| 1 |
<pre>Network Working Group: M. McGovern |
| 2 |
Request for Comments: 1707 Sunspot Graphics |
| 3 |
Category: Informational R. Ullmann |
| 4 |
Lotus Development Corporation |
| 5 |
October 1994 |
| 6 |
|
| 7 |
|
| 8 |
<span class="h1">CATNIP: Common Architecture for the Internet</span> |
| 9 |
|
| 10 |
Status of this Memo |
| 11 |
|
| 12 |
This memo provides information for the Internet community. This memo |
| 13 |
does not specify an Internet standard of any kind. Distribution of |
| 14 |
this memo is unlimited. |
| 15 |
|
| 16 |
Abstract |
| 17 |
|
| 18 |
This document was submitted to the IETF IPng area in response to <a href="./rfc1550">RFC</a> |
| 19 |
<a href="./rfc1550">1550</a> Publication of this document does not imply acceptance by the |
| 20 |
IPng area of any ideas expressed within. Comments should be |
| 21 |
submitted to the [email protected] mailing list. |
| 22 |
|
| 23 |
Executive Summary |
| 24 |
|
| 25 |
This paper describes a common architecture for the network layer |
| 26 |
protocol. The Common Architecture for Next Generation Internet |
| 27 |
Protocol (CATNIP) provides a compressed form of the existing network |
| 28 |
layer protocols. Each compression is defined so that the resulting |
| 29 |
network protocol data units are identical in format. The fixed part |
| 30 |
of the compressed format is 16 bytes in length, and may often be the |
| 31 |
only part transmitted on the subnetwork. |
| 32 |
|
| 33 |
With some attention paid to details, it is possible for a transport |
| 34 |
layer protocol (such as TCP) to operate properly with one end system |
| 35 |
using one network layer (e.g. IP version 4) and the other using some |
| 36 |
other network protocol, such as CLNP. Using the CATNIP definitions, |
| 37 |
all the existing transport layer protocols used on connectionless |
| 38 |
network services will operate over any existing network layer |
| 39 |
protocol. |
| 40 |
|
| 41 |
The CATNIP uses cache handles to provide both rapid identification of |
| 42 |
the next hop in high performance routing as well as abbreviation of |
| 43 |
the network header by permitting the addresses to be omitted when a |
| 44 |
valid cache handle is available. The fixed part of the network layer |
| 45 |
header carries the cache handles. |
| 46 |
|
| 47 |
|
| 48 |
|
| 49 |
|
| 50 |
|
| 51 |
|
| 52 |
<span class="grey">McGovern & Ullmann [Page 1]</span></pre> |
| 53 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span> |
| 54 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 55 |
|
| 56 |
|
| 57 |
The cache handles are either provided by feedback from the downstream |
| 58 |
router in response to offered traffic, or explicitly provided as part |
| 59 |
of the establishment of a circuit or flow through the network. When |
| 60 |
used for flows, the handle is the locally significant flow |
| 61 |
identifier. |
| 62 |
|
| 63 |
When used for circuits, the handle is the layer 3 peer-to-peer |
| 64 |
logical channel identifier, and permits a full implementation of |
| 65 |
network-layer connection-oriented service if the routers along the |
| 66 |
path provide sufficient features. At the same time, the packet format |
| 67 |
of the connectionless service is retained, and hop by hop fully |
| 68 |
addressed datagrams can be used at the same time. Any intermediate |
| 69 |
model between the connection oriented and the connectionless service |
| 70 |
can thus be provided over cooperating routers. |
| 71 |
|
| 72 |
CATNIP Objectives |
| 73 |
|
| 74 |
The first objective of the CATNIP is a practical recognition of the |
| 75 |
existing state of internetworking, and an understanding that any |
| 76 |
approach must encompass the entire problem. While it is common in the |
| 77 |
IP Internet to dismiss the ISO with various amusing phrases, it is |
| 78 |
hardly realistic. As the Internet moves into the realm of providing |
| 79 |
real commercial infrastructure, for telephone, cable television, and |
| 80 |
the myriad other mundane uses, compliance with international |
| 81 |
standards is an imperative. |
| 82 |
|
| 83 |
The argument that the IETF need not (or should not) follow existing |
| 84 |
ISO standards will not hold. The ISO is the legal standards |
| 85 |
organization for the planet. Every other industry develops and |
| 86 |
follows ISO standards. There is (no longer) anything special about |
| 87 |
computer software or data networking. |
| 88 |
|
| 89 |
ISO convergence is both necessary and sufficient to gain |
| 90 |
international acceptance and deployment of IPng. Non-convergence will |
| 91 |
effectively preclude deployment. |
| 92 |
|
| 93 |
The CATNIP integrates CLNP, IP, and IPX. The CATNIP design provides |
| 94 |
for any of the transport layer protocols in use, for example TP4, |
| 95 |
CLTP, TCP, UDP, IPX and SPX to run over any of the network layer |
| 96 |
protocol formats: CLNP, IP (version 4), IPX, and the CATNIP. |
| 97 |
|
| 98 |
Incremental Infrastructure Deployment |
| 99 |
|
| 100 |
The best use of the CATNIP is to begin to build a common Internet |
| 101 |
infrastructure. The routers and other components of the common system |
| 102 |
are able to use a single consistent addressing method, and common |
| 103 |
terms of reference for other aspects of the system. |
| 104 |
|
| 105 |
|
| 106 |
|
| 107 |
|
| 108 |
<span class="grey">McGovern & Ullmann [Page 2]</span></pre> |
| 109 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span> |
