headers (48 octets) |
| 144 |
consume 19.2 kbit/s with a packet rate of 50 packets/s. 50 |
| 145 |
packets/s is equivalent to having 20 ms worth of voice samples in |
| 146 |
each packet. IPv4/UDP headers consumes 11.2 kbit/s at 50 |
| 147 |
packets/s. Tunneling or routing headers, for example to support |
| 148 |
mobility, will increase the bandwidth consumed by headers by 10-20 |
| 149 |
kbit/s. This should be compared with the bandwidth required for |
| 150 |
the actual sound samples, for example 13 kbit/s with GSM encoding. |
| 151 |
Header compression can reduce the bandwidth needed for headers |
| 152 |
significantly, in the example to about 1.7 kbit/s. This enables |
| 153 |
higher quality voice transmission over 14.4 and 28.8 kbit/s |
| 154 |
modems. |
| 155 |
|
| 156 |
* Decrease header overhead. |
| 157 |
|
| 158 |
A common size of TCP segments for bulk transfers over medium-speed |
| 159 |
links is 512 octets today. When TCP segments are tunneled, for |
| 160 |
example because Mobile IP is used, the IPv6/IPv6/TCP header is 100 |
| 161 |
|
| 162 |
|
| 163 |
|
| 164 |
<span class="grey">Degermark, et. al. Standards Track [Page 3]</span></pre> |
| 165 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span> |
| 166 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 167 |
|
| 168 |
|
| 169 |
octets. Header compression will decrease the header overhead for |
| 170 |
IPv6/TCP from 19.5 per cent to less than 1 per cent, and for |
| 171 |
tunneled IPv4/TCP from 11.7 to less than 1 per cent. This is a |
| 172 |
significant gain for line-speeds as high as a few Mbit/s. |
| 173 |
|
| 174 |
The IPv6 specification prescribes path MTU discovery, so with IPv6 |
| 175 |
bulk TCP transfers should use segments larger than 512 octets when |
| 176 |
possible. Still, with 1400 octet segments (<a href="./rfc894">RFC 894</a> Ethernet |
| 177 |
encapsulation allows 1500 octet payloads, of which 100 octets are |
| 178 |
used for IP headers), header compression reduces IPv6 header |
| 179 |
overhead from 7.1% to 0.4%. |
| 180 |
|
| 181 |
* Reduce packet loss rate over lossy links. |
| 182 |
|
| 183 |
Because fewer bits are sent per packet, the packet loss rate will |
| 184 |
be lower for a given bit-error rate. This results in higher |
| 185 |
throughput for TCP as the sending window can open up more between |
| 186 |
losses, and in fewer lost packets for UDP. |
| 187 |
|
| 188 |
The mechanisms described here are intended for a point-to-point link. |
| 189 |
However, care has been taken to allow extensions for multi-access |
| 190 |
links and multicast. |
| 191 |
|
| 192 |
Headers that can be compressed include TCP, UDP, IPv4, and IPv6 base |
| 193 |
and extension headers. For TCP packets, the mechanisms of Van |
| 194 |
Jacobson [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>] are used to recover from loss. Two additional |
| 195 |
mechanisms that increase the efficiency of VJ header compression over |
| 196 |
lossy links are also described. For non-TCP packets, compression |
| 197 |
slow-start and periodic header refreshes allow minimal periods of |
| 198 |
packet discard after loss of a header that changes the context. There |
| 199 |
are hooks for adding header compression schemes on top of UDP, for |
| 200 |
example compression of RTP headers. |
| 201 |
|
| 202 |
Header compression relies on many fields being constant or changing |
| 203 |
seldomly in consecutive packets belonging to the same packet stream. |
| 204 |
Fields that do not change between packets need not be transmitted at |
| 205 |
all. Fields that change often with small and/or predictable values, |
| 206 |
e.g., TCP sequence numbers, can be encoded incrementally so that the |
| 207 |
number of bits needed for these fields decrease significantly. Only |
| 208 |
fields that change often and randomly, e.g., checksums or |
| 209 |
authentication data, need to be transmitted in every header. |
| 210 |
|
| 211 |
The general principle of header compression is to occasionally send a |
| 212 |
packet with a full header; subsequent compressed headers refer to the |
| 213 |
context established by the full header and may contain incremental |
| 214 |
changes to the context. |
| 215 |
|
| 216 |
|
| 217 |
|
| 218 |
|
| 219 |
|
| 220 |
<span class="grey">Degermark, et. al. Standards Track [Page 4]</span></pre> |
| 221 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span> |
| 222 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 223 |
|
| 224 |
|
| 225 |
This header compression scheme does not require that all packets in |
| 226 |
the same stream passes over the compressed link. However, for TCP |
| 227 |
streams the difference between subsequent headers can become more |
| 228 |
irregular and the compression rate can decrease. Neither is it |
| 229 |
required that corresponding TCP data and acknowledgment packets |
| 230 |
traverse the link in opposite directions. |
| 231 |
|
| 232 |
This header compression scheme is useful on first-hop or last-hop |
| 233 |
links as well as links in the middle of the network. When many packet |
| 234 |
streams (several hundred) traverse the link, a phenomenon that could |
| 235 |
be called CID thrashing could occur, where headers seldom can be |
| 236 |
matched with an existing context and have to be sent uncompressed or |
| 237 |
as full headers. It is up to an implementation to use techniques such |
| 238 |
as hysteresis to ensure that the packet streams that give the highest |
| 239 |
compression rates keep their context. Such techniques are more |
| 240 |
likely to be needed in the middle of the network. |
| 241 |
|
| 242 |
<span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>. Terminology</span> |
| 243 |
|
| 244 |
This section explains some terms used in this document. |
| 245 |
|
| 246 |
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", |
| 247 |
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this |
| 248 |
document are to be interpreted as described in <a href="./rfc2119">RFC 2119</a>. |
| 249 |
|
| 250 |
Subheader |
| 251 |
|
| 252 |
An IPv6 base header, an IPv6 extension header, an IPv4 header, a |
| 253 |
UDP header, or a TCP header. |
| 254 |
|
| 255 |
Header |
| 256 |
|
| 257 |
A chain of subheaders. |
| 258 |
|
| 259 |
Compress |
| 260 |
|
| 261 |
The act of reducing the size of a header by removing header fields |
| 262 |
or reducing the size of header fields. This is done in a way such |
| 263 |
that a decompressor can reconstruct the header if its context |
| 264 |
state is identical to the context state used when compressing the |
| 265 |
header. |
| 266 |
|
| 267 |
Decompress |
| 268 |
|
| 269 |
The act of reconstructing a compressed header. |
| 270 |
|
| 271 |
|
| 272 |
|
| 273 |
|
| 274 |
|
| 275 |
|
| 276 |
<span class="grey">Degermark, et. al. Standards Track [Page 5]</span></pre> |
| 277 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span> |
| 278 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 279 |
|
| 280 |
|
| 281 |
Context identifier (CID) |
| 282 |
|
| 283 |
A small unique number identifying the context that should be used |
| 284 |
to decompress a compressed header. Carried in full headers and |
| 285 |
compressed headers. |
| 286 |
|
| 287 |
Context |
| 288 |
|
| 289 |
The state which the compressor uses to compress a header and the |
| 290 |
decompressor uses to decompress a header. The context is the |
| 291 |
uncompressed version of the last header sent (compressor) or |
| 292 |
received (decompressor) over the link, except for fields in the |
| 293 |
header that are included "as-is" in compressed headers or can be |
| 294 |
inferred from, e.g., the size of the link-level frame. |
| 295 |
|
| 296 |
The context for a packet stream is associated with a context |
| 297 |
identifier. The context for non-TCP packet streams is also |
| 298 |
associated with a generation. |
| 299 |
|
| 300 |
Generation |
| 301 |
|
| 302 |
For non-TCP packet streams, each new version of the context for a |
| 303 |
given CID is associated with a generation: a small number that is |
| 304 |
incremented whenever the context associated with that CID changes. |
| 305 |
Carried by full and compressed non-TCP headers. |
| 306 |
|
| 307 |
Packet stream |
| 308 |
|
| 309 |
A sequence of packets whose headers are similar and share context. |
| 310 |
For example, headers in a TCP packet stream have the same source |
| 311 |
and final destination address, and the same port numbers in the |
| 312 |
TCP header. Similarly, headers in a UDP packet stream have the |
| 313 |
same source and destination address, and the same port numbers in |
| 314 |
the UDP header. |
| 315 |
|
| 316 |
Full header (header refresh) |
| 317 |
|
| 318 |
An uncompressed header that updates or refreshes the context for a |
| 319 |
packet stream. It carries a CID that will be used to identify the |
| 320 |
context. |
| 321 |
|
| 322 |
Full headers for non-TCP packet streams also carry the generation |
| 323 |
of the context they update or refresh. |
| 324 |
|
| 325 |
Regular header |
| 326 |
|
| 327 |
A normal, uncompressed, header. Does not carry CID or generation |
| 328 |
association. |
| 329 |
|
| 330 |
|
| 331 |
|
| 332 |
<span class="grey">Degermark, et. al. Standards Track [Page 6]</span></pre> |
| 333 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span> |
| 334 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 335 |
|
| 336 |
|
| 337 |
Incorrect decompression |
| 338 |
|
| 339 |
When a compressed and then decompressed header is different from |
| 340 |
the uncompressed header. Usually due to mismatching context |
| 341 |
between the compressor and decompressor or bit errors during |
| 342 |
transmission of the compressed header. |
| 343 |
|
| 344 |
Differential coding |
| 345 |
|
| 346 |
A compression technique where the compressed value of a header |
| 347 |
field is the difference between the current value of the field and |
| 348 |
the value of the same field in the previous header belonging to |
| 349 |
the same packet stream. A decompressor can thus obtain the value |
| 350 |
of the field by adding the value in the compressed header to its |
| 351 |
context. This technique is used for TCP streams but not for non- |
| 352 |
TCP streams. |
| 353 |
|
| 354 |
<span class="h2"><a class="selflink" id="section-3" href="#section-3">3</a>. Compression method</span> |
| 355 |
|
| 356 |
Much of the header information stays the same over the life-time of a |
| 357 |
packet stream. For non-TCP packet streams almost all fields of the |
| 358 |
headers are constant. For TCP many fields are constant and others |
| 359 |
change with small and predictable values. |
| 360 |
|
| 361 |
To initiate compression of the headers of a packet stream, a full |
| 362 |
header carrying a context identifier, CID, is transmitted over the |
| 363 |
link. The compressor and decompressor store most fields of this full |
| 364 |
header as context. The context consists of the fields of the header |
| 365 |
whose values are constant and thus need not be sent over the link at |
| 366 |
all, or change little between consecutive headers so that it uses |
| 367 |
fewer bits to send the difference from the previous value compared to |
| 368 |
sending the absolute value. |
| 369 |
|
| 370 |
Any change in fields that are expected to be constant in a packet |
| 371 |
stream will cause the compressor to send a full header again to |
| 372 |
update the context at the decompressor. As long as the context is the |
| 373 |
same at compressor and decompressor, headers can be decompressed to |
| 374 |
be exactly as they were before compression. However, if a full header |
| 375 |
or compressed header is lost during transmission, the context of the |
| 376 |
decompressor may become obsolete as it is not updated properly. |
| 377 |
Compressed headers will then be decompressed incorrectly. |
| 378 |
|
| 379 |
IPv6 is not meant to be used over links that can deliver a |
| 380 |
significant fraction of damaged packets to the IPv6 module. This |
| 381 |
means that links must have a very low bit-error rate or that link- |
| 382 |
level frames must be protected by strong checksums, forward error |
| 383 |
correction or something of that nature. Header compression SHOULD |
| 384 |
not be used for IPv4 without strong link-level checksums. Damaged |
| 385 |
|
| 386 |
|
| 387 |
|
| 388 |
<span class="grey">Degermark, et. al. Standards Track [Page 7]</span></pre> |
| 389 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span> |
| 390 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 391 |
|
| 392 |
|
| 393 |
frames will thus be discarded by the link layer. The link layer |
| 394 |
implementation might indicate to the header compression module that a |
| 395 |
frame was damaged, but it cannot say what packet stream it belonged |
| 396 |
to as it might be the CID that is damaged. Moreover, frames may |
| 397 |
disappear without the link layer implementation's knowledge, for |
| 398 |
example if the link is a multi-hop link where frames can be dropped |
| 399 |
due to congestion at each hop. The kind of link errors that a header |
| 400 |
compression module should deal with and protect against will thus be |
| 401 |
packet loss. |
| 402 |
|
| 403 |
So a header compression scheme needs mechanisms to update the context |
| 404 |
at the decompressor and to detect or avoid incorrect decompression. |
| 405 |
These mechanisms are very different for TCP and non-TCP streams, and |
| 406 |
are described in sections <a href="#section-3.2">3.2</a> and <a href="#section-3.3">3.3</a>. |
| 407 |
|
| 408 |
The compression mechanisms in this document assume that packets are |
| 409 |
not reordered between the compressor and decompressor. If the link |
| 410 |
|
| 411 |
does reorder, <a href="#section-11">section 11</a> describes mechanisms for ordering the |
| 412 |
packets before decompression. It is also assumed that the link-layer |
| 413 |
implementation can provide the length of packets, and that there is |
| 414 |
no padding in UDP packets or tunneled packets. |
| 415 |
|
| 416 |
<span class="h3"><a class="selflink" id="section-3.1" href="#section-3.1">3.1</a>. Packet types</span> |
| 417 |
|
| 418 |
This compression method uses four packet types in addition to the |
| 419 |
IPv4 and IPv6 packet types. The combination of link-level packet |
| 420 |
type and the value of the first four bits of the packet uniquely |
| 421 |
determines the packet type. Details on how these packet types are |
| 422 |
represented are in <a href="#section-13">section 13</a>. |
| 423 |
|
| 424 |
FULL_HEADER - indicates a packet with an uncompressed header, |
| 425 |
including a CID and, if not a TCP packet, a generation. It |
| 426 |
establishes or refreshes the context for the packet stream |
| 427 |
identified by the CID. |
| 428 |
|
| 429 |
COMPRESSED_NON_TCP - indicates a non-TCP packet with a compressed |
| 430 |
header. The compressed header consists of a CID identifying what |
| 431 |
context to use for decompression, a generation to detect an |
| 432 |
inconsistent context and the randomly changing fields of the |
| 433 |
header. |
