RFC 914


Network Working Group                                    David J. Farber
Request for Comments: 914                                   Gary S. Delp
                                                         Thomas M. Conte
                                                  University of Delaware
                                                          September 1984

                          A Thinwire Protocol
                   for connecting personal computers
                            to the INTERNET

Status of this Memo



   This RFC focuses discussion on the particular problems in the
   ARPA-Internet of low speed network interconnection with personal
   computers, and possible methods of solution.  None of the proposed
   solutions in this document are intended as standards for the
   ARPA-Internet.  Rather, it is hoped that a general consensus will
   emerge as to the appropriate solution to the problems, leading
   eventually to the adoption of standards.  Distribution of this memo
   unlimited.

What is the Problem Anyway ?

   As we connect workstations and personal computers to the INTERNET,
   many of the cost/speed communication tradeoffs change.  This has made
   us reconsider the way we juggle the protocol and hardware design
   tradeoffs.  With substantial computing power available in the $3--10K
   range, it is feasible to locate computers at their point of use,
   including in buildings, in our homes, and other places remote from
   the existing high speed connections.  Dedicated 56k baud lines are
   costly, have limited availability, and long lead time for
   installation.  High speed LAN's are not an applicable interconnection
   solution.  These two facts ensure that readily available 1200 / 2400
   baud phone modems over dialed or leased telephone lines will be an
   important part of the interconnection scheme in the near future.
   This paper will consider some of the problems and possibilities
   involved with using a "thin" (less than 9600 baud) data path.  A trio
   of "THINWIRE"  protocols for connecting a personal computer to the
   INTERNET are presented for discussion.

   Although the cost and flexibility of telephone modems is very
   attractive, their low speed produces some major problems.  As an
   example, a minimum TCP/IP Telnet packet (one character) is 41 bytes
   long.  At 1200 baud, the transmission time for such a packet would be
   around 0.3 seconds.  This is equivalent to using a 30 baud line for
   single character transmission.  (Throughout the paper, the assumption
   is made that the transmission speed is limited only by the speed of
   the communication line.  We also assume that the line will act as a
   synchronous link when calculating speed.  In reality, with interrupt,
   computational, and framing overhead, the times could be 10-50%
   worse.)

   In many cases, local echo and line editing can allow acceptable


Farber & Delp & Conte                                           [Page 1]



RFC 914                                                   September 1984
Thinwire Protocol


   Telnet behavior, but many applications will work only with character
   at a time transmission.  In addition, multiple data streams can be
   very useful for fully taking advantage of the personal
   computer/Internet link.  Thus this proposal.

   There are several forms that a solution to this problem can take.
   Three of these are listed below, followed by descriptions of possible
   solutions of each form.

   o    As a non-solution, one can learn to live with the slow
        communication (possibly a reasonable thing to do for background
        file transfer and one-time inquiries to time, date, or
        quote-of-the-day servers).

   o    Using TCP/IP, one can intercept the link level transmissions,
        and try various kinds of compression algorithms.  This provides
        for a symmetrical structure on either side of the "Thinwire".

   o    One could build an "asymmetrical" gateway which takes some of
        the transport and network communication overhead away from both
        the serial link and the personal computer.  The object would be
        to make the PC do the local work, and to make the
        interconnection with the extended network a benefit to the PC
        and not a drain on the facilities of the PC.

   The first form has the advantage of simplicity and ease of
   implementation. The disadvantages have been discussed above.  The
   second form, compression at link level, can be exploited in two ways.

      Thinwire I is a simple robust compressor, which will reduce the 41
      byte minimum TCP/IP Telnet packets to a series of 17 byte update
      packets.  This would improve the effective baud rate from 30 baud
      to 70 baud over a 1200 baud line (for single character packets).

      Thinwire II uses a considerably more complex technique, and takes
      advantage of the storage and processing power on either side of
      the thinwire link.  Thinwire II will compress packets from
      multiple TCP/IP connections from 41 bytes down to 13 bytes.  The
      increased communication rate is 95 (effective) baud for single
      character packets.

   The third form balances the characteristics of the personal computer,
   the communications line, the gateway, and the Internet protocols to
   optimize the utility of the communications and the workstation
   itself.  Instead of running full transport and internet layers on the
   PC, the PC and the gateway manage a single reliable stream,
   multiplexing data on this stream with control requests.  Without the
   interneting and flow control structures traveling over the
   communications line on a per/packet basis, the data flow can be


Farber & Delp & Conte                                           [Page 2]



RFC 914                                                   September 1984
Thinwire Protocol


   compressed a great deal.  As there is some switching overhead, and a
   reliable link level protocol is needed on the serial line, the
   average effective baud rate would be in the 900 baud range.

