Network Working Group                                        R. Ullmann
Request for Comments: 1475                 Process Software Corporation
                                                             June 1993


                       TP/IX: The Next Internet

Status of this Memo

  This memo defines an Experimental Protocol for the Internet
  community.  It does not specify an Internet standard.  Discussion and
  suggestions for improvement are requested.  Please refer to the
  current edition of the "IAB Official Protocol Standards" for the
  standardization state and status of this protocol.  Distribution of
  this memo is unlimited.

Abstract

  The first version of this memo, describing a possible next generation
  of Internet protocols, was written by the present author in the
  summer and fall of 1989, and circulated informally, including to the
  IESG, in December 1989.  A further informal note on the addressing,
  called "Toasternet Part II", was circulated on the IETF mail list
  during March of 1992.

Table of Contents

  1.       Introduction . . . . . . . . . . . . . . . . . . . . 3
  1.1       Objectives  . . . . . . . . . . . . . . . . . . . . 5
  1.2       Philosophy  . . . . . . . . . . . . . . . . . . . . 6
  2.       Internet numbers . . . . . . . . . . . . . . . . . . 6
  2.1       Is 64 Bits Enough?  . . . . . . . . . . . . . . . . 6
  2.2       Why version 7?  . . . . . . . . . . . . . . . . . . 7
  2.3       The version 7 IP address  . . . . . . . . . . . . . 7
  2.4       AD numbers  . . . . . . . . . . . . . . . . . . . . 8
  2.5       Mapping of version 4 numbers  . . . . . . . . . . . 8
  3.       IP: Internet datagram protocol . . . . . . . . . . . 9
  3.1       IP datagram header format . . . . . . . . . . . .  10
  3.1.1       Version . . . . . . . . . . . . . . . . . . . .  10
  3.1.2       Header length . . . . . . . . . . . . . . . . .  10
  3.1.3       Time to live  . . . . . . . . . . . . . . . . .  10
  3.1.4       Total datagram length . . . . . . . . . . . . .  11
  3.1.5       Forward route identifier  . . . . . . . . . . .  11
  3.1.6       Destination . . . . . . . . . . . . . . . . . .  11
  3.1.7       Source  . . . . . . . . . . . . . . . . . . . .  11
  3.1.8       Protocol  . . . . . . . . . . . . . . . . . . .  11
  3.1.9       Checksum  . . . . . . . . . . . . . . . . . . .  11
  3.1.10      Options . . . . . . . . . . . . . . . . . . . .  11



Ullmann                                                         [Page 1]

RFC 1475                         TP/IX                         June 1993


  3.2       Option Format . . . . . . . . . . . . . . . . . .  12
  3.2.1       Class (C) . . . . . . . . . . . . . . . . . . .  12
  3.2.2       Copy on fragmentation (F) . . . . . . . . . . .  13
  3.2.3       Type  . . . . . . . . . . . . . . . . . . . . .  13
  3.2.4       Length  . . . . . . . . . . . . . . . . . . . .  13
  3.2.5       Option data . . . . . . . . . . . . . . . . . .  13
  3.3       IP options  . . . . . . . . . . . . . . . . . . .  13
  3.3.1       Null  . . . . . . . . . . . . . . . . . . . . .  13
  3.3.2       Fragment  . . . . . . . . . . . . . . . . . . .  14
  3.3.3       Last Fragment . . . . . . . . . . . . . . . . .  14
  3.3.4       Don't Fragment  . . . . . . . . . . . . . . . .  15
  3.3.5       Don't Convert . . . . . . . . . . . . . . . . .  15
  3.4       Forward route identifier  . . . . . . . . . . . .  15
  3.4.1       Procedure description . . . . . . . . . . . . .  15
  3.4.2       Flows . . . . . . . . . . . . . . . . . . . . .  17
  3.4.3       Mobile hosts  . . . . . . . . . . . . . . . . .  17
  4.       TCP: Transport protocol  . . . . . . . . . . . . .  18
  4.1       TCP segment header format . . . . . . . . . . . .  18
  4.1.1       Data offset . . . . . . . . . . . . . . . . . .  19
  4.1.2       MBZ . . . . . . . . . . . . . . . . . . . . . .  19
  4.1.3       Flags . . . . . . . . . . . . . . . . . . . . .  19
  4.1.4       Checksum  . . . . . . . . . . . . . . . . . . .  19
  4.1.5       Source port . . . . . . . . . . . . . . . . . .  20
  4.1.6       Destination port  . . . . . . . . . . . . . . .  20
  4.1.7       Sequence  . . . . . . . . . . . . . . . . . . .  20
  4.1.8       Acknowledgement . . . . . . . . . . . . . . . .  20
  4.1.9       Window  . . . . . . . . . . . . . . . . . . . .  20
  4.1.10      Options . . . . . . . . . . . . . . . . . . . .  20
  4.2       Port numbers  . . . . . . . . . . . . . . . . . .  20
  4.3       TCP options . . . . . . . . . . . . . . . . . . .  21
  4.3.1       Option Format . . . . . . . . . . . . . . . . .  21
  4.3.2       Null  . . . . . . . . . . . . . . . . . . . . .  21
  4.3.3       Maximum Segment Size  . . . . . . . . . . . . .  21
  4.3.4       Urgent Pointer  . . . . . . . . . . . . . . . .  21
  4.3.5       32 Bit rollover . . . . . . . . . . . . . . . .  21
  5.       UDP: User Datagram protocol  . . . . . . . . . . .  22
  5.1       UDP header format . . . . . . . . . . . . . . . .  22
  5.1.1       Data offset . . . . . . . . . . . . . . . . . .  22
  5.1.2       MBZ . . . . . . . . . . . . . . . . . . . . . .  22
  5.1.3       Checksum  . . . . . . . . . . . . . . . . . . .  22
  5.1.4       Source port . . . . . . . . . . . . . . . . . .  22
  5.1.5       Destination port  . . . . . . . . . . . . . . .  22
  5.1.6       Options . . . . . . . . . . . . . . . . . . . .  23
  6.       ICMP . . . . . . . . . . . . . . . . . . . . . . .  23
  6.1       ICMP header format  . . . . . . . . . . . . . . .  23
  6.2       Conversion failed ICMP message  . . . . . . . . .  23
  7.       Notes on the domain system . . . . . . . . . . . .  25
  7.1       A records . . . . . . . . . . . . . . . . . . . .  25



Ullmann                                                         [Page 2]

RFC 1475                         TP/IX                         June 1993


  7.2       PTR zone  . . . . . . . . . . . . . . . . . . . .  25
  8.       Conversion between version 4 and version 7 . . . .  25
  8.1       Version 4 IP address extension option . . . . . .  26
  8.1.1     Option format . . . . . . . . . . . . . . . . . .  26
  8.2      Fragmented datagrams . . . . . . . . . . . . . . .  26
  8.3      Where does the conversion happen?  . . . . . . . .  27
  8.4      Hybrid IPv4 systems  . . . . . . . . . . . . . . .  28
  8.5      Maximum segment size in TCP  . . . . . . . . . . .  28
  8.6      Forwarding and redirects . . . . . . . . . . . . .  28
  8.7      Design considerations  . . . . . . . . . . . . . .  28
  8.8      Conversion from IPv4 to IPv7 . . . . . . . . . . .  29
  8.9      Conversion from IPv7 to IPv4 . . . . . . . . . . .  30
  8.10     Conversion from TCPv4 to TCPv7 . . . . . . . . . .  31
  8.11     Conversion from TCPv7 to TCPv4 . . . . . . . . . .  32
  8.12     ICMP conversion  . . . . . . . . . . . . . . . . .  33
  9.       Postscript . . . . . . . . . . . . . . . . . . . .  33
  10.      References . . . . . . . . . . . . . . . . . . . .  34
  11.      Security Considerations  . . . . . . . . . . . . .  35
  12.      Author's Address . . . . . . . . . . . . . . . . .  35

1.  Introduction

  This memo presents the specification for version 7 of the Internet
  Protocol, as well as version 7 of the TCP and the user datagram
  protocol.  Version 7 has been designed to address several major
  problems that have arisen as version 4 has evolved and been deployed,
  and to make a major step forward in the datagram switching and
  forwarding architecture of the Internet.