| 434 |
|
| 435 |
COMPRESSED_TCP - indicates a packet with a compressed TCP header, |
| 436 |
containing a CID, a flag octet indentifying what fields have |
| 437 |
changed, and the changed fields encoded as the difference from |
| 438 |
the previous value. |
| 439 |
|
| 440 |
|
| 441 |
|
| 442 |
|
| 443 |
|
| 444 |
<span class="grey">Degermark, et. al. Standards Track [Page 8]</span></pre> |
| 445 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span> |
| 446 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 447 |
|
| 448 |
|
| 449 |
COMPRESSED_TCP_NODELTA - indicates a packet with a compressed TCP |
| 450 |
header where all fields that are normally sent as the difference |
| 451 |
to the previous value are instead sent as-is. This packet type |
| 452 |
is only sent as the response to a header request from the |
| 453 |
decompressor. It must not be sent as the result of a |
| 454 |
retransmission. |
| 455 |
|
| 456 |
In addition to the packet types used for compression, regular IPv4 |
| 457 |
and IPv6 packets are used whenever a compressor decides to not |
| 458 |
compress a packet. An additional packet type may be used to speed up |
| 459 |
repair of TCP streams over links where the decompressor can send |
| 460 |
packets to the compressor. |
| 461 |
|
| 462 |
CONTEXT_STATE - indicates a special packet sent from the |
| 463 |
decompressor to the compressor to communicate a list of (TCP) |
| 464 |
CIDs for which synchronization has been lost. This packet is only |
| 465 |
sent over a single link so it requires no IP header. The format |
| 466 |
is shown in <a href="#section-10.2">section 10.2</a>. |
| 467 |
|
| 468 |
<span class="h3"><a class="selflink" id="section-3.2" href="#section-3.2">3.2</a>. Lost packets in TCP packet streams</span> |
| 469 |
|
| 470 |
Since TCP headers are compressed using the difference from the |
| 471 |
previous TCP header, loss of a packet with a compressed or full |
| 472 |
header will cause subsequent compressed headers to be decompressed |
| 473 |
incorrectly because the context used for decompression was not |
| 474 |
incremented properly. |
| 475 |
|
| 476 |
Loss of a compressed TCP header will cause the TCP sequence numbers |
| 477 |
of subsequently decompressed TCP headers to be off by k, where k is |
| 478 |
the size of the lost segment. Such incorrectly decompressed TCP |
| 479 |
headers will be discarded by the TCP receiver as the TCP checksum |
| 480 |
reliably catches "off-by-k" errors in the sequence numbers for |
| 481 |
plausible k. |
| 482 |
|
| 483 |
TCP's repair mechanisms will eventually retransmit the discarded |
| 484 |
segment and the compressor peeks into the TCP headers to detect when |
| 485 |
TCP retransmits. When this happens, the compressor sends a full |
| 486 |
header on the assumption that the retransmission was due to |
| 487 |
mismatching compression state at the decompressor. [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>] has a |
| 488 |
good explanation of this mechanism. |
| 489 |
|
| 490 |
The mechanisms of <a href="#section-10">section 10</a> should be used to speed up the repair of |
| 491 |
the context. This is important over medium speed links with high |
| 492 |
packet loss rates, for example wireless. Losing a timeout's worth of |
| 493 |
packets due to inconsistent context after each packet lost over the |
| 494 |
link is not acceptable, especially when the TCP connection is over |
| 495 |
the wide area. |
| 496 |
|
| 497 |
|
| 498 |
|
| 499 |
|
| 500 |
<span class="grey">Degermark, et. al. Standards Track [Page 9]</span></pre> |
| 501 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span> |
| 502 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 503 |
|
| 504 |
|
| 505 |
<span class="h3"><a class="selflink" id="section-3.3" href="#section-3.3">3.3</a>. Lost packets in UDP and other non-TCP packet streams</span> |
| 506 |
|
| 507 |
Incorrectly decompressed headers of UDP packets and other non-TCP |
| 508 |
packets are not so well-protected by checksums as TCP packets. There |
| 509 |
are no sequence numbers that become "off-by-k" and virtually |
| 510 |
guarantees a failed checksum as there are for TCP. The UDP checksum |
| 511 |
only covers payload, UDP header, and pseudo header. The pseudo |
| 512 |
header includes the source and destination addresses, the transport |
| 513 |
protocol type and the length of the transport packet. Except for |
| 514 |
those fields, large parts of the IPv6 header are not covered by the |
| 515 |
UDP checksum. Moreover, other non-TCP headers lack checksums |
| 516 |
altogether, for example fragments. |
| 517 |
|
| 518 |
In order to safely avoid incorrect decompression of non-TCP headers, |
| 519 |
each version of the context for non-TCP packet streams is identified |
| 520 |
by a generation, a small number that is carried by the full headers |
| 521 |
that establish and refresh the context. Compressed headers carry the |
| 522 |
generation value of the context that were used to compress them. |
| 523 |
When a decompressor sees that a compressed header carries a |
| 524 |
generation value other than the generation of its context for that |
| 525 |
packet stream, the context is not up to date and the packet must be |
| 526 |
discarded or stored until a full header establishes correct context. |
| 527 |
|
| 528 |
Differential coding is not used for non-TCP streams, so compressed |
| 529 |
non-TCP headers do not change the context. Thus, loss of a |
| 530 |
compressed header does not invalidate subsequent packets with |
| 531 |
compressed headers. Moreover, the generation changes only when the |
| 532 |
context of a full header is different from the context of the |
| 533 |
previous full header. This means that losing a full header will make |
| 534 |
the context of the decompressor obsolete only when the full header |
| 535 |
would actually have changed the context. |
| 536 |
|
| 537 |
The generation field is 6 bits long so the generation value repeats |
| 538 |
itself after 64 changes to the context. To avoid incorrect |
| 539 |
decompression after error bursts or other temporary disruptions, the |
| 540 |
compressor must not reuse the same generation value after a shorter |
| 541 |
time than MIN_WRAP seconds. A decompressor which has been |
| 542 |
disconnected MIN_WRAP seconds or more must wait for the next full |
| 543 |
header before decompressing. A compressor must wait at least MIN_WRAP |
| 544 |
seconds after booting before compressing non-TCP headers. Instead of |
| 545 |
reusing a generation value too soon, a compressor may switch to |
| 546 |
another CID or send regular headers until MIN_WRAP seconds have |
| 547 |
passed. The value of MIN_WRAP is found in <a href="#section-14">section 14</a>. |
| 548 |
|
| 549 |
|
| 550 |
|
| 551 |
|
| 552 |
|
| 553 |
|
| 554 |
|
| 555 |
|
| 556 |
<span class="grey">Degermark, et. al. Standards Track [Page 10]</span></pre> |
| 557 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span> |
| 558 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 559 |
|
| 560 |
|
| 561 |
<span class="h4"><a class="selflink" id="section-3.3.1" href="#section-3.3.1">3.3.1</a>. Compression Slow-Start</span> |
| 562 |
|
| 563 |
To allow the decompressor to recover quickly from loss of a full |
| 564 |
header that would have changed the context, full headers are sent |
| 565 |
periodically with an exponentially increasing period after a change |
| 566 |
in the context. This technique avoids an exchange of messages between |
| 567 |
compressor and decompressor used by other compression schemes, such |
| 568 |
as in [<a href="./rfc1553" title=""Compressing IPX Headers Over WAN Media (CIPX)"">RFC-1553</a>]. Such exchanges can be costly for wireless mobiles |
| 569 |
as more power is consumed by the transmitter and delay can be |
| 570 |
introduced by switching between sending and receiving. Moreover, |
| 571 |
techniques that require an exchange of messages cannot be used over |
| 572 |
simplex links, such as direct-broadcast satellite channels or cable |
| 573 |
TV systems, and are hard to adapt to multicast over multi-access |
| 574 |
links. |
| 575 |
|
| 576 |
|.|..|....|........|................|.............................. |
| 577 |
^ |
| 578 |
Change Sent packets: | with full header, . with compressed header |
| 579 |
|
| 580 |
The picture shows how packets are sent after change. The compressor |
| 581 |
keeps a variable for each non-TCP packet stream, F_PERIOD, that keeps |
| 582 |
track of how many compressed headers may be sent between full |
| 583 |
headers. When the headers of a non-TCP packet stream change so that |
| 584 |
its context changes, a full header is sent and F_PERIOD is set to |
| 585 |
one. After sending F_PERIOD compressed headers, a full header is |
| 586 |
sent. F_PERIOD is doubled each time a full header is sent during |
| 587 |
compression slow-start. |
| 588 |
|
| 589 |
<span class="h4"><a class="selflink" id="section-3.3.2" href="#section-3.3.2">3.3.2</a>. Periodic Header Refreshes</span> |
| 590 |
|
| 591 |
To avoid losing too many packets if a receiver has lost its context, |
| 592 |
there is an upper limit, F_MAX_PERIOD, on the number of non-TCP |
| 593 |
packets with compressed headers that may be sent between header |
| 594 |
refreshes. If a packet is to be sent and F_MAX_PERIOD compressed |
| 595 |
headers have been sent since the last full header for this packet |
| 596 |
stream was sent, a full header must be sent. |
| 597 |
|
| 598 |
To avoid long periods of disconnection for low data rate packet |
| 599 |
streams, there is also an upper bound, F_MAX_TIME, on the time |
| 600 |
between full headers in a non-TCP packet stream. If a packet is to be |
| 601 |
sent and more than F_MAX_TIME seconds have passed since the last full |
| 602 |
header was sent for this packet stream, a full header must be sent. |
| 603 |
The values of F_MAX_PERIOD and F_MAX_TIME are found in <a href="#section-14">section 14</a>. |
| 604 |
|
| 605 |
|
| 606 |
|
| 607 |
|
| 608 |
|
| 609 |
|
| 610 |
|
| 611 |
|
| 612 |
<span class="grey">Degermark, et. al. Standards Track [Page 11]</span></pre> |
| 613 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-12" ></span> |
| 614 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 615 |
|
| 616 |
|
| 617 |
<span class="h4"><a class="selflink" id="section-3.3.3" href="#section-3.3.3">3.3.3</a>. Rules for sending Full Headers</span> |
| 618 |
|
| 619 |
The following pseudo code can be used by the compressor to determine |
| 620 |
when to send a full header for a non-TCP packet stream. The code |
| 621 |
maintains two variables: |
| 622 |
|
| 623 |
C_NUM -- a count of the number of compressed headers sent |
| 624 |
since the last full header was sent. |
| 625 |
F_LAST -- the time of sending the last full header. |
| 626 |
|
| 627 |
and uses the functions |
| 628 |
|
| 629 |
current_time() return the current time |
| 630 |
min(a,b) return the smallest of a and b |
| 631 |
|
| 632 |
the procedures send_full_header(), increment_generation_value(), |
| 633 |
and send_compressed_header() |
| 634 |
do the obvious thing. |
| 635 |
|
| 636 |
if ( <this header changes the context> ) |
| 637 |
|
| 638 |
C_NUM := 0; |
| 639 |
F_LAST := current_time(); |
| 640 |
F_PERIOD := 1; |
| 641 |
increment_generation_value(); |
| 642 |
send_full_header(); |
| 643 |
|
| 644 |
elseif ( C_NUM >= F_PERIOD ) |
| 645 |
|
| 646 |
C_NUM := 0; |
| 647 |
F_LAST := current_time(); |
| 648 |
F_PERIOD := min(2 * F_PERIOD, F_MAX_PERIOD); |
| 649 |
send_full_header(); |
| 650 |
|
| 651 |
elseif ( current_time() > F_LAST + F_MAX_TIME ) |
| 652 |
|
| 653 |
C_NUM := 0; |
| 654 |
F_LAST := current_time(); |
| 655 |
send_full_header(); |
| 656 |
|
| 657 |
else |
| 658 |
|
| 659 |
C_NUM := C_NUM + 1 |
| 660 |
send_compressed_header(); |
| 661 |
|
| 662 |
endif |
| 663 |
|
| 664 |
|
| 665 |
|
| 666 |
|
| 667 |
|
| 668 |
<span class="grey">Degermark, et. al. Standards Track [Page 12]</span></pre> |
| 669 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-13" ></span> |
| 670 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 671 |
|
| 672 |
|
| 673 |
<span class="h4"><a class="selflink" id="section-3.3.4" href="#section-3.3.4">3.3.4</a>. Cost of sending Header Refreshes</span> |
| 674 |
|
| 675 |
If every f'th packet carries a full header, H is the size of a full |
| 676 |
header, and C is the size of a compressed header, the average header |
| 677 |
size is |
| 678 |
|
| 679 |
(H-C)/f + C |
| 680 |
|
| 681 |
For f > 1, the average header size is (H-C)/f larger than a |
| 682 |
compressed header. |
| 683 |
|
| 684 |
In a diagram where the average header size is plotted for various f |
| 685 |
values, there is a distinct knee in the curve, i.e., there is a limit |
| 686 |
beyond which further increasing f gives diminishing returns. |
| 687 |
F_MAX_PERIOD should be chosen to be a frequency well to the right of |
| 688 |
the knee of the curve. For typical sizes of H and C, say 48 octets |
| 689 |
for the full header (IPv6/UDP) and 4 octets for the compressed |
| 690 |
header, setting F_MAX_PERIOD > 44 means that full headers will |
| 691 |
contribute less than an octet to the average header size. With a |
| 692 |
four-address routing header, F_MAX_PERIOD > 115 will have the same |
| 693 |
effect. |
| 694 |
|
| 695 |
The default F_MAX_PERIOD value of 256 (<a href="#section-14">section 14</a>) puts the full |
| 696 |
header frequency well to the right of the knee and means that full |
| 697 |
headers will typically contribute considerably less than an octet to |
| 698 |
the average header size. For H = 48 and C = 4, full headers |
| 699 |
contribute about 1.4 bits to the average header size after reaching |
| 700 |
the steady-state header refresh frequency determined by the default |
| 701 |
|
| 702 |
F_MAX_PERIOD. 1.4 bits is a very small overhead. |
| 703 |
|
| 704 |
After a change in the context, the exponential backoff scheme will |
| 705 |
initially send full headers frequently. The default F_MAX_PERIOD |
| 706 |
will be reached after nine full headers and 255 compressed headers |
| 707 |
have been sent. This is equivalent to a little over 5 seconds for a |
| 708 |
typical voice stream with 20 ms worth of voice samples per packet. |
| 709 |
|
| 710 |
During the whole backoff period, full headers contribute 1.5 octets |
| 711 |
to the average header size when H = 48 and C = 4. For 20 ms voice |
| 712 |
samples, it takes less than 1.3 seconds until full headers contribute |
| 713 |
less than one octet to the average header size, and during these |
| 714 |
initial 1.3 seconds full headers add less than 4 octets to the |
| 715 |
average header size. The cost of the exponential backoff is not |
| 716 |
great and as the headers of non-TCP packet streams are expected to |
| 717 |
change seldomly, it will be amortized over a long time. |
| 718 |
|
| 719 |
|
| 720 |
|
| 721 |
|
| 722 |
|
| 723 |
|
| 724 |
<span class="grey">Degermark, et. al. Standards Track [Page 13]</span></pre> |
| 725 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-14" ></span> |
| 726 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 727 |