| 110 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 111 |
|
| 112 |
|
| 113 |
The CATNIP is designed to be incrementally deployable in the strong |
| 114 |
sense: you can plop a CATNIP system down in place of any existing |
| 115 |
network component and continue to operate normally with no |
| 116 |
reconfiguration. (Note: not "just a little". None at all. The number |
| 117 |
of "little changes" suggested by some proposals, and the utterly |
| 118 |
enormous amount of documentation, training, and administrative effort |
| 119 |
then required, astounds the present authors.) The vendors do all of |
| 120 |
the work. |
| 121 |
|
| 122 |
There are also no external requirements; no "border routers", no |
| 123 |
requirement that administrators apply specific restrictions to their |
| 124 |
network designs, define special tables, or add things to the DNS. |
| 125 |
When the end users and administrators fully understand the combined |
| 126 |
system, they will want to operate differently, but in no case will |
| 127 |
they be forced. Not even in small ways. Networks and end user |
| 128 |
organizations operate under sufficient constraints on deployment of |
| 129 |
systems anyway; they do not need a new network architecture adding to |
| 130 |
the difficulty. |
| 131 |
|
| 132 |
Typically deployment will occur as part of normal upgrade revisions |
| 133 |
of software, and due to the "swamping" of the existing base as the |
| 134 |
network grows. (When the Internet grows by a factor of 5, at least |
| 135 |
80% will then be "new" systems.) The users of the network may then |
| 136 |
take advantage of the new capabilities. Some of the performance |
| 137 |
improvements will be automatic, others may require some |
| 138 |
administrative understanding to get to the best performance level. |
| 139 |
|
| 140 |
The CATNIP definitions provide stateless translation of network |
| 141 |
datagrams to and from CATNIP and, by implication, directly between |
| 142 |
the other network layer protocols. A CATNIP-capable system |
| 143 |
implementing the full set of definitions can interoperate with any |
| 144 |
existing protocol. Various subsets of the full capability may be |
| 145 |
provided by some vendors. |
| 146 |
|
| 147 |
No Address Translation |
| 148 |
|
| 149 |
Note that there is no "address translation" in the CATNIP |
| 150 |
specification. (While it may seem odd to state a negative objective, |
| 151 |
this is worth saying as people seem to assume the opposite.) There |
| 152 |
are no "mapping tables", no magic ways of digging translations out of |
| 153 |
the DNS or X.500, no routers looking up translations or asking other |
| 154 |
systems for them. |
| 155 |
|
| 156 |
Addresses are modified with a simple algorithmic mapping, a mapping |
| 157 |
that is no more than using specific prefixes for IP and IPX |
| 158 |
addresses. Not a large set of prefixes; one prefix. The entire |
| 159 |
existing IP version 4 network is mapped with one prefix and the IPX |
| 160 |
global network with one other prefix. (The IP mapping does provide |
| 161 |
|
| 162 |
|
| 163 |
|
| 164 |
<span class="grey">McGovern & Ullmann [Page 3]</span></pre> |
| 165 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span> |
| 166 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 167 |
|
| 168 |
|
| 169 |
for future assignment of other IANA/IPv4 domains that are disjoint |
| 170 |
from the existing one.) |
| 171 |
|
| 172 |
This means that there is no immediate effect on addresses embedded in |
| 173 |
higher level protocols. |
| 174 |
|
| 175 |
Higher level protocols not using the full form (those native to IP |
| 176 |
and IPX) will eventually be extended to use the full addressing to |
| 177 |
extend their usability over all of the network layers. |
| 178 |
|
| 179 |
No Legacy Systems |
| 180 |
|
| 181 |
The CATNIP leaves no systems behind: with no reconfiguration, any |
| 182 |
system presently capable of IP, CLNP, or IPX retains at least the |
| 183 |
connectivity it has now. With some administrative changes (such as |
| 184 |
assigning IPX domain addresses to some CLNP hosts for example) on |
| 185 |
other systems, unmodified systems may gain significant connectivity. |
| 186 |
IPX systems with registered network numbers may gain the most. |
| 187 |
|
| 188 |
Limited Scope |
| 189 |
|
| 190 |
The CATNIP defines a common network layer packet format and basic |
| 191 |
architecture. It intentionally does not specify ES-IS methods, |
| 192 |
routing, naming systems, autoconfiguration and other subjects not |
| 193 |
part of the core Internet wide architecture. The related problems and |
| 194 |
their (many) solutions are not within the scope of the specification |
| 195 |
of the basic common network layer. |
| 196 |
|
| 197 |
Existing Addresses and Network Numbers |
| 198 |
|
| 199 |
The Internet's version 4 numbering system has proven to be very |
| 200 |
flexible, (mostly) expandable, and simple. In short: it works. |
| 201 |
However, there are two problems. Neither was considered serious when |
| 202 |
the CATNIP was first developed in 1988 and 1989, but both are now of |
| 203 |
major concern: |
| 204 |
|
| 205 |
|
| 206 |
o The division into network, and then subnet, is insufficient. |
| 207 |
Almost all sites need a network assignment large enough to |
| 208 |
subnet. At the top of the hierarchy, there is a need to assign |
| 209 |
administrative domains. |
| 210 |
|
| 211 |