  The major problems are threefold.  First, the address space of
  version 4 is now seen to be too small.  While it was viewed as being
  almost impossibly large when version 4 was designed, two things have
  occurred to create a problem.  The first is a success crisis:  the
  internet protocols have been more widely used and accepted than their
  designers anticipated.  Also, technology has moved forward, putting
  microprocessors into devices not anticipated except as future dreams
  a decade ago.

  The second major problem is a perceived routing explosion.  The
  present routing architecture of the internet calls for routing each
  organization's network independently.  It is becoming increasingly
  clear that this does not scale to a universal internet.  While it is
  possible to route several billion networks in a flat, structureless
  domain, it is not desireable.

  There is also the political administrative issue of assigning network
  numbers to organizations.  The version 4 administrative system calls
  for organizations to request network assignments from a single



Ullmann                                                         [Page 3]

RFC 1475                         TP/IX                         June 1993


  authority.  While to some extent this has been alleviated by
  reserving blocks to delegated assignments, the address space is not
  large enough to do this in the necessary general case, with large
  blocks allocated to (e.g.) national authority.

  The third problem is the increasing bandwidth of the networks and of
  the applications possible on the network.  The TCP, while having
  proven useful on an unprecedented range of network speeds, is now the
  limiting factor at the highest speeds, due to restrictions of window
  size, sequence-space, and port numbers.  These limitations can all be
  addressed by increasing the sizes of the relevant fields.  See
  [RFC1323].

  There is also an opportunity to move the technology forward, and take
  advantage of a combination of the best features of the hop-by-hop
  connectionless forwarding of version 4 (and CLNP) as well as the
  pre-established paths of version 5 (and, e.g., the OSI CONS).

  Internet Version 7 includes four major areas of improvement, while at
  the same time retaining interoperation with version 4 with a small
  amount of conversion knowledge imposed on version 7 hosts and
  routers.

     o  It increases the address fields to 64 bits, with sufficient
        space for visible future expansion of the internet.

     o  It adds a numbering layer for administrations, above the
        organization or network layer, as well as providing more
        space for subnetting within organizations.

     o  It increases the range of speeds and network path delays over
        which the TCP will operate satisfactorily, as well as the
        number of transactions in bounded time that can be served by
        a host.

     o  Finally, it provides a forward route identifier in each
        datagram, to support extremely fast path, circuit, or
        flow-based forwarding, or any desired combination, while
        preserving hop-by-hop connectivity.

  The result is not just a movement sideways, deploying a new network
  layer protocol to patch current problems.  It is a significant step
  forward for network layer technology,








Ullmann                                                         [Page 4]

RFC 1475                         TP/IX                         June 1993


1.1  Objectives

  The following are some of the objectives of the design.

 o  Use what has been learned from the IP version 4 protocol, fixing
    things that are troublesome, and not fixing that which is not
    broken.

 o  Retain the essential "look and feel" of the Internet protocol
    suite.  It has been very successful, and one doesn't argue with
    success.

 o  Not introduce concepts that the Internet has shown do not belong
    in the protocol definition.  Best example:  we do not want to add
    any kind of routing information into the addressing, other than
    the administrative hierarchy that has sometimes proved useful.
    Note that the one feature in version 4 addressing (the class
    system) designed to aid routing is now the most serious single
    problem.

 o  Allow current hosts to interoperate, if not universally, at least
    within an organization or larger area for the indefinite future.
    There will be version 4 hosts for 10-15 years into the future,
    the Internet must remain on good terms with them.

 o  Likewise, we must not impose the new version, telling sites they
    must convert to stay connected.  People resist imposed solutions.
    It must not be marketed as something different from IPv4; the
    differences must be down-played at every opportunity.

 o  The design must allow individual hosts and routers to be upgraded
    effectively at random, with no transition plan constraints.

 o  The design must not require renumbering the Internet.  The
    administrative work already accomplished is immense, if it is to
    be done again it will be in assigning NSAPs.

 o  It must allow IPv4 hosts to interoperate without any reduction in
    function, without any modification to their software or
    configuration.  (Universal connectivity will be lost by IPv4
    hosts, but they must be able to continue operating within their
    organization at least.)

 o  It must permit network layer state-free translation of datagrams
    between IPv4 and IPv7; this is important to the previous point,
    and essential to early testing and transitional deployment.

 o  It must be a competent alternative to CLNP.



Ullmann                                                         [Page 5]

RFC 1475                         TP/IX                         June 1993


 o  It must not involve changing the semantics of the network layer
    service in any way that invalidates the huge amount of work that
    has gone into understanding how TCP (for example) functions in
    the net, and the implementation of that understanding.

 o  It must be defined Real Soon; the window of opportunity is almost
    closed.  It will take vendors 3 years to deploy from the time the
    standard is rock-solid concrete.

  I believe all of these are accomplishable in a consistent, well-
  engineered solution, and all are essential to the survival of the
  Internet.

1.2  Philosophy

  Protocols should become simpler as they evolve.

2.  Internet numbers

  The version 4 numbering system has proven to be very flexible,
  (mostly) expandable, and simple.  In short:  it works.  There are two
  problems, neither serious when this specification was first developed
  in 1988 and 1989, but have as expected become more serious:

     o  The division into network, and then subnet, is insufficient.
        Almost all sites need a network assignment large enough to
        subnet.  At the top of the hierarchy, there is a need to
        assign administrative domains.

     o  As bit-packing is done to accomplish the desired network
        structure, the 32 bit limit causes more and more aggravation.

2.1  Is 64 Bits Enough?

  Consider:  (thought experiment) 32 bits presently numbers "all" of
  the computers in the world, and another 32 bits could be used to
  number all of the bytes of on-line storage on each computer.  (Most
  have a lot less than 4 gigabytes on-line, the ones that have more
  could be notionally assigned more than one address.)

  So: 64 bits is enough to number every byte of online storage in
  existence today, in a hierarchical structured numbering plan.

  Another way of looking at 64 bits:  it is more than 2 billion
  addresses for each person on the planet.  Even if I have
  microprocessors in my shirt buttons I'm not going to have that many.
  32 bits, on the other hand, was never going to be sufficient:  there
  are more than 2^32 people.



Ullmann                                                         [Page 6]

RFC 1475                         TP/IX                         June 1993


2.2  Why version 7?

  It was clearly recognized at the start of this project in 1988 that
  making the address 64 bits implies a new IP header format, which was
  called either "TP/IX" or "IP version 7"; there wasn't anything magic
  about the number 7, I made it up.  Version 4 is the familiar current
  version of IP.  Version 5 is the experimental ST (Stream) protocol.
  ST-II, a newer version of ST, uses the same version number, something
  I was not aware of until recently; I suspected it might have been
  allocated 6.  Besides, I liked 7.

  Apparently (as reported by Bob Braden) the IAB followed much the same
  logic, and may have had the idea planted by the mention of version 7
  in the "Toasternet Part II" memo.  The IAB in June 1992 floated a
  proposal that CLNP, or a CLNP-based design, be Internet Version 7.
  (And promptly got themselves toasted.) However, close inspection of
  the bits shows that CLNP is clearly version 8.