|
| 728 |
|
| 729 |
The cost of header refreshes in terms of bandwidth are higher than |
| 730 |
similar costs for hard state schemes like [<a href="./rfc1553" title=""Compressing IPX Headers Over WAN Media (CIPX)"">RFC-1553</a>] where full |
| 731 |
headers must be acknowledged by the decompressor before compressed |
| 732 |
headers may be sent. Such schemes typically send one full header plus |
| 733 |
a few control messages when the context changes. Hard state schemes |
| 734 |
require more types of protocol messages and an exchange of messages |
| 735 |
is necessary. Hard state schemes also need to deal explicitly with |
| 736 |
various error conditions that soft state handles automatically, for |
| 737 |
instance the case of one party disappearing unexpectedly, a common |
| 738 |
situation on wireless links where mobiles may go out of range of the |
| 739 |
base station. |
| 740 |
|
| 741 |
The major advantage of the soft state scheme is that no handshakes |
| 742 |
are needed between compressor and decompressor, so the scheme can be |
| 743 |
used over simplex links. The costs in terms of bandwidth are higher |
| 744 |
than for hard state schemes, but the simplicity of the decompressor, |
| 745 |
the simplicity of the protocol, and the lack of handshakes between |
| 746 |
compressor and decompressor justifies this small cost. Moreover, soft |
| 747 |
state schemes are more easily extended to multicast over multi-access |
| 748 |
links, for example radio links. |
| 749 |
|
| 750 |
<span class="h2"><a class="selflink" id="section-4" href="#section-4">4</a>. Grouping packets into packet streams</span> |
| 751 |
|
| 752 |
This section explains how packets MAY be grouped together into packet |
| 753 |
streams for compression. To achieve the best compression rates, |
| 754 |
packets SHOULD be grouped together such that packets in the same |
| 755 |
packet stream have similar headers. If this grouping fails, header |
| 756 |
compression performance will be bad, since the compression algorithm |
| 757 |
can rarely utilize the existing context for the packet stream and |
| 758 |
full headers must be sent frequently. |
| 759 |
|
| 760 |
Grouping is done by the compressor. A compressor may use whatever |
| 761 |
criterion it finds appropriate to group packets into packet streams. |
| 762 |
To determine what packet stream a packet belongs to, a compressor MAY |
| 763 |
|
| 764 |
a) examine the compressible chain of subheaders (see <a href="#section-7">section 7</a>), |
| 765 |
|
| 766 |
b) examine the contents of an upper layer protocol header that |
| 767 |
follows the compressible chain of subheaders, for example ICMP |
| 768 |
headers, DVMRP headers, or tunneled IPX headers, |
| 769 |
|
| 770 |
c) use information obtained from a resource manager, for example if a |
| 771 |
resource manager requests compression for a particular packet |
| 772 |
stream and provides a way to identify packets belonging to that |
| 773 |
packet stream, |
| 774 |
|
| 775 |
|
| 776 |
|
| 777 |
|
| 778 |
|
| 779 |
|
| 780 |
<span class="grey">Degermark, et. al. Standards Track [Page 14]</span></pre> |
| 781 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-15" ></span> |
| 782 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 783 |
|
| 784 |
|
| 785 |
d) use any other relevant information, for example if routes flap and |
| 786 |
the hop limit (TTL) field in a packet stream changes frequently |
| 787 |
between n and n+k, a compressor may choose to group the packets |
| 788 |
into two different packet streams. |
| 789 |
|
| 790 |
A compressor is also free not to group packets into packet streams |
| 791 |
for compression, letting some packets keep their regular headers and |
| 792 |
passing them through unmodified. |
| 793 |
|
| 794 |
As long as the rules for when to send full headers for a non-TCP |
| 795 |
packet stream are followed and subheaders are compressed as specified |
| 796 |
in this document, the decompressor is able to reconstruct a |
| 797 |
compressed header correctly regardless of how packets are grouped |
| 798 |
into packet streams. |
| 799 |
|
| 800 |
<span class="h3"><a class="selflink" id="section-4.1" href="#section-4.1">4.1</a> Guidelines for grouping packets</span> |
| 801 |
|
| 802 |
In this section we give OPTIONAL guidelines for how a compressor may |
| 803 |
group packets into packet streams for compression. |
| 804 |
|
| 805 |
Defining fields |
| 806 |
|
| 807 |
The defining fields of a header should be present and identical in |
| 808 |
all packets belonging to the same packet stream. These fields are |
| 809 |
marked DEF in <a href="#section-7">section 7</a>. The defining fields include the flow |
| 810 |
label, source and destination addresses of IP headers, final |
| 811 |
destination address in routing headers, the next header fields |
| 812 |
(for IPv6), the protocol field (IPv4), port numbers (UDP and TCP), |
| 813 |
and the SPI in authentication and encryption headers. |
| 814 |
|
| 815 |
Fragmented packets |
| 816 |
|
| 817 |
Fragmented and unfragmented packets should never be grouped |
| 818 |
together in the same packet stream. The Identification field of |
| 819 |
the Fragment header or IPv4 header should not be used to identify |
| 820 |
the packet stream. If it was, the first fragment of a new packet |
| 821 |
would cause a compression slow-start. |
| 822 |
|
| 823 |
No field after a Fragment Header, or an IPv4 header for a |
| 824 |
fragment, should be used for grouping purposes. |
| 825 |
|
| 826 |
Upper protocol identification |
| 827 |
|
| 828 |
The first next header field identifying a header not described in |
| 829 |
<a href="#section-7">section 7</a> should be used for identifying packet streams, i.e., all |
| 830 |
packets with the same DEF fields and the same upper protocol |
| 831 |
should be grouped together. |
| 832 |
|
| 833 |
|
| 834 |
|
| 835 |
|
| 836 |
<span class="grey">Degermark, et. al. Standards Track [Page 15]</span></pre> |
| 837 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-16" ></span> |
| 838 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 839 |
|
| 840 |
|
| 841 |
TTL field (Hop Limit field) |
| 842 |
|
| 843 |
A sophisticated implementation might monitor the TTL (Hop Limit) |
| 844 |
field and if it changes frequently use it as a DEF field. This can |
| 845 |
occur when there are frequent route flaps so that packets traverse |
| 846 |
different paths through the internet. |
| 847 |
|
| 848 |
Traffic Class field (IPv6), Type of Service field (IPv4) |
| 849 |
|
| 850 |
It is possible that the Traffic Class field of the IPv6 header and |
| 851 |
the Type of Service of the IPv4 header will change frequently |
| 852 |
between packets with otherwise identical DEF fields. A |
| 853 |
sophisticated implementation should watch out for this and be |
| 854 |
prepared to use these fields as defining fields. |
| 855 |
|
| 856 |
When IP packets are tunneled they are encapsulated with an additional |
| 857 |
IP header at the tunnel entry point and then sent to the tunnel |
| 858 |
endpoint. To group such packets into packet streams, the inner |
| 859 |
headers should also be examined to determine the packet stream. If |
| 860 |
this is not done, full headers will be sent each time the headers of |
| 861 |
the inner IP packet changes. So when a packet is tunneled, the |
| 862 |
identifying fields of the inner subheaders should be considered in |
| 863 |
addition to the identifying fields of the initial IP header. |
| 864 |
|
| 865 |
An implementation can use other fields for identification than the |
| 866 |
ones described here. If too many fields are used for identification, |
| 867 |
performance might suffer because more CIDs will be used and the wrong |
| 868 |
CIDs might be reused when new flows need CIDs. If too few fields are |
| 869 |
used for identification, performance might suffer because there are |
| 870 |
too frequent changes to the context. |
| 871 |
|
| 872 |
We stress that these guidelines are educated guesses. When IPv6 is |
| 873 |
widely deployed and IPv6 traffic can be analyzed, we might find that |
| 874 |
other grouping algorithms perform better. We also stress that if the |
| 875 |
grouping fails, the result will be bad performance but not incorrect |
| 876 |
decompression. The decompressor can do its task regardless of how the |
| 877 |
grouping algorithm works. |
| 878 |
|
| 879 |
<span class="h2"><a class="selflink" id="section-5" href="#section-5">5</a>. Size Issues</span> |
| 880 |
|
| 881 |
<span class="h3"><a class="selflink" id="section-5.1" href="#section-5.1">5.1</a>. Context Identifiers</span> |
| 882 |
|
| 883 |
Context identifiers can be 8 or 16 bits long. Their size is not |
| 884 |
relevant for finding the context. An 8-bit CID with value two and a |
| 885 |
16-bit CID with value two are equivalent. |
| 886 |
|
| 887 |
|
| 888 |
|
| 889 |
|
| 890 |
|
| 891 |
|
| 892 |
<span class="grey">Degermark, et. al. Standards Track [Page 16]</span></pre> |
| 893 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-17" ></span> |
| 894 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 895 |
|
| 896 |
|
| 897 |
The CID spaces for TCP and non-TCP are separate, so a TCP CID and a |
| 898 |
non-TCP CID never identify the same context. Even if they have the |
| 899 |
same value. This doubles the available CID space while using the same |
| 900 |
number of bits for CIDs. It is always possible to tell whether a |
| 901 |
full or compressed header is for a TCP or non-TCP packet, so no |
| 902 |
mixups can occur. |
| 903 |
|
| 904 |
Non-TCP compressed headers encode the size of the CID using one bit |
| 905 |
in the second octet of the compressed header. The 8-bit CID allows a |
| 906 |
minimum compressed header size of 2 octets for non-TCP packets, the |
| 907 |
CID uses the first octet and the size bit and the 6-bit Generation |
| 908 |
value fit in the second octet. |
| 909 |
|
| 910 |
For TCP the only available CID size is 8 bits as in [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>]. 8 |
| 911 |
bits is probably sufficient as TCP connections are always point-to- |
| 912 |
point. |
| 913 |
|
| 914 |
The 16 bit CID size may not be needed for point-to-point links; it is |
| 915 |
intended for use on multi-access links where a larger CID space may |
| 916 |
be needed for efficient selection of CIDs. |
| 917 |
|
| 918 |
The major difficulty with multi-access links is that several |
| 919 |
compressors share the CID space of a decompressor. CIDs can no |
| 920 |
longer be selected independently by the compressors as collisions may |
| 921 |
occur. This problem may be resolved by letting the decompressors |
| 922 |
have a separate CID space for each compressor. Having separate CID |
| 923 |
spaces requires that decompressors can identify which compressor sent |
| 924 |
the compressed packet, perhaps by utilizing link-layer information as |
| 925 |
to who sent the link-layer frame. If such information is not |
| 926 |
available, all compressors on the multi-access link may be |
| 927 |
enumerated, automatically or otherwise, and supply their number as |
| 928 |
part of the CID. This latter method requires a large CID space. |
| 929 |
|
| 930 |
<span class="h3"><a class="selflink" id="section-5.2" href="#section-5.2">5.2</a>. Size of the context</span> |
| 931 |
|
| 932 |
The size of the context SHOULD be limited to simplify implementation |
| 933 |
of compressor and decompressor, and put a limit on their memory |
| 934 |
requirements. However, there is no upper limit on the size of an |
| 935 |
IPv6 header as the chain of extension headers can be arbitrarily |
| 936 |
long. This is a problem as the context is essentially a stored |
| 937 |
header. |
| 938 |
|
| 939 |
The configurable parameter MAX_HEADER (see <a href="#section-14">section 14</a>) represents the |
| 940 |
maximum size of the context, expressed as the maximum sized header |
| 941 |
that can be stored as context. When a header is larger than |
| 942 |
MAX_HEADER, only part of it is stored as context. An implementation |
| 943 |
MUST NOT compress more than the initial MAX_HEADER octets of a |
| 944 |
header. An implementation MUST NOT partially compress a subheader. |
| 945 |
|
| 946 |
|
| 947 |
|
| 948 |
<span class="grey">Degermark, et. al. Standards Track [Page 17]</span></pre> |
| 949 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-18" ></span> |
| 950 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 951 |
|
| 952 |
|
| 953 |
Thus, the part of the header that is stored as context and is |
| 954 |
compressed is the longest initial sequence of entire subheaders that |
| 955 |
is not larger than MAX_HEADER octets. |
| 956 |
|
| 957 |
<span class="h3"><a class="selflink" id="section-5.3" href="#section-5.3">5.3</a>. Size of full headers</span> |
| 958 |
|
| 959 |
It is desirable to avoid increasing the size of packets with full |
| 960 |
headers beyond their original size, as their size may be optimized |
| 961 |
for the MTU of the link. Since we assume that the link layer |
| 962 |
implementation provides the length of packets, we can use the length |
| 963 |
fields in full headers to pass the values of the CID and the |
| 964 |
generation to the decompressor. |
| 965 |
|
| 966 |
This requires that the link-layer must not add padding to the |
| 967 |
payload, at least not padding that can be delivered to the |
| 968 |
destination link user. It is also required that no extra padding is |
| 969 |
added after UDP data or in tunneled packets. This allows values of |
| 970 |
length fields to be calculated from the length of headers and the |
| 971 |
length of the link-layer frame. |
| 972 |
|
| 973 |
The generation requires one octet and the CID may require up to 2 |
| 974 |
octets. There are length fields of 2 octets in the IPv6 Base Header, |
| 975 |
the IPv4 header, and the UDP header. |
| 976 |
|
| 977 |
A full TCP header will thus have at least 2 octets available in the |
| 978 |
IP header to pass the 8 bit CID, which is sufficient. There will be |
| 979 |
more than two octets available if there is more than one IP header. |
| 980 |
|
| 981 |
[<a id="ref-RFC-1144">RFC-1144</a>] uses the 8 bit Protocol field of the IPv4 header to pass |
| 982 |
the CID. We cannot use the corresponding method as the sequence of |
| 983 |
IPv6 extension headers is not fixed and CID values are not disjoint |
| 984 |
from the legal values of Next Header fields. |
| 985 |
|
| 986 |
An IPv6/UDP or IPv4/UDP packet will have 4 octets available to pass |
| 987 |
the generation and the CID, so all CID sizes may be used. Fragmented |
| 988 |
or encrypted packet streams may have only 2 octets available to pass |
| 989 |
the generation and CID. Thus, 8-bit CIDs may be the only CID sizes |
| 990 |
that can be used for such packet streams. When IPv6/IPv4 or |
| 991 |
IPv4/IPv6 tunneling is used, there will be at least 4 octets |
| 992 |
available, and both CID sizes may be used. |
| 993 |
|
| 994 |
The generation value is passed in the higher order octet of the first |
| 995 |
length field in the full header. When only one length field is |