o As bit-packing is done to accomplish the desired network |
| 212 |
structure, the 32-bit limit causes more and more aggravation. |
| 213 |
|
| 214 |
Another major addressing system used in open internetworking is the |
| 215 |
OSI method of specifying Network Service Access Points (NSAPs). The |
| 216 |
NSAP consists of an authority and format identifier, a number |
| 217 |
|
| 218 |
|
| 219 |
|
| 220 |
<span class="grey">McGovern & Ullmann [Page 4]</span></pre> |
| 221 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span> |
| 222 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 223 |
|
| 224 |
|
| 225 |
assigned to that authority, an address assigned by that authority, |
| 226 |
and a selector identifying the next layer (transport layer) protocol. |
| 227 |
This is actually a general multi-level hierarchy, often obscured by |
| 228 |
the details of specific profiles. (For example, CLNP doesn't specify |
| 229 |
20 octet NSAPs, it allows any length. But various GOSIPs profile the |
| 230 |
NSAP as 20 octets, and IS-IS makes specific assumptions about the |
| 231 |
last 1-8 octets. And so on.) |
| 232 |
|
| 233 |
The NSAP does not directly correspond to an IP address, as the |
| 234 |
selector in IP is separate from the address. The concept that does |
| 235 |
correspond is the NSAP less the selector, called the Network Entity |
| 236 |
Title or NET. (An unfortunate acronym, but one we will use to avoid |
| 237 |
repeating the full term.) The usual definition of NET is an NSAP with |
| 238 |
the selector set to 0; the NET used here omits the 0 selector. |
| 239 |
|
| 240 |
There is also a network numbering system used by IPX, a product of |
| 241 |
Novell, Inc. (referred to from here on as Novell) and other vendors |
| 242 |
making compatible software. While IPX is not yet well connected into |
| 243 |
a global network, it has a larger installed base than either of the |
| 244 |
other network layers. |
| 245 |
|
| 246 |
Network Layer Address |
| 247 |
|
| 248 |
The network layer address looks like: |
| 249 |
|
| 250 |
+----------+----------+---------------+---------------+ |
| 251 |
| length | AFI | IDI ... | DSP ... | |
| 252 |
+----------+----------+---------------+---------------+ |
| 253 |
|
| 254 |
The fields are named in the usual OSI terminology although that leads |
| 255 |
to an oversupply of acronyms. Here are more detailed descriptions of |
| 256 |
each field: |
| 257 |
|
| 258 |
length: the number of bytes (octets) in the remainder of the |
| 259 |
address. |
| 260 |
|
| 261 |
AFI: the Authority and Format Identifier. A single byte |
| 262 |
value, from a set of well-known values registered by |
| 263 |
ISO, that determines the semantics of the IDI field |
| 264 |
|
| 265 |
IDI: the Initial Domain Identifier, a number assigned by the |
| 266 |
authority named by the AFI, formatted according to the |
| 267 |
semantics implied by the AFI, that determines the |
| 268 |
authority for the remainder of the address. |
| 269 |
|
| 270 |
DSP: Domain Specific Part, an address assigned by the |
| 271 |
authority identified by the value of the IDI. |
| 272 |
|
| 273 |
|
| 274 |
|
| 275 |
|
| 276 |
<span class="grey">McGovern & Ullmann [Page 5]</span></pre> |
| 277 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span> |
| 278 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 279 |
|
| 280 |
|
| 281 |
Note that there are several levels of authority. ISO, for example, |
| 282 |
identifies (with the AFI) a set of numbering authorities (like X.121, |
| 283 |
the numbering plan for the PSPDN, or E.164, the numbering plan for |
| 284 |
the telephone system). Each authority numbers a set of organizations |
| 285 |
or individuals or other entities. (For example, E.164 assigns |
| 286 |
16172477959 to me as a telephone subscriber.) |
| 287 |
|
| 288 |
The entity then is the authority for the remainder of the address. I |
| 289 |
can do what I please with the addresses starting with (AFI=E.164) |
| 290 |
(IDI=16172477959). Note that this is a delegation of authority, and |
| 291 |
not an embedding of a data-link address (the telephone number) in a |
| 292 |
network layer address. The actual routing of the network layer |
| 293 |
address has nothing to do with the authority numbering. |
| 294 |
|
| 295 |
The domain-specific part is variable length, and can be allocated in |
| 296 |
whatever way the authority identified by the AFI+IDI desires. |
| 297 |
|
| 298 |
Network Layer Datagram |
| 299 |
|
| 300 |
The common architecture format for network layer datagrams is |
| 301 |
described below. The design is a balance between use on high |
| 302 |
performance networks and routers, and a desire to minimize the number |
| 303 |
of bits in the fixed header. Using the current state of processor |
| 304 |
technology as a reference, the fixed header is all loaded into CPU |
| 305 |
registers on the first memory cycle, and it all fits within the |
| 306 |
operation bandwidth. The header leaves the remaining data aligned on |
| 307 |
the header size (128 bits); with 64 bit addresses present and no |
| 308 |
options it leaves the transport header 256 bit aligned. |
| 309 |
|
| 310 |
On very slow and low performance networks, the fixed header is still |
| 311 |
fairly small, and could be further compressed by methods similar to |
| 312 |
those used with IP version 4 on links that consider every bit |
| 313 |
precious. In between, it fits nicely into ATM cells and radio |