   Each of these Thinwire possibilities will be explored in detail.

Thinwire I

   The simplest technique for the compression of packets which have
   similar headers is for both the transmitting and receiving host to
   store the most recent packet and transmit just the changes from one
   packet to the next.  The updated information is transmitted by
   sending a packet including the updated information along with a
   description of where the information should be placed.  A series of
   descriptor-data blocks would make up the update packet.  The
   descriptor consists of the offset from the last byte changed to the
   start of the data to be changed and a count of the number of data
   bytes to be substituted into the old template.  The descriptor is one
   byte long, with two four bit fields; offsets and counts of up to 15
   bytes can be described. In the most pathological case the descriptor
   adds an extra byte for every 15 bytes (or a 6% expansion).

   An example of Thinwire I in action is shown in Appendix A.  A
   sequence of two single character TCP/IP Telnet packets is shown.  The
   "update" packet which would actually be transmitted is shown
   following them.  Each Telnet packet is 41 bytes long; the typical
   update is 17 bytes.  This technique is a useful improvement over
   sending entire packets.  It is also computationally simple.  It
   suffers from two problems: the compression is modest, and, if there
   is more than one class of packets being handled, the assumption of
   common header information breaks down, causing the compression of
   each class to suffer.

Thinwire II

   Both of the problems described above suggest that a more
   computationally complex protocol may be appropriate.  Any major
   improvement in data compression must depend on knowledge of the
   protocols being used.  Thinwire II uses this knowledge to accomplish
   two things.  First, the packets are sorted into classes.  The packets
   from each TCP connection using the thinwire link, would, because of
   their header similarities, make up a class of packets.  Recognizing
   these classes and sorting by them is called "matching templates".
   Second, knowledge of the protocols is used to compress the updates.
   A bitfield indicating which fields in the header have changed,
   followed only by the changed fields, is much shorter than the general
   form of change notices.  Simple arithmetic is allowed, so 32 bit




Farber & Delp & Conte                                           [Page 3]



RFC 914                                                   September 1984
Thinwire Protocol


   fields can often be updated in 8 or 16 bits.  By using the sorting,
   protocol-specific updating, Thinwire II provides significant
   compression.

   A typical transaction is described in Appendix B.  The "template
   matching" is based on the unchanging fields in each class of packet.
   A TCP/IP packet would match on the following fields: network type
   field(IP), version, type of service, protocol(TCP), and source and
   destination address and port.  Note that the 41 bytes have been
   reduced to 13 bytes.  An additional advantage is that  multiple
   classes of packets can be transported across the same line without
   affecting the compression of each other, just by matching and storing
   multiple templates.

   Some of the implications of this system are:

      o    The necessity of saving several templates (one for each
           TCP/IP connection ) means that there will be a relatively
           large memory requirement.  This requirement for current
           personal computers is reasonable.  In addition, the gateway
           must keep tables for several connections at a time.

      o    The Thinwire links are slow (that's why we call them thin);
           much slower than normal disk access.  There is no reason that
           inactive templates cannot be swapped out to disk and
           retrieved when needed if memory is limited.  (Note that as
           memory density increases, this is less and less of a
           problem.)

      o    There is state information in the connections.  If the two
           sides get out of synchronization with each other, data flow
           stops.  This means that some method of error detection and
           recovery must be provided.

      o    To minimize the problem described above, the protocol used on
           the serial line must be reliable.  See Appendix D for details
           of SLIP, Serial Line Interface Protocol, as an example of
           such a protocol.  There must also be periodic
           resynchronization.  (For example, every Nth packet would be
           transmitted in full).

      o    The asynchronous link is not, by its nature, a packet
           oriented system; a packet structure will need to be layered
           on the character at a time transfer.  However, if the
           protocol layer below thinwire (SLIP) can be trusted, the
           formation of packets is a simple matter.

      o    Thinwire II will need to be enhanced for each new protocol



Farber & Delp & Conte                                           [Page 4]



RFC 914                                                   September 1984
Thinwire Protocol


           (TCP, UDP, TP4) it is called upon to service.  Any packet
           type not recognized by the Thinwire connection will be
           transmitted in full.

   For maintaining full network service, Thinwire II or a close variant
   seems to be the solution.