| 996 |
available, the 8-bit CID is passed in the low order octet. When two |
| 997 |
length fields are available, the lowest two octets of the CID are |
| 998 |
passed in the second length field and the low order octet of the |
| 999 |
first length field carries the highest octet of the CID. |
| 1000 |
|
| 1001 |
|
| 1002 |
|
| 1003 |
|
| 1004 |
<span class="grey">Degermark, et. al. Standards Track [Page 18]</span></pre> |
| 1005 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-19" ></span> |
| 1006 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1007 |
|
| 1008 |
|
| 1009 |
<span class="h4"><a class="selflink" id="section-5.3.1" href="#section-5.3.1">5.3.1</a>. Use of length fields in full TCP headers</span> |
| 1010 |
|
| 1011 |
Use of first length field: |
| 1012 |
|
| 1013 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1014 |
Length field | LSB of pkt nr | CID | |
| 1015 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1016 |
|
| 1017 |
Use of second length field if available: |
| 1018 |
|
| 1019 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1020 |
Second length field | MSB of pkt nr | 0 | |
| 1021 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1022 |
|
| 1023 |
Pkt nr is short for packet sequence number, described in <a href="#section-11.2">section</a> |
| 1024 |
<a href="#section-11.2">11.2</a>. |
| 1025 |
|
| 1026 |
<span class="h4"><a class="selflink" id="section-5.3.2" href="#section-5.3.2">5.3.2</a>. Use of length fields in full non-TCP headers</span> |
| 1027 |
|
| 1028 |
Full non-TCP headers with 8-bit CID: |
| 1029 |
|
| 1030 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1031 |
First length field |0|D| Generation| CID | |
| 1032 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1033 |
|
| 1034 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1035 |
Second length field (if avail.) | 0 | Data (if D=1) | |
| 1036 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1037 |
|
| 1038 |
Full non-TCP headers with 16-bit CID: |
| 1039 |
|
| 1040 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1041 |
First length field |1|D| Generation| Data (if D=1) | |
| 1042 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1043 |
|
| 1044 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1045 |
Second length field | CID | |
| 1046 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1047 |
|
| 1048 |
The first bit in the first length field indicates the length of the |
| 1049 |
CID. The Data field is zero if D is zero. The use of the D bit and |
| 1050 |
Data field is explained in <a href="#section-12">section 12</a>. |
| 1051 |
|
| 1052 |
|
| 1053 |
|
| 1054 |
|
| 1055 |
|
| 1056 |
|
| 1057 |
|
| 1058 |
|
| 1059 |
|
| 1060 |
<span class="grey">Degermark, et. al. Standards Track [Page 19]</span></pre> |
| 1061 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-20" ></span> |
| 1062 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1063 |
|
| 1064 |
|
| 1065 |
<span class="h2"><a class="selflink" id="section-6" href="#section-6">6</a>. Compressed Header Formats</span> |
| 1066 |
|
| 1067 |
This section uses some terminology (DELTA, RANDOM) defined in <a href="#section-7">section</a> |
| 1068 |
<a href="#section-7">7</a>. |
| 1069 |
|
| 1070 |
a) COMPRESSED_TCP format (similar to [<a href="./rfc1144">RFC 1144</a>]): |
| 1071 |
|
| 1072 |
+-+-+-+-+-+-+-+-+ |
| 1073 |
| CID | |
| 1074 |
+-+-+-+-+-+-+-+-+ |
| 1075 |
|R O I P S A W U| |
| 1076 |
+-+-+-+-+-+-+-+-+ |
| 1077 |
| | |
| 1078 |
+ TCP Checksum + |
| 1079 |
| | |
| 1080 |
+-+-+-+-+-+-+-+-+ |
| 1081 |
| RANDOM fields, if any (see <a href="#section-7">section 7</a>) (implied) |
| 1082 |
- - - - - - - - |
| 1083 |
| R-octet | (if R=1) |
| 1084 |
- - - - - - - - |
| 1085 |
| Urgent Pointer Value (if U=1) |
| 1086 |
- - - - - - - - |
| 1087 |
| Window Delta (if W=1) |
| 1088 |
- - - - - - - - |
| 1089 |
| Acknowledgment Number Delta (if A=1) |
| 1090 |
- - - - - - - - |
| 1091 |
| Sequence Number Delta (if S=1) |
| 1092 |
- - - - - - - - |
| 1093 |
| IPv4 Identification Delta (if I=1) |
| 1094 |
- - - - - - - - |
| 1095 |
| Options (if O=1) |
| 1096 |
- - - - - - - - |
| 1097 |
|
| 1098 |
|
| 1099 |
The latter flags in the second octet (IPSAWU) have the same meaning |
| 1100 |
as in [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>], regardless of whether the TCP segments are carried |
| 1101 |
by IPv6 or IPv4. The C bit has been eliminated because the CID is |
| 1102 |
always present. The context associated with the CID keeps track of |
| 1103 |
the IP version and what RANDOM fields are present. The order between |
| 1104 |
delta fields specified here is exactly as in [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>]. An |
| 1105 |
implementation will typically scan the context from the beginning and |
| 1106 |
insert the RANDOM fields in order. The RANDOM fields are thus placed |
| 1107 |
before the DELTA fields of the TCP header in the same order as they |
| 1108 |
occur in the original uncompressed header. |
| 1109 |
|
| 1110 |
|
| 1111 |
|
| 1112 |
|
| 1113 |
|
| 1114 |
|
| 1115 |
|
| 1116 |
<span class="grey">Degermark, et. al. Standards Track [Page 20]</span></pre> |
| 1117 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-21" ></span> |
| 1118 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1119 |
|
| 1120 |
|
| 1121 |
The I flag is zero unless an IPv4 header immediately precedes the TCP |
| 1122 |
header. The combined IPv4/TCP header is then compressed as a unit as |
| 1123 |
described in [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>]. Identification fields in IPv4 headers that |
| 1124 |
are not immediately followed by a TCP header are RANDOM. |
| 1125 |
|
| 1126 |
If the O flag is set, the Options of the TCP header were not the same |
| 1127 |
as in the previous header. The entire Option field are placed last in |
| 1128 |
the compressed TCP header. |
| 1129 |
|
| 1130 |
If the R flag is set, there were differences between the context and |
| 1131 |
the Reserved field (6 bits) in the TCP header or bit 6 or 7 of the |
| 1132 |
TOS octet (Traffic Class octet) in a IPv4 header (IPv6 header) that |
| 1133 |
immediately precedes the TCP header. An octet with the actual values |
| 1134 |
of the Reserved field and bit 6 and 7 of the TOS or Traffic Class |
| 1135 |
field is then placed immediately after the RANDOM fields. Bits 0-5 |
| 1136 |
of the passed octet is the actual value of the Reserved field, and |
| 1137 |
bits 6 and 7 are the actual values of bits 6 and 7 in the TOS or |
| 1138 |
Traffic Class field. If there is no preceding IP header, bits 6 and 7 |
| 1139 |
are 0. The octet passed with the R flag MUST NOT update the context. |
| 1140 |
|
| 1141 |
NOTE: The R-octet does not update the context because if it did, the |
| 1142 |
nTCP checksum would not guard the receiving TCP from erroneously |
| 1143 |
decompressed headers. Bits 6 and 7 of the TOS octet or Traffic Class |
| 1144 |
octet is expected to change frequently due to Explicit Congestion |
| 1145 |
Notification. |
| 1146 |
|
| 1147 |
See <a href="#section-7.12">section 7.12</a> and [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>] for further information on how to |
| 1148 |
compress TCP headers. |
| 1149 |
|
| 1150 |
b) COMPRESSED_TCP_NODELTA header format |
| 1151 |
|
| 1152 |
+-+-+-+-+-+-+-+-+ |
| 1153 |
| CID | |
| 1154 |
+-+-+-+-+-+-+-+-+ |
| 1155 |
| RANDOM fields, if any (see <a href="#section-7">section 7</a>) (implied) |
| 1156 |
+-+-+-+-+-+-+-+-+ |
| 1157 |
| Whole TCP header except for Port Numbers |
| 1158 |
+-+-+-+-+-+-+-+-+ |
| 1159 |
|
| 1160 |
|
| 1161 |
|
| 1162 |
|
| 1163 |
|
| 1164 |
|
| 1165 |
|
| 1166 |
|
| 1167 |
|
| 1168 |
|
| 1169 |
|
| 1170 |
|
| 1171 |
|
| 1172 |
<span class="grey">Degermark, et. al. Standards Track [Page 21]</span></pre> |
| 1173 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-22" ></span> |
| 1174 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1175 |
|
| 1176 |
|
| 1177 |
c) Compressed non-TCP header, 8 bit CID: |
| 1178 |
0 7 |
| 1179 |
+-+-+-+-+-+-+-+-+ |
| 1180 |
| CID | |
| 1181 |
+-+-+-+-+-+-+-+-+ |
| 1182 |
|0|D| Generation| |
| 1183 |
+-+-+-+-+-+-+-+-+ |
| 1184 |
| data | (if D=1) |
| 1185 |
- - - - - - - - |
| 1186 |
| RANDOM fields, if any (<a href="#section-7">section 7</a>) (implied) |
| 1187 |
- - - - - - - - |
| 1188 |
|
| 1189 |
d) Compressed non-TCP header, 16 bit CID: |
| 1190 |
0 7 |
| 1191 |
+-+-+-+-+-+-+-+-+ |
| 1192 |
| msb of CID | |
| 1193 |
+-+-+-+-+-+-+-+-+ |
| 1194 |
|1|D| Generation| |
| 1195 |
+-+-+-+-+-+-+-+-+ |
| 1196 |
| lsb of CID | |
| 1197 |
+-+-+-+-+-+-+-+-+ |
| 1198 |
| data | (if D=1) |
| 1199 |
- - - - - - - - |
| 1200 |
| RANDOM fields, if any (<a href="#section-7">section 7</a>) (implied) |
| 1201 |
- - - - - - - - |
| 1202 |
|
| 1203 |
The generation, CID and optional one octet data are followed by |
| 1204 |
relevant RANDOM fields (see <a href="#section-7">section 7</a>) as implied by the compression |
| 1205 |
state, placed in the same order as they occur in the original |
| 1206 |
uncompressed header, followed by the payload. |
| 1207 |
|
| 1208 |
<span class="h2"><a class="selflink" id="section-7" href="#section-7">7</a>. Compression of subheaders</span> |
| 1209 |
|
| 1210 |
This section gives rules for how the compressible chain of subheaders |
| 1211 |
is compressed. These rules MUST be followed. Subheaders that may be |
| 1212 |
compressed include IPv6 base and extension headers, TCP headers, UDP |
| 1213 |
headers, and IPv4 headers. The compressible chain of subheaders |
| 1214 |
extends from the beginning of the header |
| 1215 |
|
| 1216 |
a) up to but not including the first header that is not an IPv4 |
| 1217 |
header, an IPv6 base or extension header, a TCP header, or a UDP |
| 1218 |
header, or |
| 1219 |
|
| 1220 |
b) up to and including the first TCP header, UDP header, Fragment |
| 1221 |
Header, Encapsulating Security Payload Header, or IPv4 header for |
| 1222 |
a fragment, |
| 1223 |
|
| 1224 |
|
| 1225 |
|
| 1226 |
|
| 1227 |
|
| 1228 |
<span class="grey">Degermark, et. al. Standards Track [Page 22]</span></pre> |
| 1229 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-23" ></span> |
| 1230 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1231 |
|
| 1232 |
|
| 1233 |
whichever gives the shorter chain. For example, rules a) and b) both |
| 1234 |
fit a chain of subheaders that contain a Fragment Header and ends at |
| 1235 |
a tunneled IPX packet. Since rule b) gives a shorter chain, the |
| 1236 |
compressible chain of subheaders stops at the Fragment Header. |
| 1237 |
|
| 1238 |
The following subsections are a systematic classification of how all |
| 1239 |
fields in subheaders are expected to change. |
| 1240 |
|
| 1241 |
NOCHANGE The field is not expected to change. Any change means |
| 1242 |
that a full header MUST be sent to update the context. |
| 1243 |
|
| 1244 |
DELTA The field may change often but usually the difference |
| 1245 |
from the field in the previous header is small, so that |
| 1246 |
it is cheaper to send the change from the previous value |
| 1247 |
rather than the current value. This type of compression |
| 1248 |
is only used for TCP packet streams. |
| 1249 |
|
| 1250 |
RANDOM The field must be included "as-is" in compressed headers, |
| 1251 |
usually because it changes unpredictably. |
| 1252 |
|
| 1253 |
INFERRED The field contains a value that can be inferred from |
| 1254 |
other values, for example the size of the frame carrying |
| 1255 |
the packet, and thus must not be included in the |
| 1256 |
compressed header. |
| 1257 |
|
| 1258 |
The classification implies how a compressed header is constructed. No |
| 1259 |
field that is NOCHANGE or INFERRED is present in a compressed header. |
| 1260 |
A compressor obtains the values of NOCHANGE fields from the context |
| 1261 |
identified by the compression identifier, and obtains the values of |
| 1262 |
INFERRED fields from the link-layer implementation, e.g., from the |
| 1263 |
size of the link-layer frame, or from other fields, e.g., by |
| 1264 |
recalculating the IPv4 header checksum. DELTA fields are encoded as |
| 1265 |
the difference to the value in the previous packet in the same packet |
| 1266 |
stream. The decompressor must update the context by adding the value |
| 1267 |
in the compressed header to the value in its context. The result is |
| 1268 |
the proper value of the field. RANDOM fields must be sent "as-is" in |
| 1269 |
the compressed header. RANDOM fields must occur in the same order in |
| 1270 |
the compressed header as they occur in the full header. |
| 1271 |
|
| 1272 |
Fields that may optionally be used to identify what packet stream a |
| 1273 |
packet belongs to according to <a href="#section-4.1">section 4.1</a> are marked with the word |
| 1274 |
DEF. To a compressor using the optional guidelines from <a href="#section-4.1">section 4.1</a>, |
| 1275 |
any difference in corresponding DEF fields between two packets |
| 1276 |
implies that they belong to different packet streams. Moreover, if a |
| 1277 |
DEF field is present in one packet but not in another, the packets |
| 1278 |
belong to different packet streams. |
| 1279 |
|
| 1280 |
|
| 1281 |
|
| 1282 |
|
| 1283 |
|
| 1284 |
<span class="grey">Degermark, et. al. Standards Track [Page 23]</span></pre> |
| 1285 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-24" ></span> |
| 1286 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1287 |
|
| 1288 |
|
| 1289 |
<span class="h3"><a class="selflink" id="section-7.1" href="#section-7.1">7.1</a>. IPv6 Header [IPv6, <a href="#section-3">section 3</a>]</span> |
| 1290 |
|
| 1291 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1292 |
|Version| Traffic Class | Flow Label | |
| 1293 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1294 |
| Payload Length | Next Header | Hop Limit | |
| 1295 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1296 |
| | |
| 1297 |
+ + |
| 1298 |
| | |
| 1299 |
+ Source Address + |
| 1300 |
| | |
| 1301 |
+ + |
| 1302 |
| | |
| 1303 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1304 |
| | |
| 1305 |
+ + |
| 1306 |
| | |
| 1307 |
+ Destination Address + |
| 1308 |
| | |
| 1309 |
+ + |
| 1310 |
| | |
| 1311 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1312 |
|
| 1313 |
Version NOCHANGE (DEF) |
| 1314 |
Traffic Class NOCHANGE (might be DEF, see sect 4.1) |
| 1315 |
(see also sect 6 a) |
| 1316 |
Flow Label NOCHANGE (DEF) |
| 1317 |
Payload Length INFERRED |
| 1318 |
Next Header NOCHANGE |
| 1319 |
Hop Limit NOCHANGE (might be DEF, see sect 4.1) |
| 1320 |
Source Address NOCHANGE (DEF) |
| 1321 |
Destination Address NOCHANGE (DEF) |
| 1322 |
|
| 1323 |
The Payload Length field of encapsulated headers must correspond to |
| 1324 |
the length value of the encapsulating header. If not, the header |
| 1325 |
chain MUST NOT be compressed. |
| 1326 |
|
| 1327 |
NOTE: If this the IP header closest to a TCP header, bit 7 of the |
| 1328 |
Traffic Class field can be passed using the R-flag of the compressed |
| 1329 |