| 314 |
packets, leaving sufficient space for the transport header and |
| 315 |
application data. |
| 316 |
|
| 317 |
|
| 318 |
|
| 319 |
|
| 320 |
|
| 321 |
|
| 322 |
|
| 323 |
|
| 324 |
|
| 325 |
|
| 326 |
|
| 327 |
|
| 328 |
|
| 329 |
|
| 330 |
|
| 331 |
|
| 332 |
<span class="grey">McGovern & Ullmann [Page 6]</span></pre> |
| 333 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span> |
| 334 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 335 |
|
| 336 |
|
| 337 |
+---------------+---------------+-+-+-+-+-+-+-+-+---------------+ |
| 338 |
| NLPID (70) | Header Size |D|S|R|M|E| MBZ | Time to Live | |
| 339 |
+---------------+---------------+-+-+-+-+-+-+-+-+---------------+ |
| 340 |
| Forward Cache Identifier | |
| 341 |
+---------------------------------------------------------------+ |
| 342 |
| Datagram Length | |
| 343 |
+---------------------------------------------------------------+ |
| 344 |
| Transport Protocol | Checksum | |
| 345 |
+---------------------------------------------------------------+ |
| 346 |
| Destination Address ... | |
| 347 |
+---------------------------------------------------------------+ |
| 348 |
| Source Address ... | |
| 349 |
+---------------------------------------------------------------+ |
| 350 |
| Options ... | |
| 351 |
+---------------------------------------------------------------+ |
| 352 |
|
| 353 |
|
| 354 |
NLPID: The first byte (the network layer protocol identifier in OSI) |
| 355 |
is an 8 bit constant 70 (hex). This corresponds to Internet |
| 356 |
Version 7. |
| 357 |
|
| 358 |
Header Length: The header length is a 8-bit count of the number of |
| 359 |
32-bit words in the header. This allows the header to |
| 360 |
be up to 1020 bytes in length. |
| 361 |
|
| 362 |
Flags: This byte is a small set of flags determining the datagram |
| 363 |
header format and the processing semantics. The last three bits |
| 364 |
are reserved, and must be set to zero. (Note that the |
| 365 |
corresponding bits in CLNP version 1 are 001, since this byte |
| 366 |
is the version field. This may be useful.) |
| 367 |
|
| 368 |
Destination Address Omitted: When the destination address omitted |
| 369 |
(DAO) flag is zero, the destination address is present as shown |
| 370 |
in the datagram format diagram. When a datagram is sent with |
| 371 |
an FCI that identifies the destination and the DAO flag is |
| 372 |
set, the address does not appear in the datagram. |
| 373 |
|
| 374 |
Source Address Omitted: The source address omitted (SAO) flag is zero |
| 375 |
when the source address is present in the datagram. When |
| 376 |
datagram is sent with an FCI that identifies the source and the |
| 377 |
SAO flag is set, the source address is omitted from the |
| 378 |
datagram. |
| 379 |
|
| 380 |
Report Fragmentation Done: When this bit (RFD) is set, an intermediate |
| 381 |
router that fragments the datagram (because it is larger than |
| 382 |
the next subnetwork MTU) should report the event with an ICMP |
| 383 |
Datagram Too Big message. (Unlike IP version 4, which uses DF |
| 384 |
for MTU discovery, the RFD flag allows the fragmented datagram |
| 385 |
|
| 386 |
|
| 387 |
|
| 388 |
<span class="grey">McGovern & Ullmann [Page 7]</span></pre> |
| 389 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span> |
| 390 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 391 |
|
| 392 |
|
| 393 |
to be delivered.) |
| 394 |
|
| 395 |
Mandatory Router Option: The mandatory router option (MRO) flag |
| 396 |
indicates that routers forwarding the datagram must look at the |
| 397 |
network header options. If not set, an intermediate router |
| 398 |
should not look at the header options. (But it may anyway; |
| 399 |
this is a necessary consequence of transparent network layer |
| 400 |
translation, which may occur anywhere.) |
| 401 |
|
| 402 |
The destination host, or an intermediate router doing |
| 403 |
translation, must look at the header options regardless of |
| 404 |
the setting of the MRO flag. |
| 405 |
|
| 406 |
A router doing fragmentation will normally only use the F |
| 407 |
flag in options to determine whether options should be copied |
| 408 |
within the fragmentation code path. (It might also recognize |
| 409 |
and elide null options.) If the MRO flag is not set, the router |
| 410 |
may not act on an option even though it copies it properly |
| 411 |
during fragmentation. |
| 412 |
|
| 413 |
If there are no options present, MRO should always be zero, so |
| 414 |
that routers can follow the no-option profile path in their |
| 415 |
implementation. (Remember that the presence of options cannot |
| 416 |
be divided from the header length, since the addresses are |
| 417 |
variable length.) |
| 418 |
|
| 419 |
Error Report Suppression: The ERS flag is set to suppress the sending |
| 420 |
of error reports by any system (whether host or router) |
| 421 |
receiving or forwarding the datagram. The system may log the |
| 422 |
error, increment network management counters, and take any |
| 423 |
similar action, but ICMP error messages or CNLP error reports |
| 424 |
must not be sent. |
| 425 |
|
| 426 |
The ERS flag is normally set on ICMP messages and other network |
| 427 |
layer error reports. It does not suppress the normal response |
| 428 |
to ICMP queries or similar network layer queries (CNLP echo |
| 429 |
request). |
| 430 |
|
| 431 |
If both the RFD and ERS flags are set, the fragmentation report |