Thinwire III

   When transmissions at the local network (link) level are not
   required, if only the available services are desired, then a solution
   based on Thinwire III may be appropriate.  A star network with a
   gateway in the center serving as the connection between a number of
   Personal Computers and the Internet is the key of Thinwire III.
   Rather than providing connections at the network/link level, Thinwire
   III assumes that there is a reliable serial link (SLIP or equivalent)
   beneath it and that the workstation/personal computer has better
   things to do than manage TCP state tables, timeouts, etc.  It also
   assumes that the gateway supporting the Thinwire III connections is
   powerful enough to run many TCP connections and several SLIP's at the
   same time.  The gateway fills in for the limitations of the
   communications line and the personal computer.  It provides a gateway
   to the INTERNET, managing the transport and network functions,
   providing both reliable stream and datagram service.

   In Thinwire III, the gateway starts an interpreter for each SLIP
   connection from a personal computer.  The gateway will open TCP, UDP,
   and later TP4 connections on the request of the personal computer.
   Acting as the agent for the personal computer, it will manage the
   remote negotiations and the data flow to and from the personal
   computer.  Multiple connections can be opened, with inline logical
   switches in the reliable data flow indicating which connection the
   data is destined for.  Additional escaped sequences will send error
   and informational data between the two Thinwire III communicators.

   This protocol is not symmetric.  The gateway will open connections to
   the INTERNET world as an agent for the personal computer, but the
   gateway will not be able to open inbound connections to the personal
   computer, as the personal computer is perceived as a stub host.  The
   personal computer may however passively open connections on the
   gateway to act as a server.  Extended control sequences are specified
   to handle the multiple connection negotiation that this server
   ability will entail.

   This protocol seems to ignore the problem of flow control. Our
   thought is that the processing on either side of the communication
   link will be much speedier than the link itself.  The buffering for
   the communication line and the user process blocking for this will



Farber & Delp & Conte                                           [Page 5]



RFC 914                                                   September 1984
Thinwire Protocol


   provide most of the flow control.  For the rare instances that this
   is not sufficient, there are control messages to delay the flow to a
   port or all data flow.

   A tentative specification for Thinwire III is attached as Appendix C.

The authors acknowledge the shoulders upon which they stand, and
apologize for the toes they step on.  Ongoing work is being done by Eric
Thayer, Guru Parulkar, and John Jaggers.  Special thanks are extended to
Peter vonGlahn, Jon Postel and Helen Delp for their helpful comments on
earlier drafts.  Responses will be greatly appreciated at the following
addresses:

   Dave Farber <Farber@udel-ee>
   Gary Delp <Delp@udel-ee>
   Tom Conte <Conte@udel-ee>



































Farber & Delp & Conte                                           [Page 6]



RFC 914                                                   September 1984
Thinwire Protocol


Appendix A -- Example of Thinwire I Compression

   Here is an example of how Thinwire I would operate in a common
   situation.  The connection is a TCP/IP Telnet connection.  The first
   TCP/IP Telnet packet is on the next page; it simulates the typing of
   the character "a".  The second packet would be produced by typing
   "d"; it is shown on the following page.  The compressed version is on
   the third page following.

   [NOTE: The checksums pictured have not been calculated.  Binary in
   MSB to LSB format]








































Farber & Delp & Conte                                           [Page 7]



RFC 914                                                   September 1984
Thinwire Protocol


        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
IP     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
header:|Version|  IHL  |Type of Service|          Total Length         |
       |0 1 0 0|0 1 0 1|0 0 0 0 0 0 0 0|0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0|
P      |   4   |   5   |       0       |               41              |
a      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
c      |       Identification          |Flags|      Fragment Offset    |
k      |0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1|0 0 0|0 0 0 0 0 0 0 0 0 0 0 0 0|
e      |                1              |  0  |            0            |
t      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-      |  Time to live |   Protocol    |       Header Checksum         |
1      |0 1 1 0 0 1 0 1|0 0 0 0 0 1 1 0|0 1 1 1 0 1 1 1 0 0 0 1 0 1 0 0|
       |      101      |       6       |             nnn               |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                     Source Address                            |
       |1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 1 0 0 1 1 1 0 0 0 1 0 1 0 0|
       |    192.       |       5.      |     39.       |      20       |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                   Destination Address                         |
       |0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0|
       |     10.       |       2.      |      0.       |      52       |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
        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
TCP    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
header:|         Source Port           |       Destination Port        |
       |0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1|0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1|
       |             1025              |               27              |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                        Sequence Number                        |
       |0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 1 1 0 0|
       |                              300                              |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                     Acknowledgement Number                    |
       |0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 1 0 0|
       |                              100                              |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |offset | Reserved  |U A P R S F|            Window             |
       |0 1 0 1|0 0 0 0 0 0|0 1 0 0 0 0|0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0|
       |   5   |     0     |     16    |             512               |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |           Checksum            |         Urgent Pointer        |
       |0 0 0 0 0 1 0 0 1 0 1 1 0 0 0 0|0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0|
       |             nnn               |               0               |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                            Data                               |
       |0 1 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0|
       |        "a"                                                    |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+