TCP header. See <a href="#section-6">section 6</a> a). |
| 1330 |
|
| 1331 |
This classification implies that the entire IPv6 base header will be |
| 1332 |
compressed away. |
| 1333 |
|
| 1334 |
|
| 1335 |
|
| 1336 |
|
| 1337 |
|
| 1338 |
|
| 1339 |
|
| 1340 |
<span class="grey">Degermark, et. al. Standards Track [Page 24]</span></pre> |
| 1341 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-25" ></span> |
| 1342 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1343 |
|
| 1344 |
|
| 1345 |
<span class="h3"><a class="selflink" id="section-7.2" href="#section-7.2">7.2</a>. IPv6 Extension Headers [IPv6, <a href="#section-4">section 4</a>]</span> |
| 1346 |
|
| 1347 |
What extension headers are present and the relative order of them is |
| 1348 |
not expected to change in a packet stream. Whenever there is a |
| 1349 |
change, a full packet header must be sent. All Next Header fields in |
| 1350 |
IPv6 base header and IPv6 extension headers are NOCHANGE. |
| 1351 |
|
| 1352 |
<span class="h3"><a class="selflink" id="section-7.3" href="#section-7.3">7.3</a>. Options [IPv6, <a href="#section-4.2">section 4.2</a>]</span> |
| 1353 |
|
| 1354 |
The contents of Hop-by-hop Options and Destination Options extension |
| 1355 |
headers are encoded with TLV "options" (see [<a href="#ref-IPv6" title=""Internet Protocol, Version 6 (IPv6) Specification"">IPv6</a>]): |
| 1356 |
|
| 1357 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- - - - - - - - - |
| 1358 |
| Option Type | Opt Data Len | Option Data |
| 1359 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- - - - - - - - - |
| 1360 |
|
| 1361 |
Option Type and Opt Data Len fields are assumed to be fixed for a |
| 1362 |
given packet stream, so they are classified as NOCHANGE. The Option |
| 1363 |
data is RANDOM unless specified otherwise below. |
| 1364 |
|
| 1365 |
Padding |
| 1366 |
|
| 1367 |
Pad1 option |
| 1368 |
|
| 1369 |
+-+-+-+-+-+-+-+-+ |
| 1370 |
| 0 | |
| 1371 |
+-+-+-+-+-+-+-+-+ |
| 1372 |
|
| 1373 |
Entire option is NOCHANGE. |
| 1374 |
|
| 1375 |
PadN option |
| 1376 |
|
| 1377 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- - - - - - - - - |
| 1378 |
| 1 | Opt Data Len | Option Data |
| 1379 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- - - - - - - - - |
| 1380 |
|
| 1381 |
All fields are NOCHANGE. |
| 1382 |
|
| 1383 |
|
| 1384 |
|
| 1385 |
|
| 1386 |
|
| 1387 |
|
| 1388 |
|
| 1389 |
|
| 1390 |
|
| 1391 |
|
| 1392 |
|
| 1393 |
|
| 1394 |
|
| 1395 |
|
| 1396 |
<span class="grey">Degermark, et. al. Standards Track [Page 25]</span></pre> |
| 1397 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-26" ></span> |
| 1398 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1399 |
|
| 1400 |
|
| 1401 |
<span class="h3"><a class="selflink" id="section-7.4" href="#section-7.4">7.4</a>. Hop-by-Hop Options Header [IPv6, <a href="#section-4.3">section 4.3</a>]</span> |
| 1402 |
|
| 1403 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1404 |
| Next Header | Hdr Ext Len | | |
| 1405 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |
| 1406 |
| | |
| 1407 |
. . |
| 1408 |
. Options . |
| 1409 |
. . |
| 1410 |
| | |
| 1411 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1412 |
|
| 1413 |
Next Header NOCHANGE |
| 1414 |
Hdr Ext Len NOCHANGE |
| 1415 |
|
| 1416 |
Options TLV coded values and padding. |
| 1417 |
Classified according to 7.3 above, unless |
| 1418 |
being a Jumbo Payload option (see below). |
| 1419 |
|
| 1420 |
Jumbo Payload option |
| 1421 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1422 |
| 194 |Opt Data Len=4 | |
| 1423 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1424 |
| Jumbo Payload Length | |
| 1425 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1426 |
|
| 1427 |
First two fields are NOCHANGE and Jumbo Payload Length INFERRED. |
| 1428 |
(frame length must be supplied by link layer implementation). |
| 1429 |
|
| 1430 |
|
| 1431 |
NOTE: It is silly to compress the headers of a packet carrying a |
| 1432 |
Jumbo Payload Option since the relative header overhead is |
| 1433 |
negligible. Moreover, it is usually a bad idea to send such |
| 1434 |
large packets over low- and medium-speed links. |
| 1435 |
|
| 1436 |
<span class="h3"><a class="selflink" id="section-7.5" href="#section-7.5">7.5</a>. Routing Header [IPv6, <a href="#section-4.4">section 4.4</a>]</span> |
| 1437 |
|
| 1438 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1439 |
| Next Header | Hdr Ext Len | Routing Type | Segments Left | |
| 1440 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1441 |
| | |
| 1442 |
. . |
| 1443 |
. type-specific data . |
| 1444 |
. . |
| 1445 |
| | |
| 1446 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1447 |
|
| 1448 |
All fields of the Routing Header are NOCHANGE. |
| 1449 |
|
| 1450 |
|
| 1451 |
|
| 1452 |
<span class="grey">Degermark, et. al. Standards Track [Page 26]</span></pre> |
| 1453 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-27" ></span> |
| 1454 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1455 |
|
| 1456 |
|
| 1457 |
If the Routing Type is not recognized, it is impossible to determine |
| 1458 |
the final Destination Address unless the Segments Left field has the |
| 1459 |
value zero, in which case the Destination Address is the final |
| 1460 |
Destination Address in the basic IPv6 header. |
| 1461 |
|
| 1462 |
In the Type 0 Routing Header, the last address is DEF if (Segments |
| 1463 |
Left > 0). |
| 1464 |
|
| 1465 |
Routing Headers are compressed away completely. This is a big win as |
| 1466 |
the maximum size of the Routing Header is 392 octets. Moreover, Type |
| 1467 |
0 Routing Headers with one address, size 24 octets, are used by |
| 1468 |
Mobile IP. |
| 1469 |
|
| 1470 |
<span class="h3"><a class="selflink" id="section-7.6" href="#section-7.6">7.6</a>. Fragment Header [IPv6, <a href="#section-4.5">section 4.5</a>]</span> |
| 1471 |
|
| 1472 |
The first fragment of a packet has Fragment Offset = 0 and the chain |
| 1473 |
of subheaders extends beyond its Fragment Header. If a fragment is |
| 1474 |
not the first (Fragment Offset not 0), there are no subsequent |
| 1475 |
subheaders (unless the chain of subheaders in the first fragment |
| 1476 |
didn't fit entirely in the first fragment). |
| 1477 |
|
| 1478 |
Since packets may be reordered before reaching the compression point, |
| 1479 |
and some fragments may follow other routes through the network, a |
| 1480 |
compressor cannot rely on seeing the first fragment before other |
| 1481 |
fragments. This implies that information in subheaders following the |
| 1482 |
Fragment Header of the first fragment cannot be examined to determine |
| 1483 |
the proper packet stream for other fragments. |
| 1484 |
|
| 1485 |
It is possible to design compression schemes that can compress |
| 1486 |
subheaders after the Fragment Header, at least in the first fragment, |
| 1487 |
but to avoid complicating the rules for sending full headers and the |
| 1488 |
rules for compression and decompression, the chain of subheaders that |
| 1489 |
follow a Fragment Header MUST NOT be compressed. |
| 1490 |
|
| 1491 |
The fields of the Fragment Header are classified as follows. |
| 1492 |
|
| 1493 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1494 |
| Next Header | Reserved | Fragment Offset |Res|M| |
| 1495 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1496 |
| Identification | |
| 1497 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1498 |
|
| 1499 |
Next Header NOCHANGE |
| 1500 |
Reserved NOCHANGE |
| 1501 |
Res RANDOM |
| 1502 |
M flag RANDOM |
| 1503 |
Fragment Offset RANDOM |
| 1504 |
Identification RANDOM |
| 1505 |
|
| 1506 |
|
| 1507 |
|
| 1508 |
<span class="grey">Degermark, et. al. Standards Track [Page 27]</span></pre> |
| 1509 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-28" ></span> |
| 1510 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1511 |
|
| 1512 |
|
| 1513 |
This classification implies that a Fragment Header is compressed down |
| 1514 |
to 6 octets. The minimum IPv6 MTU is 1280 octets so most fragments |
| 1515 |
will be at least 1280 octets. Since the 6 octet overhead of the |
| 1516 |
compressed fragment header is amortized over a fairly large packet, |
| 1517 |
the additional complexity of more sophisticated compression schemes |
| 1518 |
is not justifiable. |
| 1519 |
|
| 1520 |
NOTE: The Identification field is RANDOM instead of NOCHANGE |
| 1521 |
to avoid one compression slow-start per original packet. |
| 1522 |
|
| 1523 |
Grouping of fragments according to the optional guidelines in |
| 1524 |
section4.1: |
| 1525 |
|
| 1526 |
Fragments and unfragmented packets should not be grouped |
| 1527 |
together. |
| 1528 |
|
| 1529 |
Port numbers cannot be used to identify the packet stream because |
| 1530 |
port numbers are not present in every fragment. To adhere to the |
| 1531 |
uniqueness rules for the Identification value, a fragmented |
| 1532 |
packet stream is identified by the combination of Source Address |
| 1533 |
and (final) Destination Address. |
| 1534 |
|
| 1535 |
NOTE: The Identification value is NOT used to identify the |
| 1536 |
packet stream. This avoids using a new CID for each packet and |
| 1537 |
saves the cost of the associated compression slow-start. We |
| 1538 |
expect that the unfragmentable part of the headers will not |
| 1539 |
change too frequently, if it does thrashing may occur. |
| 1540 |
|
| 1541 |
<span class="h3"><a class="selflink" id="section-7.7" href="#section-7.7">7.7</a>. Destination Options Header [IPv6, <a href="#section-4.6">section 4.6</a>]</span> |
| 1542 |
|
| 1543 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1544 |
| Next Header | Hdr Ext Len | | |
| 1545 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + |
| 1546 |
| | |
| 1547 |
. . |
| 1548 |
. Options . |
| 1549 |
. . |
| 1550 |
| | |
| 1551 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1552 |
|
| 1553 |
Next Header NOCHANGE |
| 1554 |
Hdr Ext Len NOCHANGE |
| 1555 |
|
| 1556 |
Options TLV coded values and padding. |
| 1557 |
Compressed according to 7.3 above. |
| 1558 |
|
| 1559 |
The only Destination Options defined in [<a href="#ref-IPv6" title=""Internet Protocol, Version 6 (IPv6) Specification"">IPv6</a>] are the padding |
| 1560 |
options. |
| 1561 |
|
| 1562 |
|
| 1563 |
|
| 1564 |
<span class="grey">Degermark, et. al. Standards Track [Page 28]</span></pre> |
| 1565 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-29" ></span> |
| 1566 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1567 |
|
| 1568 |
|
| 1569 |
<span class="h3"><a class="selflink" id="section-7.8" href="#section-7.8">7.8</a>. No Next Header [IPv6, <a href="#section-4.7">section 4.7</a>]</span> |
| 1570 |
|
| 1571 |
Covered by rules for IPv6 Header Extensions (7.2). |
| 1572 |
|
| 1573 |
<span class="h3"><a class="selflink" id="section-7.9" href="#section-7.9">7.9</a>. Authentication Header [RFC-2402, <a href="#section-3.2">section 3.2</a>]</span> |
| 1574 |
|
| 1575 |
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 |
| 1576 |
+---------------+---------------+---------------+---------------+ |
| 1577 |
| Next Header | Length | RESERVED | |
| 1578 |
+---------------+---------------+---------------+---------------+ |
| 1579 |
| Security Parameters Index (SPI) | |
| 1580 |
+---------------+---------------+---------------+---------------+ |
| 1581 |
| | |
| 1582 |
+ Authentication Data (variable number of 32-bit words) | |
| 1583 |
| | |
| 1584 |
+---------------+---------------+---------------+---------------+ |
| 1585 |
|
| 1586 |
Next Header NOCHANGE |
| 1587 |
Length NOCHANGE |
| 1588 |
Reserved NOCHANGE |
| 1589 |
SPI NOCHANGE (DEF) |
| 1590 |
Authentication Data RANDOM |
| 1591 |
|
| 1592 |
[<a id="ref-RFC-1828">RFC-1828</a>] specifies how to do authentication with keyed MD5, the |
| 1593 |
authentication method all IPv6 implementations must support. For |
| 1594 |
this method, the Authentication Data is 16 octets. |
| 1595 |
|
| 1596 |
<span class="h3"><a class="selflink" id="section-7.10" href="#section-7.10">7.10</a>. Encapsulating Security Payload Header [RFC-2406, <a href="#section-3.1">section 3.1</a>]</span> |
| 1597 |
|
| 1598 |
This header implies that the subsequent parts of the packet are |
| 1599 |
encrypted. Thus, no further header compression is possible on |
| 1600 |
subsequent headers as encryption is typically already performed when |
| 1601 |
the compressor sees the packet. |
| 1602 |
|
| 1603 |
However, when the ESP Header is used in tunnel mode an entire IP |
| 1604 |
packet is encrypted, and the headers of that packet MAY be compressed |
| 1605 |
before the packet is encrypted at the entry point of the tunnel. |
| 1606 |
This means that it must be possible to feed an IP packet and its |
| 1607 |
length to the decompressor, as if it came from the link-layer. The |
| 1608 |
mechanisms for dealing with reordering described in <a href="#section-11">section 11</a> MUST |
| 1609 |
also be used, as packets can be reordered in a tunnel. |
| 1610 |
|
| 1611 |
|
| 1612 |
|
| 1613 |
|
| 1614 |
|
| 1615 |
|
| 1616 |
|
| 1617 |
|
| 1618 |
|
| 1619 |
|
| 1620 |
<span class="grey">Degermark, et. al. Standards Track [Page 29]</span></pre> |
| 1621 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-30" ></span> |
| 1622 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1623 |
|
| 1624 |
|
| 1625 |
+---------------+---------------+---------------+---------------+ |
| 1626 |
| Security Association Identifier (SPI), 32 bits | |
| 1627 |
+===============+===============+===============+===============+ |
| 1628 |
| Opaque Transform Data, variable length | |
| 1629 |
+---------------+---------------+---------------+---------------+ |
| 1630 |
|
| 1631 |
SPI NOCHANGE (DEF) |
| 1632 |
Opaque Transform Data RANDOM |
| 1633 |
|
| 1634 |
Everything after the SPI is encrypted and is not compressed. |
| 1635 |
|
| 1636 |
<span class="h3"><a class="selflink" id="section-7.11" href="#section-7.11">7.11</a>. UDP Header</span> |
| 1637 |
|
| 1638 |
The UDP header is described in [<a href="./rfc768" title=""User Datagram Protocol"">RFC-768</a>]. |
| 1639 |
|
| 1640 |
|
| 1641 |
The Next Header field (IPv6) or Protocol field (IPv4) in the |
| 1642 |
preceding subheader is DEF. |
| 1643 |
|
| 1644 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1645 |
| Source Port | Destination Port | |
| 1646 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1647 |
| Length | Checksum | |
| 1648 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1649 |
|
| 1650 |
Source Port NOCHANGE (DEF) |
| 1651 |
Destination Port NOCHANGE (DEF) |
| 1652 |
Length INFERRED |
| 1653 |
Checksum RANDOM, unless it is zero, |
| 1654 |
in which case it is NOCHANGE. |
| 1655 |
|
| 1656 |
The Length field of the UDP header MUST match the Length field(s) of |
| 1657 |
preceding subheaders, i.e, there must not be any padding after the |
| 1658 |
UDP payload that is covered by the IP Length. |
| 1659 |
|
| 1660 |
The UDP header is typically compressed down to 2 octets, the UDP |
| 1661 |
checksum. When the UDP checksum is zero (which it cannot be with |