| 432 |
is sent. (This definition allows a larger range of |
| 433 |
possibilities than simply over-riding the RFD flag would; a |
| 434 |
sender not desiring this behavior can see to it that RFD is |
| 435 |
clear.) |
| 436 |
|
| 437 |
Time To Live: The time to live is a 8-bit count, nominally in seconds. |
| 438 |
Each hop is required to decrement TTL by at least one. A hop |
| 439 |
that holds a datagram for an unusual amount of time (more than |
| 440 |
2 seconds, a typical example being a wait for a subnetwork |
| 441 |
|
| 442 |
|
| 443 |
|
| 444 |
<span class="grey">McGovern & Ullmann [Page 8]</span></pre> |
| 445 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span> |
| 446 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 447 |
|
| 448 |
|
| 449 |
connection establishment) should subtract the entire waiting |
| 450 |
time in seconds (rounded upward) from the TTL. |
| 451 |
|
| 452 |
Forward Cache Identifier: Each datagram carries a 32 bit field, called |
| 453 |
"forward cache identifier", that is updated (if the information |
| 454 |
is available) at each hop. This field's value is derived from |
| 455 |
ICMP messages sent back by the next hop router, a routing |
| 456 |
protocol (e.g., RAP), or some other method. The FCI is used to |
| 457 |
expedite routing decisions by preserving knowledge where |
| 458 |
possible between consecutive routers. It can also be used to |
| 459 |
make datagrams stay within reserved flows, circuits, and mobile |
| 460 |
host tunnels. If an FCI is not available, this field must be |
| 461 |
zero, the SAO and DAO flags must be clear, and both destination |
| 462 |
and source addresses must appear in the datagram. |
| 463 |
|
| 464 |
Datagram Length: The 32-bit length of the entire datagram in octets. |
| 465 |
A datagram can therefore be up to 4294967295 bytes in overall |
| 466 |
length. Particular networks normally impose lower limits. |
| 467 |
|
| 468 |
Transport Protocol: The transport layer protocol. For example, TCP is |
| 469 |
6. |
| 470 |
|
| 471 |
Checksum: The checksum is a 16-bit checksum of the entire header, |
| 472 |
using the familiar algorithm used in IP version 4. |
| 473 |
|
| 474 |
Destination: The destination address, a count byte followed by the |
| 475 |
destination NSAP with the zero selector omitted. This field is |
| 476 |
present only if the DAO flag is zero. If the count field is not |
| 477 |
3 modulo 4 (the destination is not an integral multiple of |
| 478 |
32-bit words) zero bytes are added to pad to the next multiple |
| 479 |
of 32 bits. These pad bytes are not required to be ignored: |
| 480 |
routers may rely on them being zero. |
| 481 |
|
| 482 |
Source: The source address, in the same format as the destination. |
| 483 |
Present only if the SAO flag is zero. The source is padded in |
| 484 |
the same way as destination to arrive at a 32-bit boundary. |
| 485 |
|
| 486 |
Options: Options may follow. They are variable length, and always |
| 487 |
32-bit aligned. If the MRO flag in the header is not set, |
| 488 |
routers will usually not look at or take action on any option, |
| 489 |
regardless of the setting of the class field. |
| 490 |
|
| 491 |
Multicasting |
| 492 |
|
| 493 |
The multicast-enable option permits multicast forwarding of the |
| 494 |
CATNIP datagram on subnetworks that directly support media layer |
| 495 |
multicasting. This is a vanishing species, even in 10 Mbps Ethernet, |
| 496 |
given the increasing prevalence of switching hubs. It also (perhaps |
| 497 |
|
| 498 |
|
| 499 |
|
| 500 |
<span class="grey">McGovern & Ullmann [Page 9]</span></pre> |
| 501 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span> |
| 502 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 503 |
|
| 504 |
|
| 505 |
more usefully) permits a router to forward the datagram on multiple |
| 506 |
paths when a multicast routing algorithm has established such paths. |
| 507 |
There is no option data. |
| 508 |
|
| 509 |
Note that there is no special address space for multicasting in the |
| 510 |
CATNIP. Multicast destination addresses can be allocated anywhere by |
| 511 |
any administration or authority. This supports a number of differing |
| 512 |
models of addressing. It does require that the transport layer |
| 513 |
protocol know that the destination is multicast; this is desirable in |
| 514 |
any case. (For example, the transport will probably want to set the |
| 515 |
ERS flag.) |
| 516 |
|
| 517 |
On an IEEE 802.x (ISO 8802.x) type media, the last 23 bits of the |
| 518 |
address (not including the 0 selector) are used in combination with |
| 519 |
the multicast group address assigned to the Internet to form the |
| 520 |
media address when forwarding a datagram with the multicast enable |
| 521 |
option from a router to an attached network provided that the |
| 522 |
datagram was not received on that network with either multicast or |
| 523 |
broadcast media addressing. A host may send a multicast datagram |
| 524 |
either to the media multicast address (the IP catenet model,) or |
| 525 |
media unicast to a router which is expected to repeat it to the |
| 526 |
multicast address within the entire level I area or to repeat copies |
| 527 |
to the appropriate end systems within the area on non-broadcast media |
| 528 |
(the more general CLNP model.) |
| 529 |
|
| 530 |
Network Layer Translation |
| 531 |
|
| 532 |
The objective of translation is to be able to upgrade systems, both |