Farber & Delp & Conte                                           [Page 8]



RFC 914                                                   September 1984
Thinwire Protocol


        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
IP     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
header:|Version|  IHL  |Type of Service|          Total Length         |
       |0 1 0 0|0 1 0 1|0 0 0 0 0 0 0 0|0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0|
       |   4   |   5   |       0       |               41              |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
P      |       Identification*         |Flags|      Fragment Offset    |
a      |0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0|0 0 0|0 0 0 0 0 0 0 0 0 0 0 0 0|
c      |                2              |  0  |            0            |
k      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
e      |  Time to live*|   Protocol    |       Header Checksum*        |
t      |0 1 1 0 0 1 1 0|0 0 0 0 0 1 1 0|0 1 1 1 0 1 1 1 0 0 0 1 0 1 0 0|
-      |      102      |       6       |             nnn               |
2      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                     Source Address                            |
       |1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 1 0 0 1 1 1 0 0 0 1 0 1 0 0|
       |    192.       |       5.      |     39.       |      20       |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                   Destination Address                         |
       |0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0|
       |     10.       |       2.      |      0.       |      52       |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
        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
TCP    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
header:|         Source Port           |       Destination Port        |
       |0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1|0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1|
       |             1025              |               27              |
* 's   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
show   |                        Sequence Number*                       |
changed|0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 1 1 0 1|
fields |                              301                              |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                     Acknowledgement Number*                   |
       |0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 1 0 1|
       |                              101                              |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |offset | Reserved  |U A P R S F|            Window             |
       |0 1 0 1|0 0 0 0 0 0|0 1 0 0 0 0|0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0|
       |   5   |     0     |     16    |             512               |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |           Checksum*           |         Urgent Pointer        |
       |0 0 0 0 0 1 0 0 1 0 1 1 0 0 0 0|0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0|
       |             nnn               |               0               |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                            Data*                              |
       |0 1 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0|
       |        "d"                                                    |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+



Farber & Delp & Conte                                           [Page 9]



RFC 914                                                   September 1984
Thinwire Protocol


   The Thinwire Driver finds the template (which is the previous packet
   sent), compares the template to the packet and creates a change
   message (field names of change record data have been added for
   comparison):

       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
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |Descriptor byte|   Data:       |Descriptor byte|  Data:        |
      |offset |length | Identification|offset |length |  Time to live |
      |0 0 1 0|0 0 0 1|0 0 0 0 0 0 1 0|0 0 1 0|0 0 0 1|0 1 1 1 0 1 1 0|
      |   2   |   1   |      2        |   2   |   1   |     102       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |Descriptor byte|   Data:                       |Descriptor byte|
      | offset| length|         Header Checksum       |offset |length |
      |0 0 1 0|0 0 1 0|1 1 1 1 0 0 1 0 1 0 1 1 0 1 0 0|1 1 1 1|0 0 1 0|
      |    2  |   2   |              nn               |  15   |   2   |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |   Data:       |Descriptor byte|   Data:       |Descriptor byte|
      |   Seq Number  |offset |length |   Ack Number  |offset |length |
      |0 0 1 0 1 1 0 1|0 0 1 1|0 0 0 1|0 1 1 0 0 1 0 1|0 1 1 1|0 0 1 0|
      |      301      |   3   |   1   |      101      |   7   |   2   |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |   Data:                       |Descriptor byte|   Data:       |
      |       -- TCP Checksum --      |offset |length |     data      |
      |0 0 0 0 0 1 0 0 1 0 1 1 0 0 0 0|0 0 1 0|0 0 0 1|0 1 1 0 0 1 0 0|
      |             nn                |   2   |   1   |     "d"       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |Descriptor byte|
      |offset |length |
      |0 0 0 0|0 0 0 0|  the 0 0 offset/length record ends the update.
      |   0   |   0   |
      +-+-+-+-+-+-+-+-+

   Thinwire I then sends this message over the line where the previous
   packet is updated to form the new packet.  Note: One can see that a
   series of null descriptor bytes will reset the connection.















Farber & Delp & Conte                                          [Page 10]



RFC 914                                                   September 1984
Thinwire Protocol


Appendix B -- Examples of Thinwire II Compression

   This Appendix provides an example of how the Thinwire II would
   operate in a common situation.  The same original packets are used as
   in Appendix A, so only the updates are shown.