| 1662 |
IPv6), it is likely to be so for all packets in the flow and is |
| 1663 |
defined to be NOCHANGE. This saves 2 octets in the compressed header. |
| 1664 |
|
| 1665 |
<span class="h3"><a class="selflink" id="section-7.12" href="#section-7.12">7.12</a>. TCP Header</span> |
| 1666 |
|
| 1667 |
The TCP header is described in [<a href="./rfc793" title=""Transmission Control Protocol"">RFC-793</a>]. |
| 1668 |
|
| 1669 |
The Next Header field (IPv6) or Protocol field (IPv4) in the |
| 1670 |
preceding subheader is DEF. |
| 1671 |
|
| 1672 |
|
| 1673 |
|
| 1674 |
|
| 1675 |
|
| 1676 |
<span class="grey">Degermark, et. al. Standards Track [Page 30]</span></pre> |
| 1677 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-31" ></span> |
| 1678 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1679 |
|
| 1680 |
|
| 1681 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1682 |
| Source Port | Destination Port | |
| 1683 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1684 |
| Sequence Number | |
| 1685 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1686 |
| Acknowledgment Number | |
| 1687 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1688 |
| Offset| Reserved |U|A|P|R|S|F| Window | |
| 1689 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1690 |
| Checksum | Urgent Pointer | |
| 1691 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1692 |
| Options | Padding | |
| 1693 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1694 |
|
| 1695 |
U, A, P, R, S, and F stands for Urg, Ack, Psh, Rst, Syn, and Fin. |
| 1696 |
|
| 1697 |
There are two ways to compress the TCP header. |
| 1698 |
|
| 1699 |
<span class="h4"><a class="selflink" id="section-7.12.1" href="#section-7.12.1">7.12.1</a>. Compressed with differential encoding</span> |
| 1700 |
|
| 1701 |
Source Port NOCHANGE (DEF) |
| 1702 |
Destination Port NOCHANGE (DEF) |
| 1703 |
Sequence Number DELTA |
| 1704 |
Acknowledgment Number DELTA |
| 1705 |
Offset NOCHANGE |
| 1706 |
Reserved DELTA (if differs from context, |
| 1707 |
set R-flag in flag octet |
| 1708 |
and send absolute value |
| 1709 |
as described in 6 a.) |
| 1710 |
Urg,Psh RANDOM (placed in flag octet) |
| 1711 |
Ack INFERRED to be 1 |
| 1712 |
Rst,Syn,Fin INFERRED to be 0 |
| 1713 |
Window DELTA (if change in Window, |
| 1714 |
set W-flag in flag octet |
| 1715 |
and send difference) |
| 1716 |
Checksum RANDOM |
| 1717 |
Urgent Pointer DELTA (if Urg is set, send |
| 1718 |
absolute value) |
| 1719 |
Options, Padding DELTA (if change in Options, |
| 1720 |
set O-flag and send |
| 1721 |
whole Options, Padding) |
| 1722 |
|
| 1723 |
A packet with a TCP header compressed according to the above must be |
| 1724 |
indicated to be of type COMPRESSED_TCP. The compressed header is |
| 1725 |
described in <a href="#section-6">section 6</a>. |
| 1726 |
|
| 1727 |
|
| 1728 |
|
| 1729 |
|
| 1730 |
|
| 1731 |
|
| 1732 |
<span class="grey">Degermark, et. al. Standards Track [Page 31]</span></pre> |
| 1733 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-32" ></span> |
| 1734 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1735 |
|
| 1736 |
|
| 1737 |
This method is essentially the differential encoding techniques of |
| 1738 |
Jacobson, described in [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>], the differences being the placement |
| 1739 |
of the compressed TCP header fields (see <a href="#section-6">section 6</a>), the use of the |
| 1740 |
O-flag, the use of the R-flag, and elimination of the C-flag. The |
| 1741 |
O-flag allows compression of the TCP header when the Timestamp option |
| 1742 |
is used and the Options fields changes with each header. |
| 1743 |
|
| 1744 |
DELTA values (except for Reserved field and Options, Padding) MUST be |
| 1745 |
coded as in [<a href="./rfc1144" title=""Compressing TCP/IP Headers for Low- Speed Serial Links"">RFC-1144</a>]. A Reserved field value passed with the R-flag |
| 1746 |
MUST NOT update the context at compressor or decompressor. |
| 1747 |
|
| 1748 |
<span class="h4"><a class="selflink" id="section-7.12.2" href="#section-7.12.2">7.12.2</a>. Without differential encoding</span> |
| 1749 |
|
| 1750 |
Source Port NOCHANGE (DEF) |
| 1751 |
Destination Port NOCHANGE (DEF) |
| 1752 |
|
| 1753 |
(all the rest) RANDOM |
| 1754 |
|
| 1755 |
The Identification field in a preceding IPv4 header is RANDOM. |
| 1756 |
|
| 1757 |
A packet with a TCP header compressed according to the above must be |
| 1758 |
indicated to be of type COMPRESSED_TCP_NODELTA. It uses the same CID |
| 1759 |
space as COMPRESSED_TCP packets, and the header MUST be saved as |
| 1760 |
context. The compressed header is described in <a href="#section-6">section 6</a>. |
| 1761 |
|
| 1762 |
This packet type can be sent as the response to a header request |
| 1763 |
instead of sending a full header, can be used over links that reorder |
| 1764 |
packets, and can be sent instead of a full header when there are |
| 1765 |
changes that cannot be represented by a compressed header. A |
| 1766 |
sophisticated compressor can switch to sending only |
| 1767 |
COMPRESSED_TCP_NODELTA headers when the packet loss frequency is high. |
| 1768 |
|
| 1769 |
|
| 1770 |
|
| 1771 |
|
| 1772 |
|
| 1773 |
|
| 1774 |
|
| 1775 |
|
| 1776 |
|
| 1777 |
|
| 1778 |
|
| 1779 |
|
| 1780 |
|
| 1781 |
|
| 1782 |
|
| 1783 |
|
| 1784 |
|
| 1785 |
|
| 1786 |
|
| 1787 |
|
| 1788 |
<span class="grey">Degermark, et. al. Standards Track [Page 32]</span></pre> |
| 1789 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-33" ></span> |
| 1790 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1791 |
|
| 1792 |
|
| 1793 |
<span class="h3"><a class="selflink" id="section-7.13" href="#section-7.13">7.13</a>. IPv4 header [RFC-791, <a href="#section-3.1">section 3.1</a>]</span> |
| 1794 |
|
| 1795 |
0 1 2 3 |
| 1796 |
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 |
| 1797 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1798 |
|Version| IHL |Type of Service| Total Length | |
| 1799 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1800 |
| Identification |Flags| Fragment Offset | |
| 1801 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1802 |
| Time to Live | Protocol | Header Checksum | |
| 1803 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1804 |
| Source Address | |
| 1805 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1806 |
| Destination Address | |
| 1807 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1808 |
| Options | Padding | |
| 1809 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1810 |
|
| 1811 |
There are two ways to compress the IPv4 header |
| 1812 |
|
| 1813 |
a) If the IPv4 header is not for a fragment (MF flag is not set and |
| 1814 |
Fragment Offset is zero) and there are no options (IHL is 5), it |
| 1815 |
is classified as follows |
| 1816 |
|
| 1817 |
Version NOCHANGE (DEF) |
| 1818 |
IHL NOCHANGE (DEF, must be 5) |
| 1819 |
Type of Service NOCHANGE (might be DEF, see sect 4.1) |
| 1820 |
(see also 6 a) |
| 1821 |
Total Length INFERRED (from link-layer implementation |
| 1822 |
or encapsulating IP header) |
| 1823 |
|
| 1824 |
Identification DELTA/ (If the Protocol field has the |
| 1825 |
(value corresponding to TCP) |
| 1826 |
RANDOM (otherwise) |
| 1827 |
|
| 1828 |
Flags NOCHANGE (MF flag must not be set) |
| 1829 |
Fragment Offset NOCHANGE (must be zero) |
| 1830 |
Time to Live NOCHANGE (might be DEF, see sect 4.1) |
| 1831 |
Protocol NOCHANGE |
| 1832 |
Header Checksum INFERRED (calculated from other fields) |
| 1833 |
Source Address NOCHANGE (DEF) |
| 1834 |
Destination Address NOCHANGE (DEF) |
| 1835 |
Options, Padding (not present) |
| 1836 |
|
| 1837 |
|
| 1838 |
|
| 1839 |
|
| 1840 |
|
| 1841 |
|
| 1842 |
|
| 1843 |
|
| 1844 |
<span class="grey">Degermark, et. al. Standards Track [Page 33]</span></pre> |
| 1845 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-34" ></span> |
| 1846 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1847 |
|
| 1848 |
|
| 1849 |
Note: When a TCP header immediately follows, the IPv4 and TCP |
| 1850 |
header MUST be compressed as a unit as described in <a href="#section-6">section 6</a>. |
| 1851 |
Bits 6 and 7 of the Type of Service field (bits 14 and 15 of the |
| 1852 |
first word) can then be passed using the R-flag (see <a href="#section-6">section 6</a> |
| 1853 |
a). |
| 1854 |
|
| 1855 |
b) If the IPv4 header is for a fragment (MF bit set or Fragment |
| 1856 |
Offset nonzero), or there are options (IHL > 5), all fields are |
| 1857 |
RANDOM (i.e., if the header is compressed all fields are sent |
| 1858 |
as-is and not compressed). This classification allows compression |
| 1859 |
of the tunnel header, but not the fragment header, when fragments |
| 1860 |
are tunneled. If the IPv4 header is for a fragment it ends the |
| 1861 |
compressible chain of subheaders, i.e., it must be the last |
| 1862 |
subheader to be compressed. If the IPv4 header has options but |
| 1863 |
is not for a fragment it does not end the compressible chain of |
| 1864 |
subheaders, so subsequent subheaders can be compressed. |
| 1865 |
|
| 1866 |
A compressor that follows the optional guidelines of <a href="#section-4.1">section 4.1</a> will |
| 1867 |
in case a) use the Version, Source Address and Destination Address to |
| 1868 |
define the packet stream, together with the fact that there are no |
| 1869 |
IPv4 options and that this is not a fragment. |
| 1870 |
|
| 1871 |
Case b) can define two kinds of packet streams depending on whether |
| 1872 |
the IPv4 header is for a fragment or not. |
| 1873 |
|
| 1874 |
If the IPv4 header in case b) is for a fragment, a compressor |
| 1875 |
following the optional guidelines will use that fact together with |
| 1876 |
the Version, Source Address, and Destination Address to determine the |
| 1877 |
packet stream. |
| 1878 |
|
| 1879 |
If the IPv4 header in case b) is not for a fragment, it must have |
| 1880 |
options. A compressor following the optional guidelines will use that |
| 1881 |
fact, but not the size of the options, together with the Version, |
| 1882 |
Source Address, and Destination Address to determine the packet |
| 1883 |
stream. |
| 1884 |
|
| 1885 |
<span class="h3"><a class="selflink" id="section-7.14" href="#section-7.14">7.14</a>. Minimal Encapsulation header [RFC-2004, <a href="#section-3.1">section 3.1</a>]</span> |
| 1886 |
|
| 1887 |
0 1 2 3 |
| 1888 |
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 |
| 1889 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1890 |
| Protocol |S| reserved | Header Checksum | |
| 1891 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1892 |
| Original Destination Address | |
| 1893 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1894 |
: (if present) Original Source Address : |
| 1895 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
| 1896 |
|
| 1897 |
|
| 1898 |
|
| 1899 |
|
| 1900 |
<span class="grey">Degermark, et. al. Standards Track [Page 34]</span></pre> |
| 1901 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-35" ></span> |
| 1902 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1903 |
|
| 1904 |
|
| 1905 |
Protocol NOCHANGE |
| 1906 |
Original Source Address Present (S) NOCHANGE |
| 1907 |
reserved NOCHANGE |
| 1908 |
Header Checksum INFERRED (calculated from |
| 1909 |
other values) |
| 1910 |
Original Destination Address NOCHANGE |
| 1911 |
Original Source Address NOCHANGE (present only |
| 1912 |
if S=1) |
| 1913 |
|
| 1914 |
This header is likely to be used by Mobile IP. |
| 1915 |
|
| 1916 |
<span class="h2"><a class="selflink" id="section-8" href="#section-8">8</a>. Changing context identifiers</span> |
| 1917 |
|
| 1918 |
On a point-to-point link, the compressor has total knowledge of what |
| 1919 |
CIDs are in use at the decompressor and may change what CID a packet |
| 1920 |
stream uses or reuse CIDs at will. |
| 1921 |
|
| 1922 |
Each non-TCP CID is associated with a context with a generation |
| 1923 |
value. To avoid too rapid generation wrap-around and potential |
| 1924 |
incorrect decompression, an implementation MUST avoid wrap-around of |
| 1925 |
the generation value in less than MIN_WRAP seconds (see <a href="#section-14">section 14</a>). |
| 1926 |
|
| 1927 |
To aid in avoiding wrap-around, the generation value associated with |
| 1928 |
a CID MUST NOT be reset when changing to a new packet stream. |
| 1929 |
Instead, a compressor MUST increment the generation value by one when |
| 1930 |
using the CID for a new non-TCP packet stream. |
| 1931 |
|
| 1932 |
<span class="h2"><a class="selflink" id="section-9" href="#section-9">9</a>. Rules for dropping or temporarily storing packets</span> |
| 1933 |
|
| 1934 |
When a decompressor receives a packet with a compressed TCP header |
| 1935 |
with CID C, it MUST be discarded when the context for C has not been |
| 1936 |
initialized by a full header. |
| 1937 |
|
| 1938 |
When a decompressor receives a packet with a compressed non-TCP |
| 1939 |
header with CID C and generation G, the header must not be |
| 1940 |
decompressed using the current context when |
| 1941 |
|
| 1942 |
a) the decompressor has been disconnected from the compressor for |
| 1943 |
more than MIN_WRAP seconds, because the context might be |
| 1944 |
obsolete even if it has generation G. |
| 1945 |
|
| 1946 |
b) the context for C has a generation other than G. |
| 1947 |
|
| 1948 |
In case a) and b) the packet may either be |
| 1949 |
|
| 1950 |
i) discarded immediately, or else |
| 1951 |
|
| 1952 |
|
| 1953 |
|
| 1954 |
|
| 1955 |
|
| 1956 |
<span class="grey">Degermark, et. al. Standards Track [Page 35]</span></pre> |
| 1957 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-36" ></span> |
| 1958 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 1959 |
|
| 1960 |
|
| 1961 |
ii) stored temporarily until the context is updated by a packet |
| 1962 |
with a full non-TCP header with CID C and generation G, after |
| 1963 |
which the header can be decompressed. |
| 1964 |
|
| 1965 |
Packets stored in this manner MUST be discarded when |
| 1966 |
|
| 1967 |
*) receiving full or compressed non-TCP headers with CID C |
| 1968 |
and a generation other than G, |
| 1969 |
|
| 1970 |
*) the decompressor has not received packets with CID C in |
| 1971 |
the last MIN_WRAP seconds. |
| 1972 |
|
| 1973 |
When full headers are lost, a decompressor can receive compressed |
| 1974 |
non-TCP headers with a generation value other than the generation of |
| 1975 |
its context. Rule ii) allows the decompressor to store such headers |
| 1976 |
until they can be decompressed using the correct context. |
| 1977 |
|
| 1978 |
<span class="h2"><a class="selflink" id="section-10" href="#section-10">10</a>. Low-loss header compression for TCP</span> |
| 1979 |
|
| 1980 |
Since fewer bits are transmitted per packet with header compression, |
| 1981 |
the packet loss rate is lower with header compression than without, |
| 1982 |
for a fixed bit-error rate. This is beneficial for links with high |
| 1983 |
bit-error rates such as wireless links. |
| 1984 |
|
| 1985 |