| 533 |
hosts and routers, in whatever order desired by their owners. |
| 534 |
Organizations must be able to upgrade any given system without |
| 535 |
reconfiguration or modification of any other, and existing hosts must |
| 536 |
be able to interoperate essentially forever. (Non-CATNIP routers will |
| 537 |
probably be effectively eliminated at some point, except where they |
| 538 |
exist in their own remote or isolated corners.) |
| 539 |
|
| 540 |
Each CATNIP system, whether host or router, must be able to recognize |
| 541 |
adjacent systems in the topology that are (only) IP version 4, CLNP, |
| 542 |
or IPX and call the appropriate translation routine just before |
| 543 |
sending the datagram. |
| 544 |
|
| 545 |
OSI CNLP |
| 546 |
|
| 547 |
The translation between CLNP and the CATNIP compressed form of the |
| 548 |
datagrams is the simplest case for CATNIP, since the addresses are |
| 549 |
the same and need not be extended. The resulting CATNIP datagrams may |
| 550 |
omit the source and destination addresses as explained previously, |
| 551 |
and may be mixed with uncompressed datagrams on the same subnetwork |
| 552 |
link. Alternatively, a subnetwork may operate entirely in the CATNIP, |
| 553 |
|
| 554 |
|
| 555 |
|
| 556 |
<span class="grey">McGovern & Ullmann [Page 10]</span></pre> |
| 557 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span> |
| 558 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 559 |
|
| 560 |
|
| 561 |
converting all transit traffic to CATNIP datagrams, even if FCIs that |
| 562 |
would make the compression effective are not available. |
| 563 |
|
| 564 |
Similarly, all network datagram formats with CATNIP mappings may be |
| 565 |
compressed into the common form, providing a uniform transit network |
| 566 |
service, with common routing protocols (such as IS-IS). |
| 567 |
|
| 568 |
Internet Protocol |
| 569 |
|
| 570 |
All existing version 4 numbers are defined as belonging to the |
| 571 |
Internet by using a new AFI, to be assigned to IANA by the ISO. This |
| 572 |
document uses 192 at present for clarity in examples; it is to be |
| 573 |
replaced with the assigned AFI. The AFI specifies that the IDI is two |
| 574 |
bytes long, containing an administrative domain number. |
| 575 |
|
| 576 |
The AD (Administrative Domain), identifies an administration which |
| 577 |
may be an international authority (such as the existing InterNIC), a |
| 578 |
national administration, or a large multi-organization (e.g., a |
| 579 |
government). The idea is that there should not be more than a few |
| 580 |
hundred of these at first, and eventually thousands or tens of |
| 581 |
thousands at most. |
| 582 |
|
| 583 |
AD numbers are assigned by IANA. Initially, the only assignment is |
| 584 |
the number 0.0, assigned to the InterNIC, encompassing the entire |
| 585 |
existing version 4 Internet. |
| 586 |
|
| 587 |
The mapping from/to version 4 IP addresses: |
| 588 |
|
| 589 |
+----------+----------+---------------+---------------------+ |
| 590 |
| length | AFI | IDI ... | DSP ... | |
| 591 |
+----------+----------+---------------+---------------------+ |
| 592 |
| 7 | 192 | AD number | version 4 address | |
| 593 |
+----------+----------+---------------+---------------------+ |
| 594 |
|
| 595 |
While the address (DSP) is initially always the 4 byte, version 4 |
| 596 |
address, it can be extended to arbitrary levels of subnetting within |
| 597 |
the existing Internet numbering plan. Hosts with DSPs longer than 4 |
| 598 |
bytes will not be able to interoperate with version 4 hosts. |
| 599 |
|
| 600 |
Novell IPX |
| 601 |
|
| 602 |
The Internetwork Packet Exchange protocol, developed by Novell based |
| 603 |
on the XNS protocol (Xerox Network System) has many of the same |
| 604 |
capabilities as the Internet and OSI protocols. At first look, it |
| 605 |
appears to confuse the network and transport layers, as IPX includes |
| 606 |
both the network layer service and the user datagram service of the |
| 607 |
transport layer, while SPX (sequenced packet exchange) includes the |
| 608 |
IPX network layer and provides service similar to TCP or TP4. This |
| 609 |
|
| 610 |
|
| 611 |
|
| 612 |
<span class="grey">McGovern & Ullmann [Page 11]</span></pre> |
| 613 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-12" ></span> |
| 614 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 615 |
|
| 616 |
|
| 617 |
turns out to be mostly a matter of the naming and ordering of fields |
| 618 |
in the packets, rather than any architectural difference. |
| 619 |
|
| 620 |
IPX uses a 32-bit LAN network number, implicitly concatenated with |
| 621 |
the 48-bit MAC layer address to form an Internet address. Initially, |
| 622 |
the network numbers were not assigned by any central authority, and |
| 623 |
thus were not useful for inter-organizational traffic without |
| 624 |
substantial prior arrangement. There is now an authority established |
| 625 |
by Novell to assign unique 32-bit numbers and blocks of numbers to |
| 626 |
organizations that desire inter-organization networking with the IPX |
| 627 |
protocol. |
| 628 |
|
| 629 |
The Novell/IPX numbering plan uses an ICD, to be assigned, to |
| 630 |
designate an address as an IPX address. This means Novell uses the |
| 631 |
authority (AFI=47)(ICD=Novell) and delegates assignments of the |
| 632 |
following 32 bits. |
| 633 |
|
| 634 |