2.3  The version 7 IP address

  The Version 7 IP 64 bit address looks like:

   +-------+-------+-------+-------+-------+-------+-------+-------+
   |      Admin Domain     |        Network        |     Host      |
   +-------+-------+-------+-------+-------+-------+-------+-------+

  Note:  the boundary between "network" and "host" is no more fixed
  than it is today; each (sub)network will have its own mask.  Just as
  the mask today can be anywhere from FF00 0000 (8/24) to FFFF FFFC
  (30/2), the mask for the 64 bit address can reasonably be FFFF FF00
  0000 0000 (24/40) to FFFF FFFF FFFF FFFC (62/2).

  The AD (Administrative Domain), identifies an administration which
  may be a service provider, a national administration, or a large
  multi-organization (e.g.  a government).  The idea is that there
  should not be more than a few hundred of these at first, and
  eventually thousands or tens of thousands at most.  (But note that we
  do not introduce a hard limit of 2^16 here; this estimate may be off
  by a few orders of magnitude.) Since only 1/4th of the address space
  is initially used (first two bits are 01), the remainder can then be
  allocated in the future with more information available.

  Most individual organizations would not be ADs.  In the short term,
  ADs are known to the "core routing"; it pays to keep the number
  smallish, a few thousand given current routing technology.  In the
  long term, this is not necessary.  Big administrations (i.e., with
  tens of millions of networks) get small blocks where needed, or
  additional single AD numbers when needed.



Ullmann                                                         [Page 7]

RFC 1475                         TP/IX                         June 1993


  While the AD may be used for last resort routing (with a 24/40 mask),
  it is primarily only an administrative device.  Most routing will be
  done on the entire 48 bit AD+network number, or sub and super-sets of
  those numbers.  (I.e., masks between about 32/32 and 56/8.)

  Some ADs (e.g., NSF) may make permanent assignments; others (such as
  a telephone company defining a network number for each subscriber
  line) may tie the assignment to such a subscription.  But in no case
  does this require traffic to be routed via the AD.

2.4  AD numbers

  AD numbers are allocated out of the same numbering space as network
  numbers.  This means that a version 4 address can be distinguished
  from the first 32 bits of a version 7 address.  This is useful to
  help prevent the inadvertent use of the first half of the longer
  address by a version 4 host.

  There is a non-trivial amount of software that assumes that an "int"
  is the same size and shape as an IP address, and does things like
  "ipaddr = *(int *)ptr".  This usage has always been incorrect, but
  does occur with disturbing frequency.  As IPv7 8 byte addresses
  appear in the application layers, this software will find those
  addresses unreachable; this is preferable to interacting with a
  random host.

  One possible method would be to allocate ADs in the range 96.0.0 to
  192.255.255, using the top 1/4 of the version 4 class A space.  It is
  probably best to allocate the first component downwards from 192, so
  that the boundary between class A and AD can be moved if desired
  later.  This initial allocation provides for 2031616 ADs, many more
  than there should be even in full deployment.

  Eventually, both AD and network will use the full 24 bit space
  available to them.  Knowledge of the AD range should not be coded
  into software.  If it was coded in, that software would break when
  the entire 24 bit space is used for ADs.  (This lesson should have
  been learned from CIDR.)

2.5  Mapping of version 4 numbers

  Initially, all existing Internet numbers are defined as belonging to
  the NSF/Internet AD, number 192.0.0.








Ullmann                                                         [Page 8]

RFC 1475                         TP/IX                         June 1993


  The mapping from/to version 4 IP addresses:

   +-------+-------+-------+-------+-------+-------+-------+-------+
   |      Admin Domain     |        Network        |     Host      |
   +-------+-------+-------+-------+-------+-------+-------+-------+
    [  fixed at A0 00 00  ] [ 1st 24 bits of V4 IP]   [1]   [last 8]

  So, for example, 192.42.95.15 (V4) becomes 192.0.0.192.42.95.1.15.

  And the "standard" loopback interface address becomes
  192.0.0.127.0.0.1.1 (I can see explaining that in 2015 to someone
  born in 1995.)

  The present protocol multicast (192.0.0.224.x.y.1.z) and loopback
  addresses are permanently allocated in the NSF AD.

3.  IP:  Internet datagram protocol

  The Internet datagram protocol is revised to expand some fields (most
  notably the addresses), while removing and relegating to options all
  fields not universally useful (imperative) in every datagram in every
  environment.

  This results in some simplification, a length less than twice the
  size of IPv4 even though most fields are doubled in size, and an
  expanded space for options.

  There is also a change in the option philosophy from IPv4:  it
  specified that implementation of options was not optional, what was
  optional was the existence of options in any given datagram.  This is
  changed in IPv7:  no option need be implemented to be fully
  conformant.  However, implementations must understand the option
  classes; and a future Host Requirements specification for hosts and
  routers used in the "connected Internet" may require some options in
  its profile, e.g., Fragment would probably be required.

  Digression:  In IPv4, options are often "considered harmful".  It is
  the opinion of the present author that this is because they are
  rarely needed, and not designed to be processed rapidly on most
  architectures.  This leads to little or no attempt to improve
  performance in implementations, while at the same time enormous
  effort is dedicated to optimization of the no-option case.  IPv7 is
  expected to be different on both counts.

  Fields are always aligned on their own size; the 64 bit fields on 64
  bit intervals from the start of the datagram.

  Options are all 32 bit aligned, and the null option can be used to



Ullmann                                                         [Page 9]

RFC 1475                         TP/IX                         June 1993


  push a subsequent option (or the transport layer header) into 64 bit
  or 64+32 off-phase alignment as desired.

3.1  IP datagram header format

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |version|     header length     |        time to live           |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        total datagram length                                  |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +        forward route identifier                               +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +        destination address                                    +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +        source address                                         +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        protocol               |           checksum            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        options                                                |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  A description of each field follows.

3.1.1  Version

  This document describes version 7 of the protocol.

3.1.2  Header length

  The header length is a 12 bit count of the number of 32 bit words in
  the IPv7 header.  This allows a header to be (theoretically at least)
  up to 16380 bytes in length.

3.1.3  Time to live

  The time to live is a 16 bit count, nominally in 1/16 seconds.  Each
  hop is required to decrement TTL by at least one.

  This definition should allow continuation of the useful (even though
  not entirely valid) interpretation of TTL as a hop count, while we



Ullmann                                                        [Page 10]

RFC 1475                         TP/IX                         June 1993


  move to faster networks and routers.  (The most familiar use is by
  "traceroute", which really ought to be directly implemented by one or
  more ICMP messages.)

  The scale factor converts the usual version 4 default TTL into a
  larger number of hops.  This is desireable because the forward route
  architecture of version 7 enables the construction of simpler, faster
  switches, and this may cause the network diameter to increase.

3.1.4  Total datagram length

  The 32 bit length of the entire datagram in octets.  A datagram can
  therefore be up to 4294967295 bytes in overall length.  Particular
  networks will normally impose lower limits.

3.1.5  Forward route identifier

  The identifier from the routing protocol to be used by the next hop
  router to find its next hop.  (A more complete description is given
  below.)

3.1.6  Destination

  The 64 bit IPv7 destination address.

3.1.7  Source

  The 64 bit IPv7 source address.

3.1.8  Protocol

  The transport layer protocol, e.g., TCP is 6.  The present code space
  for this layer of demultiplexing is about half full.  Expanding it to
  16 bits, allowing 65535 registered "transport" layers seems prudent.