However, since TCP headers are compressed using differential |
| 1986 |
encoding, a single lost TCP segment can ruin an entire TCP sending |
| 1987 |
window because the context is not incremented properly at the |
| 1988 |
decompressor. Subsequent headers will therefore be decompressed to |
| 1989 |
be different than before compression and discarded by the TCP |
| 1990 |
receiver because the TCP checksum fails. |
| 1991 |
|
| 1992 |
A TCP connection in the wide area where the last hop is over a |
| 1993 |
medium-speed lossy link, for example a wireless LAN, will then have |
| 1994 |
poor performance with traditional header compression because the |
| 1995 |
delay-bandwidth product is relatively large and the bit-error rate |
| 1996 |
relatively high. For a 2 Mbit/s wireless LAN and an end-to-end RTT of |
| 1997 |
200 ms, the delay-bandwidth product is 50 kbyte. That is equivalent |
| 1998 |
to about 97 512-octet segments with compressed headers. Each loss |
| 1999 |
can thus be multiplied by a factor of 100. |
| 2000 |
|
| 2001 |
This section describes two simple mechanisms for quick repair of the |
| 2002 |
context. With these mechanisms header compression will improve TCP |
| 2003 |
throughput over lossy links as well as links with low bit-error |
| 2004 |
rates. |
| 2005 |
|
| 2006 |
|
| 2007 |
|
| 2008 |
|
| 2009 |
|
| 2010 |
|
| 2011 |
|
| 2012 |
<span class="grey">Degermark, et. al. Standards Track [Page 36]</span></pre> |
| 2013 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-37" ></span> |
| 2014 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2015 |
|
| 2016 |
|
| 2017 |
<span class="h3"><a class="selflink" id="section-10.1" href="#section-10.1">10.1</a>. The "twice" algorithm</span> |
| 2018 |
|
| 2019 |
The decompressor may compute the TCP checksum to determine if its |
| 2020 |
context is not updated properly. If the checksum fails, the error is |
| 2021 |
assumed to be caused by a lost segment that did not update the |
| 2022 |
context properly. The delta of the current segment is then added to |
| 2023 |
the context again on the assumption that the lost segment contained |
| 2024 |
the same delta as the current. By decompressing and computing the TCP |
| 2025 |
checksum again, the decompressor checks if the repair succeeded or if |
| 2026 |
the delta should be applied once more. |
| 2027 |
|
| 2028 |
Analysis of traces of various TCP bulk transfers show that applying |
| 2029 |
the delta of the current segment one or two times will repair the |
| 2030 |
context for between 83 and 99 per cent of all single-segment losses |
| 2031 |
in the data stream. For the acknowledgment stream, the success rate |
| 2032 |
is smaller due to the delayed ack mechanism of TCP. The "twice" |
| 2033 |
mechanism repairs the context for 53 to 99 per cent of the losses in |
| 2034 |
the acknowledgment stream. A sophisticated implementation of this |
| 2035 |
idea would determine whether the TCP stream is an acknowledgment or |
| 2036 |
data stream and determine the segment size by observing the stream of |
| 2037 |
full and compressed headers. Trying deltas that are small multiples |
| 2038 |
of the segment size will result in even higher rates of successful |
| 2039 |
repairs for acknowledgment streams. |
| 2040 |
|
| 2041 |
<span class="h3"><a class="selflink" id="section-10.2" href="#section-10.2">10.2</a>. Header Requests</span> |
| 2042 |
|
| 2043 |
The relatively low success rate for the "twice" algorithm for TCP |
| 2044 |
acknowledgment streams calls for an additional mechanism for |
| 2045 |
repairing the context at the decompressor. When the decompressor |
| 2046 |
fails to repair the context after a loss, the decompressor may |
| 2047 |
optionally request a full header from the compressor. This is |
| 2048 |
possible on links where the decompressor can identify the compressor |
| 2049 |
and send packets to it. |
| 2050 |
|
| 2051 |
On such links, a decompressor may send a CONTEXT_STATE packet back to |
| 2052 |
the compressor to indicate that one or more contexts are invalid. A |
| 2053 |
decompressor SHOULD NOT transmit a CONTEXT_STATE packet every time a |
| 2054 |
compressed packet refers to an invalid context, but instead should |
| 2055 |
limit the rate of transmission of CONTEXT_STATE packets to avoid |
| 2056 |
flooding the reverse channel. A CONTEXT_STATE packet can indicate |
| 2057 |
that several contexts are out of date, this technique SHOULD be used |
| 2058 |
instead of sending several separate packets. The following diagram |
| 2059 |
shows the format of a CONTEXT_STATE packet. |
| 2060 |
|
| 2061 |
|
| 2062 |
|
| 2063 |
|
| 2064 |
|
| 2065 |
|
| 2066 |
|
| 2067 |
|
| 2068 |
<span class="grey">Degermark, et. al. Standards Track [Page 37]</span></pre> |
| 2069 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-38" ></span> |
| 2070 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2071 |
|
| 2072 |
|
| 2073 |
0 1 2 3 4 5 6 7 |
| 2074 |
+---+---+---+---+---+---+---+---+ |
| 2075 |
| TCP header request = 3 | |
| 2076 |
+---+---+---+---+---+---+---+---+ |
| 2077 |
| CID count | |
| 2078 |
+---+---+---+---+---+---+---+---+ |
| 2079 |
| CID | |
| 2080 |
+---+---+---+---+---+---+---+---+ |
| 2081 |
| CID | |
| 2082 |
+---+---+---+---+---+---+---+---+ |
| 2083 |
... |
| 2084 |
+---+---+---+---+---+---+---+---+ |
| 2085 |
| CID | |
| 2086 |
+---+---+---+---+---+---+---+---+ |
| 2087 |
|
| 2088 |
The first octet is a type code to allow the CONTEXT_STATE packet type |
| 2089 |
to be shared for other compression protocols that are (see [<a href="#ref-CRTP" title=""Compressing IP/UDP/RTP Headers for Low-Speed Serial Links"">CRTP</a>]) or |
| 2090 |
may be defined in parallel with this one. When used for TCP header |
| 2091 |
requests the type code has the value 3, and the remainder of the |
| 2092 |
packet is a sequence of CIDs preceded by a one-octet count of the |
| 2093 |
number of CIDs. |
| 2094 |
|
| 2095 |
On receipt of a CONTEXT_STATE packet, the compressor MUST mark the |
| 2096 |
CIDs invalid to ensure that the next packet emitted in those packet |
| 2097 |
streams are FULL_HEADER or COMPRESSED_TCP_NODELTA packets. |
| 2098 |
|
| 2099 |
Header requests are an optimization, so loss of a CONTEXT_STATE |
| 2100 |
packet does not affect the correct operation of TCP header |
| 2101 |
compression. When a CONTEXT_STATE packet is lost, eventually a new |
| 2102 |
one will be transmitted or TCP will timeout and retransmit. The big |
| 2103 |
advantage of using header requests is that TCP acknowledgment streams |
| 2104 |
can be repaired after a roundtrip-time over the lossy link. This |
| 2105 |
will typically avoid a TCP timeout and unnecessary retransmissions. |
| 2106 |
The lower packet loss rate due to smaller packets will then result in |
| 2107 |
higher throughput because the TCP window can grow larger between |
| 2108 |
losses. |
| 2109 |
|
| 2110 |
<span class="h2"><a class="selflink" id="section-11" href="#section-11">11</a>. Links that reorder packets</span> |
| 2111 |
|
| 2112 |
Some links reorder packets, for example multi-hop radio links that |
| 2113 |
use deflection routing to route around congested nodes. Packets |
| 2114 |
routed different ways can then arrive at the destination in a |
| 2115 |
different order than they were sent. |
| 2116 |
|
| 2117 |
|
| 2118 |
|
| 2119 |
|
| 2120 |
|
| 2121 |
|
| 2122 |
|
| 2123 |
|
| 2124 |
<span class="grey">Degermark, et. al. Standards Track [Page 38]</span></pre> |
| 2125 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-39" ></span> |
| 2126 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2127 |
|
| 2128 |
|
| 2129 |
<span class="h3"><a class="selflink" id="section-11.1" href="#section-11.1">11.1</a>. Reordering in non-TCP packet streams</span> |
| 2130 |
|
| 2131 |
Compressed non-TCP headers do not change the context, and neither do |
| 2132 |
full headers that refresh it. There can be problems only when a full |
| 2133 |
header that changes the context arrives out of order. There are two |
| 2134 |
cases: |
| 2135 |
|
| 2136 |
- A packet with a full header with generation G arrives *after* |
| 2137 |
a packet with a compressed header with generation G. This case |
| 2138 |
is covered by rule b) ii) in <a href="#section-9">section 9</a>. |
| 2139 |
|
| 2140 |
- A packet with a full header with generation G arrives *before* |
| 2141 |
a packet with a compressed header with generation G-1 (modulo |
| 2142 |
64). The decompressor MAY then keep both versions of the |
| 2143 |
context around for a while to be able to decompress subsequent |
| 2144 |
compressed headers with generation G-1 (modulo 64). The old |
| 2145 |
context MUST be discarded after MIN_WRAP seconds. |
| 2146 |
|
| 2147 |
<span class="h3"><a class="selflink" id="section-11.2" href="#section-11.2">11.2</a>. Reordering in TCP packet streams</span> |
| 2148 |
|
| 2149 |
A compressor may avoid sending COMPRESSED_TCP headers and only send |
| 2150 |
COMPRESSED_TCP_NODELTA headers when there is reordering over the |
| 2151 |
link. Compressed headers will typically be 17 octets with that |
| 2152 |
method, significantly larger than the usual 4-7 octets. |
| 2153 |
|
| 2154 |
To achieve better compression rates the following method, adding only |
| 2155 |
two octets to the compressed header for a total of 6-9 octets, may be |
| 2156 |
used. A packet sequence number, incremented by one for every packet |
| 2157 |
in the TCP stream, is then associated with each compressed and full |
| 2158 |
header. This allows the decompressor to place the packets in the |
| 2159 |
correct sequence and apply their deltas to the context in the correct |
| 2160 |
order. A simple sliding window scheme is used to place the packets |
| 2161 |
in the correct order. |
| 2162 |
|
| 2163 |
Two octets are needed for the packet sequence numbers. One octet |
| 2164 |
gives only 256 sequence numbers. In a sliding window scheme the |
| 2165 |
window should be no larger than half of the sequence number space, so |
| 2166 |
packets can not arrive more than 127 positions out-of-sequence. This |
| 2167 |
is equivalent to a delay of 260 ms on 2 Mbit/s links with 512 octet |
| 2168 |
segments. Delays of that order are not uncommon over wide-area |
| 2169 |
Internet connections. However, two octets giving 2^16 = 65536 values |
| 2170 |
should be sufficient. |
| 2171 |
|
| 2172 |
Full TCP/IP headers will only have space for one octet of sequence |
| 2173 |
number when there is no tunneling. It is not feasible to increase the |
| 2174 |
size of full headers since the packet size might be optimized for the |
| 2175 |
MTU of the link. Therefore only the least significant octet of the |
| 2176 |
packet sequence number can be placed in such full headers. We believe |
| 2177 |
|
| 2178 |
|
| 2179 |
|
| 2180 |
<span class="grey">Degermark, et. al. Standards Track [Page 39]</span></pre> |
| 2181 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-40" ></span> |
| 2182 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2183 |
|
| 2184 |
|
| 2185 |
that such full headers can be positioned correctly frequently enough |
| 2186 |
with only the least significant octet of the packet sequence number |
| 2187 |
available. |
| 2188 |
|
| 2189 |
The packet sequence number zero MUST be skipped over. Avoiding zero |
| 2190 |
takes care of a problem that can occur when the TCP window scale |
| 2191 |
option is used to enlarge the TCP window. When exactly 2^16 octets of |
| 2192 |
TCP data is lost, a compressed header will be decompressed |
| 2193 |
incorrectly without being detected by the TCP checksum. TCP segment |
| 2194 |
sizes are often a power of two. So by using a packet sequence number |
| 2195 |
space that is not a power of two either the TCP sequence number or |
| 2196 |
the packet sequence number will differ when 2^16 octets are lost. |
| 2197 |
Whenever a compressor sees the window scale option on a SYN segment, |
| 2198 |
it MUST use packet sequence numbers when subsequently compressing |
| 2199 |
that packet stream. |
| 2200 |
|
| 2201 |
In compressed TCP headers the two octet packet sequence number MUST |
| 2202 |
be placed immediately after the TCP Checksum. See <a href="#section-5.3">section 5.3</a> for |
| 2203 |
placement of packet sequence numbers in full headers. |
| 2204 |
|
| 2205 |
<span class="h2"><a class="selflink" id="section-12" href="#section-12">12</a>. Hooks for additional header compression</span> |
| 2206 |
|
| 2207 |
The following hook is supplied to allow additional header compression |
| 2208 |
schemes for headers on top of UDP. The initial chain of subheaders is |
| 2209 |
then compressed as described here, and the other header compression |
| 2210 |
scheme is applied to the header above the UDP header. An example of |
| 2211 |
such additional header compression is Compressed RTP by Casner and |
| 2212 |
Jacobson [<a href="#ref-CRTP" title=""Compressing IP/UDP/RTP Headers for Low-Speed Serial Links"">CRTP</a>]. To allow some error detection, such schemes |
| 2213 |
typically need a sequence number that may need to be passed in full |
| 2214 |
headers as well as compressed UDP headers. |
| 2215 |
|
| 2216 |
The D-bit and Data octet (see <a href="#section-6">section 6</a>) provides the necessary |
| 2217 |
mechanism. When a sequence number, say, needs to be passed in a |
| 2218 |
FULL_HEADER or COMPRESSED_NON_TCP header, the D-bit is set and the |
| 2219 |
sequence number is placed in the Data field. The decompressor must |
| 2220 |
then extract and make the Data field available to the additional |
| 2221 |
header compression scheme. |
| 2222 |
|
| 2223 |
Use of additional header compression schemes like CRTP must be |
| 2224 |
negotiated. The D-bit and Data octet mechanism must automatically be |
| 2225 |
enabled whenever use of additional header compression schemes has |
| 2226 |
been negotiated. |
| 2227 |
|
| 2228 |
|
| 2229 |
|
| 2230 |
|
| 2231 |
|
| 2232 |
|
| 2233 |
|
| 2234 |
|
| 2235 |
|
| 2236 |
<span class="grey">Degermark, et. al. Standards Track [Page 40]</span></pre> |
| 2237 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-41" ></span> |
| 2238 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2239 |
|
| 2240 |
|
| 2241 |
<span class="h2"><a class="selflink" id="section-13" href="#section-13">13</a>. Demultiplexing</span> |
| 2242 |
|
| 2243 |
For each link layer, there must be a document specifying how the |
| 2244 |
various packet types used by IP header compression is indicated. |
| 2245 |
Such a document exists for PPP [<a href="#ref-PPP-HC" title=""IP Header Compression for PPP"">PPP-HC</a>]. This section gives OPTIONAL |
| 2246 |
guidelines on how packet types may be indicated by a specific link- |
| 2247 |
layer. |
| 2248 |
|
| 2249 |
It is necessary to distinguish packets with regular IPv4 headers, |
| 2250 |
regular IPv6 headers, full IPv6 packets, full IPv4 packets, |
| 2251 |
compressed TCP packets, compressed non-TCP packets, and CONTEXT_STATE |
| 2252 |
packets. |
| 2253 |
|
| 2254 |
The decision to use a distinct ethertype (or equivalent) for IPv6 has |
| 2255 |
already been taken, which means that link-layers must be able to |