An IPX address in the common form looks like: |
| 635 |
|
| 636 |
+----------+----------+---------------+---------------------+ |
| 637 |
| length | AFI | IDI ... | DSP ... | |
| 638 |
+----------+----------+---------------+---------------------+ |
| 639 |
| 13 | 47 (hex) | Novell ICD | network+MAC address | |
| 640 |
+----------+----------+---------------+---------------------+ |
| 641 |
|
| 642 |
This will always be followed by two bytes of zero padding when it |
| 643 |
appears in a common network layer datagram. Note that the socket |
| 644 |
numbers included in the native form IPX address are part of the |
| 645 |
transport layer. |
| 646 |
|
| 647 |
SIPP |
| 648 |
|
| 649 |
It may seem a little odd to describe the interaction with SIPP-16 |
| 650 |
(version 6 of IP) which is another proposed candidate for the next |
| 651 |
generation of network layer protocols. However, if SIPP-16 is |
| 652 |
deployed, whether or not as the protocol of choice for replacement of |
| 653 |
IP version 4, there will then be four network protocols to |
| 654 |
accommodate. It is prudent to investigate how SIPP-16 could then be |
| 655 |
integrated into the common addressing plan and datagram format. |
| 656 |
|
| 657 |
SIPP-16 defines 128 bit addresses, which are included in the NSAP |
| 658 |
addressing plan under the Internet AFI as AD number 0.1. It is not |
| 659 |
clear at this time what administration will hold the authority for |
| 660 |
the SIPP-16 numbering plan. This produces a 20 byte NSAPA, with the |
| 661 |
system ID field positioned exactly as expected by (e.g.) IS-IS. |
| 662 |
|
| 663 |
|
| 664 |
|
| 665 |
|
| 666 |
|
| 667 |
|
| 668 |
<span class="grey">McGovern & Ullmann [Page 12]</span></pre> |
| 669 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-13" ></span> |
| 670 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 671 |
|
| 672 |
|
| 673 |
+----------+----------+---------------+---------------------+ |
| 674 |
| length | AFI | IDI ... | DSP ... | |
| 675 |
+----------+----------+---------------+---------------------+ |
| 676 |
| 19 | 192 | AD (0.1) | SIPP-16 address | |
| 677 |
+----------+----------+---------------+---------------------+ |
| 678 |
|
| 679 |
The SIPP-16 addressing method (the definition of the 128 bits) will |
| 680 |
not be described here. |
| 681 |
|
| 682 |
The SIPP proposal also includes an encapsulated-tunnel proposal |
| 683 |
called IPAE, to address some of the issues that are designed into |
| 684 |
CATNIP. The CATNIP direct translation does not use the SIPP-IPAE |
| 685 |
packet formats. IPAE also specifies a "mapping table" for prefixes. |
| 686 |
This table is kept up-to-date by periodic FTP transfers from a |
| 687 |
"central site." The CATNIP definitions leave the problem of prefix |
| 688 |
selection when converting into SIPP firmly within the scope of the |
| 689 |
SIPP-IPAE proposal, and possible methods are not described here. |
| 690 |
|
| 691 |
In translating from SIPP (IPv6) to CATNIP (IPv7), the only unusual |
| 692 |
aspect is that SIPP defines some things that are normally considered |
| 693 |
options to be "payloads" overloaded onto the transport protocol |
| 694 |
numbering space. Fortunately, the only one that need be considered |
| 695 |
is fragmentation; a fragmented SIPP datagram may need to be |
| 696 |
reassembled prior to conversion. Other "payloads" such as routing |
| 697 |
are ignored (translated verbatim) and will normally simply fail to |
| 698 |
achieve the desired effect. |
| 699 |
|
| 700 |
Translation to SIPP is simple, except for the difficult problem of |
| 701 |
inventing the "prefix" if an implementation wants to support |
| 702 |
translating Internet AD 0.0 numbers into the SIPP addressing domain. |
| 703 |
|
| 704 |
Internet DNS |
| 705 |
|
| 706 |
CATNIP addresses are represented in the DNS with the NSAP RR. The |
| 707 |
data in the resource record is the NSAP, including the zero selector |
| 708 |
at the end. The zone file syntax for the data is a string of |
| 709 |
hexadecimal digits, with a period "." inserted between any two octets |
| 710 |
where desired for readability. For example: |
| 711 |
|
| 712 |
The inverse (PTR) zone is .NSAP.INT, with the CATNIP address |
| 713 |
(reversed). That is, like .IN-ADDR.ARPA, but with .NSAP.INT instead. |
| 714 |
The nibbles are represented as hexadecimal digits. |
| 715 |
|
| 716 |
This respects the difference in actual authority: the IANA is the |
| 717 |
authority for the entire space rooted in .IN-ADDR.ARPA. in the |
| 718 |
version 4 Internet, while in the new Internet it holds the authority |
| 719 |
only for 0.C.NSAP.INT. (Following the example of 192 as the AFI |
| 720 |
value.) The domain 0.0.0.0.0.C.NSAP.INT is to be delegated by IANA to |
| 721 |
|
| 722 |
|
| 723 |
|
| 724 |
<span class="grey">McGovern & Ullmann [Page 13]</span></pre> |
| 725 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-14" ></span> |
| 726 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 727 |
|
| 728 |
|
| 729 |
the InterNIC. (Understanding that in present practice the InterNIC is |
| 730 |
the operator of the authoritative root.) |
| 731 |
|
| 732 |
Security Considerations |
| 733 |
|
| 734 |
The CATNIP design permits the direct use of the present proposals for |
| 735 |
network layer security being developed in the IPSEC WG of the IETF. |
| 736 |
There are a number of detailed requirements; the most relevant being |
| 737 |