3.1.9  Checksum

  The checksum is a 16 bit checksum of the entire IP header, using the
  familiar algorithm used in IPv4.

3.1.10  Options

  Options may follow.  They are variable length, and always 32 bit
  aligned, as discussed previously.







Ullmann                                                        [Page 11]

RFC 1475                         TP/IX                         June 1993


3.2  Option Format

  Each option begins with a 32 bit header:

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | C |F|    type                 |   length                      |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        option data                 ...          |   padding   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  A description of each field:

3.2.1  Class (C)

  This field tells implementations what to do with datagrams containing
  options they do not understand.  No implementation is required to
  implement (i.e., understand) any given option by the TCP/IP
  specification itself.

  Classes:

      0        use or forward and include this option unmodified
      1        use this datagram, but do not forward the datagram
      2        discard, or forward and include this option unmodified
      3        discard this datagram

  A host receiving a datagram addressed to itself will use it if there
  are no unknown options of class 2 or 3.  A router receiving a
  datagram not addressed to it will forward the datagram if and only if
  there are no unknown options of class 1 or 3.  (The astute reader
  will note that the bits can also be seen as having individual
  interpretations, one allowing use even if unknown, one allowing
  forwarding if unknown.)

  Note that classes 0 and 2 are imperative:  if the datagram is
  forwarded, the unknown option must be included.

  Class and type are entirely orthogonal, different implementations
  might use different classes for the same option, except where
  restricted by the option definition.

  Also note that for options that are known (implemented by) the host
  or router, the class has no meaning; the option definition totally
  determines the behavior.  (Although it should be noted that the
  option might explicitly define a class dependent behavior.)




Ullmann                                                        [Page 12]

RFC 1475                         TP/IX                         June 1993


3.2.2  Copy on fragmentation (F)

  If the F bit is set, this option must be copied into all fragments
  when a datagram is fragmented.  If the F bit is reset (zero), the
  option must only be copied into the first (zero-offset) fragment.

3.2.3  Type

  The type field identifies the particular option, types being
  registered as well known values in the internet.  A few of the
  options with their types are described below.

3.2.4  Length

  Length of the option data, in bytes.

3.2.5  Option data

  Variable length specified by the length field, plus 0-3 bytes of
  zeros to pad to a 32 bit boundary.  Fields within the option data
  that are 64 bits long are normally placed on the assumption that the
  option header is off-phase aligned, the usual case when the option is
  the only one present, and immediately follows the IP header.

3.3  IP options

  The following sections describe the options defined to emulate IPv4
  features, or necessary in the basic structure of the protocol.

3.3.1  Null

  The null option, type 0, provides for a space filler in the option
  area.  The data may be of any size, including 0 bytes (perhaps the
  most useful case.)

  It may be used to change alignment of the following options or to
  replace an option being deleted, by setting type to 0 and class to 0,
  leaving the length and content of the data unmodified.  (Note that
  this implies that options must not contain "secret" data, relying on
  class 3 to prevent the data from leaving the domain of routers that
  understand the option.)

  Null is normally class 0, and need not be implemented to serve its
  function.







Ullmann                                                        [Page 13]

RFC 1475                         TP/IX                         June 1993


3.3.2  Fragment

  Fragment (type 1) indicates that the datagram is part of a complete
  IP datagram.  It is always class 2.

  The data consists of (one of) the 64 bit IP address(es) of the router
  doing the fragmentation, a 64 bit datagram ID generated by that
  router, and a 32 bit fragment offset.  The IDs should be generated so
  as to be very likely unique over a period of time larger than the TCP
  MSL (maximum segment lifetime).  (The TCP ISN (initial sequence
  number) generator might be used to initialize the ID generator in a
  router.)

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   | C |F|    type                 |   length                      |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +          fragmenting router IP address                        +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +          datagram ID                                          +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |          offset                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  If a datagram must be refragmented, the original 128 bit address+ID
  is preserved, so that the datagram can be reassembled from any
  sufficient set of the resulting fragments.  The 64 bits fields are
  positioned so that they are aligned in the usual case of the fragment
  option following the IP header.

  A router implementing Fragment (doing fragmentation) must recognize
  the Don't Fragment option.

3.3.3  Last Fragment

  Last Fragment (type 2) has the same format as Fragment, but implies
  that this datagram is the last fragment needed to reassemble the
  original datagram.

  Note that an implementation can reasonably add arriving datagrams
  with Fragment to a cache, and then attempt a reassembly when a
  datagram with Last Fragment arrives (and the the total length is
  known); this will work well when datagrams are not reordered in the



Ullmann                                                        [Page 14]

RFC 1475                         TP/IX                         June 1993


  network.

3.3.4  Don't Fragment

  This option (type 3, class 0) indicates that the datagram may not be
  fragmented.  If it can not be forwarded without fragmentation, it is
  discarded, and the appropriate ICMP message sent.  (Unless, of
  course, the datagram is an ICMP message.) There is no data present.

3.3.5  Don't Convert

  The Don't Convert option prohibits conversion from IPv7 to IPv4
  protocol, requiring instead that the datagram be discarded and an
  ICMP message sent (conversion failed/don't convert set).  It is type
  4, usually class 0, and must be implemented by any router
  implementing conversion.  A host is under no such constraint; like
  any protocol specification, only the "bits on the wire" can be
  specified, the host receiving the datagram may convert it as part of
  its procedure.  There is no data present in this option.

3.4  Forward route identifier

  Each IP datagram carries a 64 bit field, called "forward route
  identifier", that is updated (if the information is available) at
  each hop.  This field's value is derived from the routing protocol
  (e.g., RAP [RFC1476]).  It is used to expedite routing decisions by
  preserving knowledge where possible between consecutive routers.  It
  can also be used to make datagrams stay within reserved flows and
  mobile-host tunnels where required.

3.4.1  Procedure description

  Consider 3 routers, A, B, and C.  Traffic is passing through them,
  between two other hosts (or networks), X and Y, packets are going
  XABCY and YCBAX.  Consider only one direction:  routing info flowing
  from C to A, to provide a route from A to C.  The same thing will be
  happening in the other direction.

  An explanation of the notation:

    R(r,d,i,h)    A route that means: "from router r, to go toward
                  final destination d, replace the forward route
                  identifier in the packet with i, and take next
                  hop h."

    Ri(r,d)       An opaque (outside of router r) identifier, that can
                  be used by r to find R(r,d,...).




Ullmann                                                        [Page 15]

RFC 1475                         TP/IX                         June 1993


    Flowi(r,rt)   An opaque (outside of router r) identifier, that
                  router r can use to find a flow or tunnel with which
                  the datagram is associated, and from that the route
                  rt on which the flow or tunnel is built, as well as
                  the Flowi() for the subsequent hop.

    Ri(Dgram)     The forward route identifier in a datagram.

  Router C announces a route R(C,Y,0,Y) to router B.  It includes in it
  an identifier Ri(C,Y) internal to C, that will allow C to find the
  route rapidly.  (A table index, or an actual memory address.)

  Router B creates a route R(B,Y,Ri(C,Y),C) via router C, it announces
  it to A, including an identifier Ri(B,Y), internal to B, and used by
  A as an opaque object.

  Router A creates a route R(A,Y,Ri(B,Y),B) via router B.  It has no
  one to announce it to.

  Now:  X originates a datagram addressed to Y.  It has no routing
  information, and sets Ri(Dgram) to zero.  It forwards the datagram to
  router A (X's default gateway).

  A finds no valid Ri(Dgram), and looks up the destination (Y) in its
  routing tables.  It finds R(A,Y,Ri(B,Y),B), sets Ri(Dgram) <-
  Ri(B,Y), and forwards the datagram to B.