| 2256 |
indicate that a packet is an IPv6 packet. |
| 2257 |
|
| 2258 |
IP header compression requires that the link-layer implementation can |
| 2259 |
indicate four kinds of packets: COMPRESSED_TCP for format a) in |
| 2260 |
<a href="#section-6">section 6</a>, COMPRESSED_TCP_NODELTA for format b), COMPRESSED_NON_TCP |
| 2261 |
for formats c) and d), and CONTEXT_STATE as described in <a href="#section-11.2">section</a> |
| 2262 |
<a href="#section-11.2">11.2</a>. It is also desirable to indicate FULL_HEADERS at the link |
| 2263 |
layer. |
| 2264 |
|
| 2265 |
Full headers can be indicated by setting the first bit of the Version |
| 2266 |
field in a packet indicated to be an IPv6 packet. In addition, one |
| 2267 |
bit of the Version field is used to indicate if the first subheader |
| 2268 |
is an IPv6 or an IPv4 header, and one bit is used to indicate if this |
| 2269 |
full header carries a TCP CID or a non-TCP CID. The first four bits |
| 2270 |
are encoded as follows: |
| 2271 |
|
| 2272 |
Version Meaning |
| 2273 |
------- ------- |
| 2274 |
|
| 2275 |
0110 regular IPv6 header |
| 2276 |
|
| 2277 |
1T*0 T=1 indicates a TCP header, T=0 indicates a non-TCP header |
| 2278 |
1*V0 V=1 indicates a IPv6 header, V=0 indicates a IPv4 header |
| 2279 |
|
| 2280 |
If a link-layer cannot indicate the packet types for the compressed |
| 2281 |
headers or CONTEXT_STATE, packet types that cannot be indicated could |
| 2282 |
start with an octet indicating the packet type, followed by the |
| 2283 |
header. |
| 2284 |
|
| 2285 |
|
| 2286 |
|
| 2287 |
|
| 2288 |
|
| 2289 |
|
| 2290 |
|
| 2291 |
|
| 2292 |
<span class="grey">Degermark, et. al. Standards Track [Page 41]</span></pre> |
| 2293 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-42" ></span> |
| 2294 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2295 |
|
| 2296 |
|
| 2297 |
First octet Type of compressed header |
| 2298 |
----------- ------------------------- |
| 2299 |
|
| 2300 |
0 COMPRESSED_TCP |
| 2301 |
1 COMPRESSED_TCP_NODELTA |
| 2302 |
2 COMPRESSED_NON_TCP |
| 2303 |
3 CONTEXT_STATE |
| 2304 |
|
| 2305 |
The currently assigned CONTEXT_STATE type values are |
| 2306 |
|
| 2307 |
Value Type Reference |
| 2308 |
----- ----- ---------- |
| 2309 |
0 Reserved - |
| 2310 |
1 IP/UDP/RTP w. 8-bit CID [<a href="#ref-CRTP" title=""Compressing IP/UDP/RTP Headers for Low-Speed Serial Links"">CRTP</a>] |
| 2311 |
2 IP/UDP/RTP w. 16-bit CID [<a href="#ref-CRTP" title=""Compressing IP/UDP/RTP Headers for Low-Speed Serial Links"">CRTP</a>] |
| 2312 |
3 TCP header request <a href="#section-10.2">Section 10.2</a> |
| 2313 |
|
| 2314 |
<span class="h2"><a class="selflink" id="section-14" href="#section-14">14</a>. Configuration Parameters</span> |
| 2315 |
|
| 2316 |
Header compression parameters are negotiated in a way specific to the |
| 2317 |
link-layer implementation. Such procedures for link-layer xxx needs |
| 2318 |
to be specified in a document "IP header compression over xxx". Such |
| 2319 |
a document exists for PPP [<a href="#ref-PPP-HC" title=""IP Header Compression for PPP"">PPP-HC</a>]. |
| 2320 |
|
| 2321 |
The following parameter is fixed for all implementations of this |
| 2322 |
header compression scheme. |
| 2323 |
|
| 2324 |
MIN_WRAP - minimum time of generation value wrap around |
| 2325 |
|
| 2326 |
3 seconds. |
| 2327 |
|
| 2328 |
The following parameters can be negotiated between the compressor and |
| 2329 |
decompressor. If not negotiated their values must be as specified by |
| 2330 |
DEFAULT. |
| 2331 |
|
| 2332 |
F_MAX_PERIOD - Largest number of compressed non-TCP headers that |
| 2333 |
may be sent without sending a full header. |
| 2334 |
|
| 2335 |
DEFAULT is 256 |
| 2336 |
|
| 2337 |
F_MAX_PERIOD must be at least 1 and at most 65535. |
| 2338 |
|
| 2339 |
|
| 2340 |
F_MAX_TIME - Compressed headers may not be sent more than |
| 2341 |
F_MAX_TIME seconds after sending last full header. |
| 2342 |
|
| 2343 |
DEFAULT is 5 |
| 2344 |
|
| 2345 |
|
| 2346 |
|
| 2347 |
|
| 2348 |
<span class="grey">Degermark, et. al. Standards Track [Page 42]</span></pre> |
| 2349 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-43" ></span> |
| 2350 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2351 |
|
| 2352 |
|
| 2353 |
F_MAX_TIME must be at least 1 and at most 255. |
| 2354 |
|
| 2355 |
NOTE: F_MAX_PERIOD and F_MAX_TIME should be lower when it is |
| 2356 |
likely that a decompressor loses its state. |
| 2357 |
|
| 2358 |
|
| 2359 |
MAX_HEADER - The largest header size in octets that may |
| 2360 |
be compressed. |
| 2361 |
|
| 2362 |
DEFAULT is 168 octets, which covers |
| 2363 |
|
| 2364 |
- Two IPv6 base headers |
| 2365 |
- A Keyed MD5 Authentication Header |
| 2366 |
- A maximum-sized TCP header |
| 2367 |
|
| 2368 |
MAX_HEADER must be at least 60 octets and |
| 2369 |
at most 65535 octets. |
| 2370 |
|
| 2371 |
|
| 2372 |
TCP_SPACE - Maximum CID value for TCP. |
| 2373 |
|
| 2374 |
DEFAULT is 15 (which gives 16 CID values) |
| 2375 |
|
| 2376 |
TCP_SPACE must be at least 3 and at most 255. |
| 2377 |
|
| 2378 |
|
| 2379 |
NON_TCP_SPACE - Maximum CID value for non-TCP. |
| 2380 |
|
| 2381 |
DEFAULT is 15 (which gives 16 CID values) |
| 2382 |
|
| 2383 |
NON_TCP_SPACE must be at least 3 and at most 65535. |
| 2384 |
|
| 2385 |
|
| 2386 |
EXPECT_REORDERING - The mechanisms in <a href="#section-11">section 11</a> are used. |
| 2387 |
|
| 2388 |
DEFAULT no. |
| 2389 |
|
| 2390 |
<span class="h2"><a class="selflink" id="section-15" href="#section-15">15</a>. Implementation Status</span> |
| 2391 |
|
| 2392 |
A prototype using UDP as the link layer has been operational since |
| 2393 |
March 1996. A NetBSD implementation for PPP has been operational |
| 2394 |
since October 1996. |
| 2395 |
|
| 2396 |
|
| 2397 |
|
| 2398 |
|
| 2399 |
|
| 2400 |
|
| 2401 |
|
| 2402 |
|
| 2403 |
|
| 2404 |
<span class="grey">Degermark, et. al. Standards Track [Page 43]</span></pre> |
| 2405 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-44" ></span> |
| 2406 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2407 |
|
| 2408 |
|
| 2409 |
<span class="h2"><a class="selflink" id="section-16" href="#section-16">16</a>. Acknowledgments</span> |
| 2410 |
|
| 2411 |
This protocol uses many ideas originated by Van Jacobson in the |
| 2412 |
design of header compression for TCP/IP over slow-speed links [RFC- |
| 2413 |
1144]. It has benefited from discussions with Stephen Casner and |
| 2414 |
Carsten Bormann. |
| 2415 |
|
| 2416 |
We thank Craig Partridge for pointing out a problem that can occur |
| 2417 |
when the TCP window scale option is used. A solution to this problem |
| 2418 |
relying on the packet sequence numbers used for reordering is |
| 2419 |
described in <a href="#section-11.2">section 11.2</a>. |
| 2420 |
|
| 2421 |
<span class="h2"><a class="selflink" id="section-17" href="#section-17">17</a>. Security Considerations</span> |
| 2422 |
|
| 2423 |
The compression protocols in this document run on top of a link-layer |
| 2424 |
protocol. The compression protocols themselves introduce no new |
| 2425 |
additional vulnerabilities beyond those associated with the specific |
| 2426 |
link-layer technology being used. |
| 2427 |
|
| 2428 |
Denial-of-service attacks are possible if an intruder can introduce |
| 2429 |
(for example) bogus Full Header packets onto the link. However, an |
| 2430 |
intruder having the ability to inject arbitrary packets at the link- |
| 2431 |
layer in this manner raises additional security issues that dwarf |
| 2432 |
those related to the use of header compression. |
| 2433 |
|
| 2434 |
We advise implementors against identifying packet streams with the |
| 2435 |
aid of information that is encrypted, even if such information |
| 2436 |
happens to be available to the compressor. Doing so may expose |
| 2437 |
traffic patterns. |
| 2438 |
|
| 2439 |
|
| 2440 |
|
| 2441 |
|
| 2442 |
|
| 2443 |
|
| 2444 |
|
| 2445 |
|
| 2446 |
|
| 2447 |
|
| 2448 |
|
| 2449 |
|
| 2450 |
|
| 2451 |
|
| 2452 |
|
| 2453 |
|
| 2454 |
|
| 2455 |
|
| 2456 |
|
| 2457 |
|
| 2458 |
|
| 2459 |
|
| 2460 |
<span class="grey">Degermark, et. al. Standards Track [Page 44]</span></pre> |
| 2461 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-45" ></span> |
| 2462 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2463 |
|
| 2464 |
|
| 2465 |
<span class="h2"><a class="selflink" id="section-18" href="#section-18">18</a>. Authors' Addresses</span> |
| 2466 |
|
| 2467 |
Mikael Degermark |
| 2468 |
Department of Computer Science and Electrical Engineering |
| 2469 |
Lulea University of Technology |
| 2470 |
SE-971 87 Lulea, Sweden |
| 2471 |
|
| 2472 |
Phone: +46 920 91188 |
| 2473 |
Fax: +46 920 72831 |
| 2474 |
Mobile: +46 70 833 8933 |
| 2475 |
EMail: [email protected] |
| 2476 |
|
| 2477 |
|
| 2478 |
Bjorn Nordgren |
| 2479 |
CDT/Telia Research AB |
| 2480 |
Aurorum 6 |
| 2481 |
S-977 75 Lulea, Sweden |
| 2482 |
|
| 2483 |
Phone: +46 920 75400 |
| 2484 |
Fax: +46 920 75490 |
| 2485 |
EMail: [email protected], [email protected] |
| 2486 |
|
| 2487 |
|
| 2488 |
Stephen Pink |
| 2489 |
Department of Computer Science and Electrical Engineering |
| 2490 |
Lulea University of Technology |
| 2491 |
SE-971 87 Lulea, Sweden |
| 2492 |
|
| 2493 |
Phone: +46 920 752 29 |
| 2494 |
Fax: +46 920 728 31 |
| 2495 |
Mobile: +46 70 532 0007 |
| 2496 |
EMail: [email protected] |
| 2497 |
|
| 2498 |
|
| 2499 |
|
| 2500 |
|
| 2501 |
|
| 2502 |
|
| 2503 |
|
| 2504 |
|
| 2505 |
|
| 2506 |
|
| 2507 |
|
| 2508 |
|
| 2509 |
|
| 2510 |
|
| 2511 |
|
| 2512 |
|
| 2513 |
|
| 2514 |
|
| 2515 |
|
| 2516 |
<span class="grey">Degermark, et. al. Standards Track [Page 45]</span></pre> |
| 2517 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-46" ></span> |
| 2518 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2519 |
|
| 2520 |
|
| 2521 |
<span class="h2"><a class="selflink" id="section-19" href="#section-19">19</a>. References</span> |
| 2522 |
|
| 2523 |
[<a id="ref-RFC-768">RFC-768</a>] Postel, J., "User Datagram Protocol", STD 6, <a href="./rfc768">RFC 768</a>, |
| 2524 |
August 1980. |
| 2525 |
|
| 2526 |
[<a id="ref-RFC-791">RFC-791</a>] Postel, J., "Internet Protocol", STD 5, <a href="./rfc791">RFC 791</a>, |
| 2527 |
September 1981. |
| 2528 |
|
| 2529 |
[<a id="ref-RFC-793">RFC-793</a>] Postel, J., "Transmission Control Protocol", STD 7, |
| 2530 |
<a href="./rfc793">RFC 793</a>, September 1981. |
| 2531 |
|
| 2532 |
[<a id="ref-RFC-1144">RFC-1144</a>] Jacobson, V., "Compressing TCP/IP Headers for Low- |
| 2533 |
Speed Serial Links", <a href="./rfc1144">RFC 1144</a>, February 1990. |
| 2534 |
|
| 2535 |
[<a id="ref-RFC-1553">RFC-1553</a>] Mathur, A. and M. Lewis, "Compressing IPX Headers |
| 2536 |
Over WAN Media (CIPX)", <a href="./rfc1553">RFC 1553</a>, December 1993. |
| 2537 |
|
| 2538 |
[<a id="ref-RFC-1700">RFC-1700</a>] Reynolds, J. and J. Postel, "Assigned Numbers", STD |
| 2539 |
2, <a href="./rfc1700">RFC 1700</a>, October 1994. See also: |
| 2540 |
<a href="http://www.iana.org/numbers.html">http://www.iana.org/numbers.html</a> |
| 2541 |
|
| 2542 |
[<a id="ref-RFC-2402">RFC-2402</a>] Kent, S. and R. Atkinson, "IP Authentication Header", |
| 2543 |
<a href="./rfc2402">RFC 2402</a>, November 1998. |
| 2544 |
|
| 2545 |
[<a id="ref-RFC-2406">RFC-2406</a>] Kent, S. and R. Atkinson, "IP Encapsulating Security |
| 2546 |
Protocol (ESP)", <a href="./rfc2406">RFC 2406</a>, November 1998. |
| 2547 |
|
| 2548 |
[<a id="ref-RFC-1828">RFC-1828</a>] Metzger, W., "IP Authentication using Keyed MD5", <a href="./rfc1828">RFC</a> |
| 2549 |
<a href="./rfc1828">1828</a>, August 1995. |
| 2550 |
|
| 2551 |
[<a id="ref-IPv6">IPv6</a>] Deering, S. and R. Hinden, "Internet Protocol, |
| 2552 |
Version 6 (IPv6) Specification", <a href="./rfc2460">RFC 2460</a>, December |
| 2553 |
1998. |
| 2554 |
|
| 2555 |
[<a id="ref-ICMPv6">ICMPv6</a>] Conta, A. and S. Deering, "Internet Control Message |
| 2556 |
Protocol (ICMPv6) for the Internet Protocol Version 6 |
| 2557 |
(IPv6) Specification.", <a href="./rfc2463">RFC 2463</a>, December 1998. |
| 2558 |
|
| 2559 |
[<a id="ref-RFC-2004">RFC-2004</a>] Perkins, C., "Minimal Encapsulation within IP", <a href="./rfc2004">RFC</a> |
| 2560 |
<a href="./rfc2004">2004</a>, October 1996. |
| 2561 |
|
| 2562 |
[<a id="ref-CRTP">CRTP</a>] Casner, S. and V. Jacobson, "Compressing IP/UDP/RTP |
| 2563 |
Headers for Low-Speed Serial Links", <a href="./rfc2508">RFC 2508</a>, |
| 2564 |
February 1999. |
| 2565 |
|
| 2566 |
[<a id="ref-PPP-HC">PPP-HC</a>] Engan, M., Casner, S. and C. Bormann, "IP Header |
| 2567 |
Compression for PPP", <a href="./rfc2509">RFC 2509</a>, February 1999. |
| 2568 |
|
| 2569 |
|
| 2570 |
|
| 2571 |
|
| 2572 |
<span class="grey">Degermark, et. al. Standards Track [Page 46]</span></pre> |
| 2573 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-47" ></span> |
| 2574 |
<span class="grey"><a href="./rfc2507">RFC 2507</a> IP Header Compression February 1999</span> |
| 2575 |
|
| 2576 |
|
| 2577 |
<span class="h2"><a class="selflink" id="section-20" href="#section-20">20</a>. Full Copyright Statement</span> |
| 2578 |
|
| 2579 |
Copyright (C) The Internet Society (1999). All Rights Reserved. |
| 2580 |
|
| 2581 |
This document and translations of it may be copied and furnished to |
| 2582 |
others, and derivative works that comment on or otherwise explain it |
| 2583 |
or assist in its implementation may be prepared, copied, published |
| 2584 |
and distributed, in whole or in part, without restriction of any |
| 2585 |
kind, provided that the above copyright notice and this paragraph are |
| 2586 |
included on all such copies and derivative works. However, this |
| 2587 |
document itself may not be modified in any way, such as by removing |
| 2588 |
the copyright notice or references to the Internet Society or other |
| 2589 |
Internet organizations, except as needed for the purpose of |
| 2590 |
developing Internet standards in which case the procedures for |
| 2591 |
copyrights defined in the Internet Standards process must be |
| 2592 |
followed, or as required to translate it into languages other than |
| 2593 |
English. |
| 2594 |
|
| 2595 |
The limited permissions granted above are perpetual and will not be |
| 2596 |
revoked by the Internet Society or its successors or assigns. |
| 2597 |
|
| 2598 |
This document and the information contained herein is provided on an |
| 2599 |
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING |
| 2600 |
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING |
| 2601 |
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION |
| 2602 |
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF |
| 2603 |
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. .fi |
| 2604 |
|
| 2605 |
|
| 2606 |
|
| 2607 |
|
| 2608 |
|
| 2609 |
|
| 2610 |
|
| 2611 |
|
| 2612 |
|
| 2613 |
|
| 2614 |
|
| 2615 |
|
| 2616 |
|
| 2617 |
|
| 2618 |
|
| 2619 |
|
| 2620 |
|
| 2621 |
|
| 2622 |
|
| 2623 |
|
| 2624 |
|
| 2625 |
|
| 2626 |
|
| 2627 |
|
| 2628 |
Degermark, et. al. Standards Track [Page 47] |
| 2629 |
</pre> |