that network layer datagram translation must not affect (cannot |
| 738 |
affect) the transport layers, since the TPDU is mostly inaccessible |
| 739 |
to the router. For example, the translation into IPX will only work |
| 740 |
if the port numbers are shadowed into the plaintext security header. |
| 741 |
|
| 742 |
References |
| 743 |
|
| 744 |
[<a id="ref-Chapin93">Chapin93</a>] Chapin, L., and D. Piscitello, "Open Systems |
| 745 |
Networking", Addison-Wesley, Reading, Massachusetts, |
| 746 |
1993. |
| 747 |
|
| 748 |
[<a id="ref-Perlman92">Perlman92</a>] Perlman, R., "Interconnections: Bridges and Routers" |
| 749 |
Addison-Wesley, Reading, Massachusetts, 1992. |
| 750 |
|
| 751 |
[<a id="ref-RFC791">RFC791</a>] Postel, J., Editor, "Internet Protocol - DARPA |
| 752 |
Internet Program Protocol Specification", STD 5, <a href="./rfc791">RFC</a> |
| 753 |
<a href="./rfc791">791</a> USC/Information Sciences Institute, September |
| 754 |
1981. |
| 755 |
|
| 756 |
[<a id="ref-RFC792">RFC792</a>] Postel, J., Editor, "Internet Control Message |
| 757 |
Protocol - DARPA Internet Program Protocol |
| 758 |
Specification", STD 5, <a href="./rfc792">RFC 792</a>, USC/Information |
| 759 |
Sciences Institute, September 1981. |
| 760 |
|
| 761 |
[<a id="ref-RFC793">RFC793</a>] Postel, J., Editor, "Transmission Control Protocol - |
| 762 |
DARPA Internet Program Protocol Specification, |
| 763 |
STD 7, <a href="./rfc793">RFC 793</a>, USC/Information Sciences Institute, |
| 764 |
September, 1981. |
| 765 |
|
| 766 |
[<a id="ref-RFC801">RFC801</a>] Postel, J., "NCP/TCP Transition Plan", <a href="./rfc801">RFC 801</a>, |
| 767 |
USC/Information Sciences Institute, November, 1981. |
| 768 |
|
| 769 |
[<a id="ref-RFC1191">RFC1191</a>] Mogul, J., and S. Deering, "Path MTU Discovery", |
| 770 |
<a href="./rfc1191">RFC 1191</a>, DECWRL, Stanford University, November, |
| 771 |
1990. |
| 772 |
|
| 773 |
[<a id="ref-RFC1234">RFC1234</a>] Provan, D., "Tunneling IPX Traffic Through IP |
| 774 |
Networks", <a href="./rfc1234">RFC 1234</a>, Novell, Inc., June 1991. |
| 775 |
|
| 776 |
|
| 777 |
|
| 778 |
|
| 779 |
|
| 780 |
<span class="grey">McGovern & Ullmann [Page 14]</span></pre> |
| 781 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-15" ></span> |
| 782 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 783 |
|
| 784 |
|
| 785 |
[<a id="ref-RFC1247">RFC1247</a>] Moy, J., "OSPF Version 2", <a href="./rfc1247">RFC 1247</a>, Proteon, Inc., |
| 786 |
July 1991. |
| 787 |
|
| 788 |
[<a id="ref-RFC1287">RFC1287</a>] Clark, D., Chapin, L., Cerf, V., Braden, R., and |
| 789 |
R. Hobby, "Towards the Future Internet Architecture", |
| 790 |
<a href="./rfc1287">RFC 1287</a>, MIT, BBN, CNRI, ISI, UCDavis, December, |
| 791 |
1991. |
| 792 |
|
| 793 |
[<a id="ref-RFC1335">RFC1335</a>] Wang, Z., and J. Crowcroft, "A Two-Tier Address |
| 794 |
Structure for the Internet: A Solution to the |
| 795 |
Problem of Address Space Exhaustion", <a href="./rfc1335">RFC 1335</a>, |
| 796 |
University College London, May 1992. |
| 797 |
|
| 798 |
[<a id="ref-RFC1338">RFC1338</a>] Fuller, V., Li, T., Yu, J., and K. Varadhan, |
| 799 |
"Supernetting: an Address Assignment and Aggregation |
| 800 |
Strategy", <a href="./rfc1338">RFC 1338</a>, BAARNet, cicso, Merit, OARnet, |
| 801 |
June 1992. |
| 802 |
|
| 803 |
[<a id="ref-RFC1347">RFC1347</a>] Callon, R., "TCP and UDP with Bigger Addresses |
| 804 |
(TUBA), A Simple Proposal for Internet Addressing |
| 805 |
and Routing", <a href="./rfc1347">RFC 1347</a>, DEC, June 1992. |
| 806 |
|
| 807 |
[<a id="ref-RFC1466">RFC1466</a>] Gerich, E., "Guidelines for Management of IP Address |
| 808 |
Space", <a href="./rfc1466">RFC 1466</a>, Merit, May 1993. |
| 809 |
|
| 810 |
[<a id="ref-RFC1475">RFC1475</a>] Ullmann, R., "TP/IX: The Next Internet", <a href="./rfc1475">RFC 1475</a>, |
| 811 |
Process Software Corporation, June 1993. |
| 812 |
|
| 813 |
[<a id="ref-RFC1476">RFC1476</a>] Ullmann, R., "RAP: Internet Route Access Protocol", |
| 814 |
<a href="./rfc1476">RFC 1476</a>, Process Software Corporation, June 1993. |
| 815 |
|
| 816 |
[<a id="ref-RFC1561">RFC1561</a>] Piscitello, D., "Use of ISO CLNP in TUBA |
| 817 |
Environments", <a href="./rfc1561">RFC 1561</a>, Core Competence, December |
| 818 |
1993. |
| 819 |
|
| 820 |
|
| 821 |
|
| 822 |
|
| 823 |
|
| 824 |
|
| 825 |
|
| 826 |
|
| 827 |
|
| 828 |
|
| 829 |
|
| 830 |
|
| 831 |
|
| 832 |
|
| 833 |
|
| 834 |
|
| 835 |
|
| 836 |
<span class="grey">McGovern & Ullmann [Page 15]</span></pre> |
| 837 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-16" ></span> |
| 838 |
<span class="grey"><a href="./rfc1707">RFC 1707</a> CATNIP October 1994</span> |
| 839 |
|
| 840 |
|
| 841 |
Authors' Addresses |
| 842 |
|
| 843 |
Michael McGovern |
| 844 |
Sunspot Graphics |
| 845 |
|
| 846 |
EMail: [email protected] |
| 847 |
|
| 848 |
|
| 849 |
Robert Ullmann |
| 850 |
Lotus Development Corporation |
| 851 |
1 Rogers Street |
| 852 |
Cambridge, MA 02142 |
| 853 |
|
| 854 |
Phone: +1 617 693 1315 |
| 855 |
EMail: [email protected] |
| 856 |
|
| 857 |
|
| 858 |
|
| 859 |
|
| 860 |
|
| 861 |
|
| 862 |
|
| 863 |
|
| 864 |
|
| 865 |
|
| 866 |
|
| 867 |
|
| 868 |
|
| 869 |
|
| 870 |
|
| 871 |
|
| 872 |
|
| 873 |
|
| 874 |
|
| 875 |
|
| 876 |
|
| 877 |
|
| 878 |
|
| 879 |
|
| 880 |
|
| 881 |
|
| 882 |
|
| 883 |
|
| 884 |
|
| 885 |
|
| 886 |
|
| 887 |
|
| 888 |
|
| 889 |
|
| 890 |
|
| 891 |
|
| 892 |
McGovern & Ullmann [Page 16] |
| 893 |
</pre> |