  Router B looks at Ri(Dgram) which directly identifies the next hop
  route R(B,Ri(C,Y),C), sets Ri(Dgram) <- Ri(C,Y) and forwards it to
  router C.

  Router C looks at Ri(Dgram) which directly locates R(C,0,Y), sets
  Ri(Dgram) <- 0 and forwards to Y.

  Y recognizes its own address in Dest(Dgram), ignores Ri(Dgram).

  Of course, the routers will validate the Ri's received, particularily
  if they are memory addresses (e.g., M(a) < Ri < M(b), Ri mod N == 0),
  and probably check that the route in fact describes the destination
  of the datagram.  If the Ri is invalid, the router must use the
  ordinary method of finding a route (i.e., what it would have done if
  Ri was 0), and silently ignore the invalid Ri.

  When a route has been aggregated at some router, implicitly or
  explicitly, it will find that the incoming Ri(Dgram) at most can
  identify the aggregation, and it must make a decision; the forwarded
  datagram then contains the Ri for the specific route.  (Note this may
  happen well upstream of the point at which the routes actually



Ullmann                                                        [Page 16]

RFC 1475                         TP/IX                         June 1993


  diverge.)

  This allows all cooperating routers to make immediate forwarding
  decisions, without any searching of tables or caches once the
  datagram has entered the routing domain.  If the host participates in
  the routing, at least to the extent of acquiring the initial Ri
  required from the first router, then only routers that have done
  aggregations need make decisions.  (If the routing changes with
  datagrams in flight, some router will be required to make a decision
  to re-rail each datagram.)

3.4.2  Flows

  If a "flow" is to be set up, the identifiers are replaced by
  Flowi(router,route), where each router's structure for the flow
  contains a pointer to the route on which the flow is built.
  Datagrams can drop out of the flow at some point, and can be inserted
  either by the originating host or by a cooperating router near the
  originator.  Since the forward route identifier field is opaque to
  the sending router, and implicitly meaningful only to the next hop
  router, use for flows (or similar optimizations) need not be
  otherwise defined by the protocol.  (One presumes that a router
  issuing both Ri's and Flowi's will take care to make sure that it can
  distinguish them by some private method.)

  If a flow has been set up by a restricted target RAP route
  announcement, it is no different from a route in the implementation.
  If this announcement originates from the host itself, the Ri in
  incoming datagrams can be used to determine whether they followed the
  flow, or to optimize delivery of the datagrams to the next layer
  protocol.

3.4.3  Mobile hosts

  First, a definition:  A "mobile host" is a host that can move around,
  connecting via different networks at different times, while
  maintaining open TCP connections.  It is distinguished from a
  "portable host", which is simply a host that can appear in various
  places in the net, without continuity.  A portable host can be
  implemented by assigning a new address for each location (more or
  less automatically), and arranging to update the domain system.
  Supporting truly mobile hosts is the more interesting problem.

  To implement mobile host support in a general way, either some layer
  of the protocol suite must provide network-wide routing, or the
  datagrams must be tunnelled from the "home" network of the host to
  its present location.  In the real network, some combination of these
  is probable:  most of the net will forward datagrams toward the home



Ullmann                                                        [Page 17]

RFC 1475                         TP/IX                         June 1993


  network, and then the datagrams will follow a specific host route to
  the mobile host.

  The requirement on the routing system is that it must be able to
  propagate a host route at least to the home network; any other
  distribution is useful optimization.  When a host route is propagated
  by RAP as a targeted route, and the routers use the resulting Ri's,
  the datagram follows an effective tunnel to the mobile host.  (Not a
  real tunnel, in the strict sense; the datagrams are following an
  actual route at the network protocol layer.)

  As explained in RAP [RFC14XX-RAP], a targeted route can be issued
  when desired; in particular, it can be triggered by the establishment
  of a TCP connection or by the arrival of datagrams that do not carry
  an Ri indicating that they have followed a (non-tunnel) route.

4.  TCP:  Transport protocol

  Internet version 7 expands the sizes of the sequence and
  acknowledgement fields, the window, and the port numbers.  This is to
  remove limitations in version 4 that begin to restrict throughput at
  (for example) the bandwidth of FDDI and round trip delay of more than
  60 milliseconds.  At gigabit speeds and delays typical of
  international links, the version 4 TCP would be a serious limitation.
  See [RFC1323].

  The port numbers are also expanded.  This alleviates the problem of
  going through the entire port number range with a rapid sequence of
  transactions in less than the lifetime of datagrams in the network.

4.1  TCP segment header format

  The 64 bit fields (sequence and acknowledgement) in the TCP header
  are off-phase aligned, in anticipation of the usual case of the TCP
  header following the 9 32-bit word IP header.  If IP options add up
  to an odd number of 32 bit words, a null option may be added to push
  the transport header to off-phase alignment.














Ullmann                                                        [Page 18]

RFC 1475                         TP/IX                         June 1993


    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  data offset  | MBZ |A|P|R|S|F|           checksum            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        source port                                            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        destination port                                       |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +        sequence number                                        +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                                                               |
   +        acknowledgement number                                 +
   |                                                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        window                                                 |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        options                          ...                   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  A description of each field:

4.1.1  Data offset

  An 8 bit count of the number of 32 bit words in the TCP header,
  including any options.

4.1.2  MBZ

  Reserved bits, must be zero, and must be ignored.

4.1.3  Flags

  These are the protocol state flags, use exactly as in TCPv4, except
  that there is no urgent data flag.

4.1.4  Checksum

  This is a 16 bit checksum of the segment.  The pseudo-header used in
  the checksum consists of the destination address, the source address,
  the protocol field (constant 6 for TCP), and the 32 bit length of the
  TCP segment.







Ullmann                                                        [Page 19]

RFC 1475                         TP/IX                         June 1993


4.1.5  Source port

  The source port number, a 32 bit identifier.  See the section on port
  numbers below.

4.1.6  Destination port.

  The 32 bit destination port number.

4.1.7  Sequence

  A 64 bit sequence number, the sequence number of the first octet of
  user data in the segment.

  The ISN (Initial Sequence Number) generator used in TCPv4 is used in
  TCPv7, with the 32 bit value loaded into both the high and low 32
  bits of the TCPv7 sequence number.  This provides reasonable behavior
  when the 32 bit rollover option is used (see below) for TCPv4
  interoperation.  V7 hosts must implement the full 64 bit sequence
  number rollover.

4.1.8  Acknowledgement

  The 64 bit acknowledgement number, acknowledging receipt of octets up
  to but not including the octet identified.  Valid if the A flag is
  set, if A is reset (0), this field should be zero, and must be
  ignored.

4.1.9  Window

  The 32 bit offered window.

4.1.10  Options

  TCP options, some of which are documented below.

4.2  Port numbers

  Port numbers are divided into several ranges:  (all numbers are
  decimal)

   0             reserved
   1-32767       Internet registered ("well-known") protocols
   32768-98303   reserved, to allow TCPv7-TCPv4 conversion
   98304 up      dynamic assignment

  It must also be remembered that hosts are free to dynamically assign
  for active connections any port not actually in use by that host:



Ullmann                                                        [Page 20]

RFC 1475                         TP/IX                         June 1993


  hosts must not reject connections because the "client" port is in the
  registered range.

4.3  TCP options

4.3.1  Option Format

  Each option begins with a 32 bit header:

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        type                   |   length                      |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        option data                 ...          |   padding   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

4.3.2  Null

  The null option (type = 0), is to be ignored.

4.3.3  Maximum Segment Size

  Maximum segment size (type = 1) specifies the largest segment that
  the other TCP should send, in terms of the number of data octets.
  When sent on a SYN segment, it is mandatory; if sent on any other
  segment it is advisory.

  Data is one 32 bit word specifying the size in octets.

4.3.4  Urgent Pointer

  The urgent pointer (type = 2) emulates the urgent field in TCPv4.
  Its presence is equivalent to the U flag being set.  The data is a 64
  bit sequence number identifying the last octet of urgent data.  (Not
  an offset, as in v4.)

4.3.5  32 Bit rollover

  The 32 bit rollover option (type = 3) indicates that only the low
  order 32 bits of the sequence and acknowledgement packets are
  significant in the packet.

  This is necessary because a converting internet layer gateway has no
  retained state, and cannot properly set the high order bits.  This
  option must be implemented by version 7 hosts that want to
  interoperate with version 4 hosts.




Ullmann                                                        [Page 21]

RFC 1475                         TP/IX                         June 1993


5.  UDP:  User Datagram protocol

  The user datagram protocol is also expanded to include larger port
  numbers, for reasons similar to the TCP.

5.1  UDP header format

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  data offset  |     MBZ       |           checksum            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        source port                                            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        destination port                                       |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        options                          ...                   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  A description of each field:

5.1.1  Data offset

  An 8 bit count of the number of 32 bit words in the UDP header,
  including any options.

5.1.2  MBZ

  Reserved bits, must be zero, and must be ignored.

5.1.3  Checksum

  This is a 16 bit checksum of the datagram.  The pseudo-header used in
  the checksum consists of the destination address, the source,
  address, and the protocol field (constant 17 for UDP), and the 32 bit
  length of the user datagram.

5.1.4  Source port

  The source port number, a 32 bit identifier.  See the section on TCP
  port numbers above.

5.1.5  Destination port.

  The 32 bit destination port number.






Ullmann                                                        [Page 22]

RFC 1475                         TP/IX                         June 1993


5.1.6  Options

  UDP options, none are presently defined.

6.  ICMP

  The ICMP protocol is very similar to ICMPv4, in some cases not
  requiring any conversion.

  The complication is that IP datagrams are nested within ICMP
  messages, and must be converted.  This is discussed later.

6.1  ICMP header format

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |     type      |     code      |           checksum            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        type-specific parameter                                |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        type-specific data               ...                   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  Type and code are well-known values, defined in [RFC792].  The codes
  have meaning only within a particular type, they are not orthogonal.

  The next 32 bit word is usually defined for the specific type,
  sometimes it is unused.

  For many types, the data consists of a nested IP datagram, usually
  truncated, which is a copy of the datagram causing the event being
  reported.  In IPv4, the nested datagram consists of the IP header,
  and another 64 bits (at least) of the original datagram.

  For IPv7, the nested datagram must include the IP header plus 96 bits
  of the remaining datagram (thus including the port numbers within TCP
  and UDP), and should include the first 256 bytes of the datagram.
  I.e., in most cases where the original datagram was not large, it
  will return the entire datagram.

6.2  Conversion failed ICMP message

  The introduction of network layer conversion requires a new message
  type, to report conversion errors.  Note that an invalid datagram
  should result in the sending of some other ICMP message (e.g.,
  parameter problem) or the silent discarding of the datagram.  This
  message is only sent when a valid datagram cannot be converted.



Ullmann                                                        [Page 23]

RFC 1475                         TP/IX                         June 1993


  Note:  implementations are not expected to, and should not, check the
  validity of incoming datagrams just to accomplish this; it simply
  means that an error detected during conversion that is known to be an
  actual error in the incoming datagram should be reported as such, not
  as a conversion failure.

  Note that the conversion failed ICMP message may be sent in either
  the IPv4 or IPv7 domain; it is a valid ICMP message type for IPv4.

    0                   1                   2                   3
    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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |     type      |     code      |           checksum            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        pointer to problem area                                |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |        copy of datagram that could not be converted ....      |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  The type for Conversion Failed is 31.

  The codes are:

       0       Unknown/unspecified error
       1       Don't Convert option present
       2       Unknown mandatory option present
       3       Known unsupported option present
       4       Unsupported transport protocol
       5       Overall length exceeded
       6       IP header length exceeded
       7       Transport protocol > 255
       8       Port conversion out of range
       9       Transport header length exceeded
       10      32 Bit Rollover missing and ACK set
       11      Unknown mandatory transport option present

  The use of code 0 should be avoided, any other condition found by
  implementors should be assigned a new code requested from IANA.  When
  code 0 is used, it is particularily important that the pointer be set
  properly.

  The pointer is an offset from the start of the original datagram to
  the beginning of the offending field.

  The data is part of the datagram that could not be converted.  It
  must be at least the IP and transport headers, and must include the
  field pointed to by the previous parameter.  For code 4, the
  transport header is probably not identifiable; the data should



Ullmann                                                        [Page 24]

RFC 1475                         TP/IX                         June 1993


  include 256 bytes of the original datagram.

7.  Notes on the domain system

7.1  A records

  Address records will be added to the IN (Internet) zone with IPv7
  addresses for all hosts as IPv7 is deployed.  Eventually the IPv4
  addresses will be removed.  As mentioned above, the AD
  (Administrative Domain) space is initially assigned so that the first
  4 octets of a v7 address cannot be confused with a v4 address (or,
  rather, the confusion will be to no effect.)

  For example:

  DELTA.Process.COM.      A       192.42.95.68
                          A       192.0.0.192.42.95.1.68

  It is important that the A record be used, to avoid the cache
  consistancy problem that would arise when different records had
  different remaining TTLs.

  Note that if an unmodified version of the more popular public domain
  nameserver is a secondary for a zone containing IPv7 addresses, it
  will erroneously issue RRs with only the first four bytes.  (I.e.,
  192.0.0.192 in the example.) This is another reason to ensure that
  the AD numbers are initially reserved out of the IPv4 network number
  space.  Eventually, zones with IPv7 addresses would be expected to be
  served only by upgraded servers.

7.2  PTR zone

  The inverse (PTR) zone is .#, with the IPv7 address (reversed).
  I.e., just like .IN-ADDR.ARPA, but with .# instead.

  This respects the difference in actual authority:  the NSF/DDN NIC is
  the authority for the entire space rooted in .IN-ADDR.ARPA.  in the
  v4 Internet, while in the new Internet it holds the authority only
  for the AD 0.0.192.#.  (Plus, of course, any other ADs assigned to it
  over time.)

8.  Conversion between version 4 and version 7

  As noted in the description of datagram format, it is possible to
  provide a mostly-transparent bridge between version 4 and version 7.

  This discusses TCP and ICMP at the session/transport layer; UDP is a
  subset of the TCP conversion.  Most protocols at this layer will



Ullmann                                                        [Page 25]

RFC 1475                         TP/IX                         June 1993


  probably need no translation; however it will probably be necessary
  to specify exactly which will have translations done.

  New protocols at the session/transport layer defined over IPv7 should
  have protocol numbers greater than 255, and will not be translated to
  IPv4.

  Most of the translations should consist of copying various fields,
  verifying fixed values in the datagram being translated, and setting
  fixed values in the datagram being produced.  In general, the
  checksum must be verified first, and then a new checksum computed for
  the generated datagram.

8.1  Version 4 IP address extension option

  A new option is defined for IP version 4, to carry the extended
  addresses of IPv7.  This will be particularily useful in the initial
  testing of IPv7, during a time when most of the fabric of the
  internet is IPv4.  An IPv7 host will be able to connect to another
  IPv7 host anywhere in the internet even though most of the paths and
  routers are IPv4, and still use the full addressing.  This will
  continue to work until non-unique network numbers are assigned, by
  which time most of the infrastructure should be IPv7.

8.1.1  Option format

   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  type (147)   | length = 10   |     source IPv7 AD number     |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  ...          | src 7th octet |     destination IPv7 AD       |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  number ...   | dst 7th octet |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

  The source and destination are in IPv4 order (source first), for
  consistancy.  The type code is 147.

8.2  Fragmented datagrams

  Datagrams that have been fragmented must be reassembled by the
  converting host or router before conversion.  Where the conversion is
  being done by the destination host (i.e.,  the case of a v7 host
  receiving v4 datagrams), this is similar to the present fragmentation
  model.

  When it is being done by an intermediate router (acting as an
  internetwork layer gateway) the router should use all of source,
  destination, and datagram ID for identification of IPv4 fragments;



Ullmann                                                        [Page 26]

RFC 1475                         TP/IX                         June 1993


  note that destination is used implicitly in the usual reassembly at
  the destination.  When reassembling an IPv7 datagram, the 128 bit
  fragment ID is used as usual.

  If the fragments take different paths through the net, and arrive at
  different conversion points, the datagram is lost.

8.3  Where does the conversion happen?

  The objective of conversion is to be able to upgrade systems, both
  hosts and routers, in whatever order desired by their owners.
  Organizations must be able to upgrade any given system without
  reconfiguration or modification of any other; and IPv4 hosts must be
  able to interoperate essentially forever.  (IPv4 routers will
  probably be effectively eliminated at some point, except where they
  exist in their own remote or isolated corners.)

  Each TCP/IP v7 system, whether host or router, must be able to
  recognize adjacent systems in the topology that are (only) v4, and
  call the appropriate conversion routine just before sending the
  datagram.

  Digression:  I believe v7 hosts will get much better performance by
  doing everything internally in v7, and using conversion to filter
  datagrams when necessary.  This keeps the usual code path simple,
  with only a "hook" right after receiving to convert incoming IPv4
  datagrams, and just before sending to convert to IPv4.  Routers may
  prefer to keep datagrams in their incoming version, at least until
  after the routing decision is made, and then doing the conversion
  only if necessary.  In either case, this is an implementation
  specific decision.

  It must be noted that any forwarding system may convert datagrams to
  IPv7, then back to IPv4, even if that loses information such as
  unknown options.  The reverse is not acceptable:  a system that
  receives an IPv7 datagram should not convert it to IPv4, then back to
  IPv7 on forwarding.

  The preferred method for identifying which hosts require conversion
  is to ARP first for the IPv7 address, and then again if no response
  is received, for the IPv4 address.  The reservation of ADs out of the
  v4 network number space is useful again here, protecting hosts that
  fail to properly use the ARP address length fields.

  On networks where ARP is not normally used, the method is to assume
  that a remote system is v7.  If an IPv7 datagram is received from it,
  the assumption is confirmed.  If, after a short time, no IPv7
  datagram is received, a v7 ICMP echo is sent.  If a reply is received



Ullmann                                                        [Page 27]

RFC 1475                         TP/IX                         June 1993


  (in either version) the assumption is confirmed.

  If no reply is recieved, the remote system is assumed not to
  understand IPv7, and datagrams are converted to IPv4 just before
  transmitting them.

  Implementations should also provide for explicit configuration where
  desired.

8.4  Hybrid IPv4 systems

  In the course of implementing IPv7, especially in constrained
  environments such as small terminal servers, it may be useful to
  implement the IPv4 address extension option directly, thereby
  regaining universal connectivity.

  This may also be a useful interim step for vendors not prepared to do
  a major rework of an implementation; but it is important not to get
  stalled in this step.

  A hybrid IPv4 + address extension system does not have to implement
  the conversion, it places this onus on its neighbors.  It may itself
  have an address with the subnet extension (7th byte) not equal to 1.

  The implication of hybrid systems is that it is not valid to assume
  that a host with a IPv7 address is a native IPv7 implementation.

8.5  Maximum segment size in TCP

  It is probably advisable for IPv4 implementations to reduce the MSS
  offered by a small amount where possible, to avoid fragmentation when
  datagrams are converted to IPv7.  This arises when IPv4 hosts are
  communicating through an IPv7 infrastructure, with the same MTU as
  the local networks of the hosts.

8.6  Forwarding and redirects

  It may be important for a router to not send ICMP redirects when it
  finds that it must do a conversion as part of forwarding the
  datagram.  In this case, the hosts involved may not be able to
  interact directly.  The IPv7 host could ignore the redirect, but this
  results in an unpleasant level of noise as the sequence continually
  recurs.

8.7  Design considerations

  The conversion is designed to be fairly efficient in implementation,
  especially on RISC architectures, assuming they can either do a



Ullmann                                                        [Page 28]

RFC 1475                         TP/IX                         June 1993


  conditional move (or store), or do a short forward branch without
  losing the instruction cache.  The other conditional branches in the
  body of the code are usually not-taken out to the failure/discard
  case.

  Handling options does involve a loop and a dispatch (case) operation.
  The options in IPv4 are more difficult to handle, not being designed
  for speed on a 32 bit aligned RISCish architecture, but they do not
  occur often, except perhaps the address extension option.

  For CISC machines, the same considerations will lead to fairly
  efficient code.

  The conversion code must be extremely careful to be robust when
  presented with invalid input; in particular, it may be presented with
  truncated transport layer headers when called recursively from the
  ICMP conversion.

8.8  Conversion from IPv4 to IPv7

  Individual steps in the conversion; the order is in most cases not
  significant.

     o  Verify checksum.

     o  Verify fragment offset is 0, MF flag is 0.

     o  Verify version is 4.

     o  Extend TTL to 16 bits, multiply by 16.

     o  Set forward route identifier to 0.

     o  Set first 3 octets of destination to AD (i.e., 192.0.0), copy
        first three octets from v4 address, set next octet to 1, copy
        last octet.  (This can be done with shift/mask/or operations
        on most architectures.)

     o  Do the same translation on source address.

     o  Copy protocol, set high 8 bits to zero.

     o  If DF flag set, add Don't Fragment option.

     o  If Address Extension option present, copy ADs and subnet
        extension numbers into destination and source.

     o  Convert other options where possible.  If an unknown option



Ullmann                                                        [Page 29]

RFC 1475                         TP/IX                         June 1993


        with copy-on-fragment is found, fail.  If copy-on-fragment is
        not set, ignore the option.  I.e., the flag is (ab)used as an
        indicator of whether the option is mandatory.

     o  Compute new IP header length.

     o  Convert session/transport layer (TCP) header and data.

     o  Compute new overall datagram length.

     o  Calculate IPv7 checksum.

8.9  Conversion from IPv7 to IPv4

  The steps to convert IPv7 to IPv4 follow.  Note that the converting
  router or host is partly in the role of destination host; it checks
  both bits of class in IP options, and (as in the other direction)
  must reassemble fragmented datagrams.

     o  Verify checksum.

     o  Verify version is 7

     o  Set type-of-service to 0 (there may be an option defined,
        that will be handled later).

     o  If length is greater than (about) 65563, fail.  (That number
        is not a typographical error.  Note that the IPv7+TCPv7
        headers add up to 28 bytes more than the corresponding v4
        headers in the usual case.) This check is only to avoid
        useless work, the precise check is later.

     o  Generate an ID (using an ISN based sequence generator,
        possibly also based on destination or source or both).

     o  Set flags and fragment field to 0.

     o  Divide TTL by 16, if zero, fail (send ICMP Time Exceeded).
        If greater that 255, set to 255.

     o  If next layer protocol is greater than 255, fail.  Else copy.

     o  Copy first 3 octets and 8th octet of destination to
        destination address.

     o  Same for source address.

     o  Generate v4 address extension option.  (If enabled; this



Ullmann                                                        [Page 30]

RFC 1475                         TP/IX                         June 1993


        probably should be a configuration option, should default to
        on.)

     o  Process v7 options.  If any unknown options of class not 0
        found, fail.

     o  If Don't Fragment option found, set DF flag.

     o  If Don't Convert option found, fail.

     o  Convert other options where possible, or fail.

     o  Compute new IP header length.  This may fail (too large),
        fail conversion if so.

     o  Convert session/transport layer (e.g., TCP).

     o  Compute new overall datagram length.  If greater than 65535,
        fail.

     o  Compute IPv4 checksum.

8.10  Conversion from TCPv4 to TCPv7

     o  Subtract header words from v4 checksum.  (Note that this is
        actually done with one's complement addition.)

     o  Copy flags (except for Urgent).

     o  If source port is less than 32768 (a sign condition test will
        suffice on most architectures), copy it.  If equal or
        greater, add 65536.

     o  Same operation on destination port.

     o  Copy sequence to low 32 bits, set high to 0.

     o  Copy acknowledgement to low 32 bits, set high to 0.

     o  Copy window.  (The TCPv4 performance extension [RFC1323]
        window-scale cannot be used, as it would require state; we
        use the basic window offered.)

     o  Add 32 bit rollover option.

     o  Convert maximum segment size option if present.

     o  Compute data offset and copy data.



Ullmann                                                        [Page 31]

RFC 1475                         TP/IX                         June 1993


     o  Add header words into saved checksum.  It is important not to
        recompute the checksum over the data; it must remain an
        end-to-end checksum.

     o  Return to IP layer conversion.

8.11  Conversion from TCPv7 to TCPv4

     o  Subtract header from v7 checksum.

     o  If source port is greater than 65535, subtract 65536.  If
        result is still greater than 65535, fail.  (Send ICMP
        conversion failed/port conversion out of range.  The sending
        host may then reset its port number generator to 98304.)

     o  Same translation for destination port.

     o  Copy low 32 bits of sequence number.

     o  If A bit set, copy low 32 bits of acknowledgement.

     o  Copy flags.

     o  If window is greater than 61440, set it to 24576.  If less,
        copy it unchanged.  (Rationale for the 24K figure:  this has
        been found to be a good default for IPv4 hosts.  If the IPv7
        host is offering a very large window, the IPv4 host probably
        isn't prepared to play at that level.)

     o  Process options.  If 32 Bit Rollover is not present, and A
        flag is set, fail.  (Send ICMP conversion failed/32 bit
        Rollover missing.)

     o  If Urgent is present, compute offset.  If in segment, set U
        flag and offset field.  If not, ignore.

     o  Convert Maximum Segment Size option.  If greater than 16384,
        set to 16384.

     o  Compute new data offset.

     o  Add header words into v4 checksum.

     o  Return to IP layer conversion.







Ullmann                                                        [Page 32]

RFC 1475                         TP/IX                         June 1993


8.12  ICMP conversion

  ICMP messages are converted by copying the type and code into the new
  packet, and copying the other type-specific fields directly.

  If the message contains an encapsulated, and usually truncated, IP
  datagram, the conversion routine is called recursively to translate
  it as far as possible.  There are some special considerations:

     o  The encapsulated datagram is less likely to be valid, given
        that it did generate an error of some kind.

     o  The conversion should attempt to complete all fields
        available, even if some would cause failures in the general
        case.  Note, in particular, that in the course of converting
        a datagram, when a failure occurs, an ICMP message
        (conversion failed) is sent; this message itself may
        immediately require conversion.  Part of that conversion will
        involve converting the original datagram.

     o  Conditions such as overall datagram length too large are not
        checked.

     o  The AD and subnet byte assumed in the nested conversion may
        not be sensible if the IPv4 address extension option is not
        present and the datagram has strayed from the expected AD.
        (Not unlikely, given that we know a priori that some error
        occured.)

     o  The conversion must be very sure not to make another
        recursive call if the nested datagram is an ICMP message.
        (This should not occur, but obviously may.)

     o  It is probably impossible to generate a correct transport
        layer checksum in the nested datagram.  The conversion may
        prefer to just zero the checksum field.  Likewise, validating
        the original checksum is pointless.

  It may be best in a given implementation to have a separate code path
  for the nested conversion, that handles these issues out of the
  optimized usual path.

9.  Postscript

  The present version of TCP/IP has been a success partly by accident,
  for reasons that weren't really designed in.  Perhaps the most
  significant is the low level of network integration required to make
  it work.



Ullmann                                                        [Page 33]

RFC 1475                         TP/IX                         June 1993


  We must be careful to retain the successful ingredients, even where
  we may be unaware of them.  Tread lightly, and use all that we have
  learned, especially about not changing things that work.

  This document has described a fairly conservative step forward, with
  clear extensibility for future developments, but without jumping into
  the abyss.

10.  References

  [RFC768]    Postel, J., "User Datagram Protocol", STD 6, RFC 768,
              USC/Information Sciences Institute, August 1980.

  [RFC791]    Postel, J., "Internet Protocol - DARPA Internet Program
              Protocol Specification", STD 5, RFC 791, DARPA,
              September 1981.

  [RFC792]    Postel, J., "Internet Control Message Protocol -
              DARPA Internet Program Protocol Specification"
              STD 5, RFC 792, USC/Information Sciences Institute,
              September 1981.

  [RFC793]    Postel, J., "Transmission Control Protocol - DARPA
              Internet Program Protocol Specification", STD 7, RFC 793,
              USC/Information Sciences Institute, September 1981.

  [RFC801]    Postel, J., "NCP/TCP Transition Plan", USC/Information
              Sciences Institute, November 1981.

  [RFC1287]   Clark, D., Chapin, L., Cerf, V., Braden, R., and
              R. Hobby, "Towards the Future Internet Architecture", RFC
              1287, MIT, BBN, CNRI, ISI, UCDavis, December 1991.

  [RFC1323]   Jacobson, V., Braden, R, and D. Borman, "TCP Extensions
              for High Performance", RFC 1323, LBL, USC/Information
              Sciences Institute, Cray Research, May 1992.

  [RFC1335]   Wang, Z., and J. Crowcroft, A Two-Tier Address Structure
              for the Internet: A Solution to the Problem of Address
              Space Exhaustion", RFC 1335, University College London,
              May 1992.

  [RFC1338]   Fuller, V., Li, T., Yu, J., and K. Varadhan,
              "Supernetting: an Address Assignment and Aggregation
              Strategy", RFC 1338, BARRNet, cicso, Merit, OARnet,
              June 1992.





Ullmann                                                        [Page 34]

RFC 1475                         TP/IX                         June 1993


  [RFC1347]   Callon, R., "TCP and UDP with Bigger Addresses (TUBA),
              A Simple Proposal for Internet Addressing and Routing",
              RFC 1347, DEC, June 1992.

  [RFC1476]   Ullmann, R., "RAP: Internet Route Access Protocol",
              RFC 1476, Process Software Corporation, June 1993.

  [RFC1379]   Braden, R., "Extending TCP for Transactions -- Concepts",
              RFC 1379, USC/Information Sciences Institute,
              November 1992.

11.  Security Considerations

  Security issues are not discussed in this memo.

12.  Author's Address

  Robert Ullmann
  Process Software Corporation
  959 Concord Street
  Framingham, MA 01701
  USA

  Phone: +1 508 879 6994 x226
  Email: [email protected]


























Ullmann                                                        [Page 35]