   As the later field definitions depend on the contents of earlier
   fields, a field by field analysis of the update packets will be
   useful.

                    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3
                   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      Thinwire II  |U|L|Template no| Len of change | Type of Packet|
       minimum     |0|0|0 0 0 1 0 1|0 0 0 1 1 0 0 1|0 0 0 0 0 0 0 1|
       header:     |N N|     5     |          41   |     TCP/IP    |
                   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

      The first bit is the UPDATE bit. If it is a 0 this packet
      describes a new template, and the entire new packet is included,
      following the header.  If there was a previous template with the
      same number, it will be cleared and replaced by the new template.
      If the UPDATE bit is a 1, then this packet should be used to
      update the template with the number given in the template number
      field.

      The second bit is the LONG bit. If it is a 1 it indicates a LONG
      packet.  This means that the update length field will be 16 bits
      instead of 8 bits.

      The remaining 6 bits in the first byte indicate the template
      number that this packet is an update to.

      The template number is followed by 1 or 2 bytes (depending on the
      value of the LONG bit) which give the length of the packet. This
      is the number of data bytes following the variable length header.

      If the UPDATE bit is 0 on this packet, the next byte will be a
      flag telling what type of packet the sender thinks this packet is.
      The flag will be saved by the receiver to interpret the update
      packets.  Type 0 is for unknown types. If the type 0 flag is set,
      there will be no updates to this template number.  Type 1 is
      TCP/IP; the method of updating will be described below.  Type 2 is
      UDP/IP; the method of update is not described at this time.

   At this time we have enough information to encode packet 1 of the
   example. Assuming for the moment that this is the first packet for
   this connection, the UPDATE bit would be set to 0.  As the packet has
   a length of 41 and so can be described in 8 bits, the LONG bit would
   be set to 0.  A template number not in use (or the oldest in use


Farber & Delp & Conte                                          [Page 11]



RFC 914                                                   September 1984
Thinwire Protocol


   template number) would be assigned to this packet.  The number 5 is
   illustrated.  The Length of Packet would be given as 41, and the Type
   Flag set to TCP/IP (1).  The 41 bytes of the packet would follow.

   The transmission of packet 2 requires the specification of Type 1
   (TCP/IP) updating.  There are portions of the packets which will
   always be the same; these are described in the body of the paper, and
   are used to match the template.  These do not need to be transmitted
   for an update.  There are portions of the packet which will always
   (well almost always) change.  These are the IP Header checksum, the
   IP Identification number, and the TCP checksum.  These are
   transmitted, in that order, with each template update immediately
   after the packet length byte/bytes.  Following the invariant portion
   of the header are updates to the fields which change some of the
   time.  Which fields are different is indicated with a bitfield
   describing the changes.

   The Bitfield is used to indicate which fields (of those that may stay
   the same) have changed.  The technique for updating the field varies
   with the field description.  The specifications for TCP/IP are shown
   in Table B-1.

           0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Thin-  |U|L|Template no| Len of change | Type of Packet|
wire II|0|0|0 0 0 1 0 1|0 0 0 1 1 0 0 1|0 0 0 0 0 0 0 1|
header:|N N|     5     |          41   |     TCP/IP    |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
        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
IP     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
header:|Version|  IHL  |Type of Service|          Total Length         |
       |0 1 0 0|0 1 0 1|0 0 0 0 0 0 0 0|0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0|
P      |   4   |   5   |       0       |               44              |
a      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
c      |       Identification          |Flags|      Fragment Offset    |
k      |0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1|0 0 0|0 0 0 0 0 0 0 0 0 0 0 0 0|
e      |                1              |  0  |            0            |
t      +~+~+~+~+~.~+~+~+~+~+~+~+~+~+~+.+~+~+~+~+~+~+~+~+~+~+~.~+~+~+~+~+
-                .                    .                      .
1                .                    .                      .
       +~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+~+
       |           Checksum            |         Urgent Pointer        |
       |0 0 0 0 0 1 0 0 1 0 1 1 0 0 0 0|0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0|
       |             nnn               |               0               |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |    Data       |
       |0 1 1 0 0 0 0 1|
       |        "a"    |
       +-+-+-+-+-+-+-+-+


Farber & Delp & Conte                                          [Page 12]



RFC 914                                                   September 1984
Thinwire Protocol


   The changed field update information is added to the update header in
   the order that the bits appear in the field.  That is, if both the IP
   packet length bit and the Time to Live  bit are set, the 2 new bytes
   of the IP Packet length will precede the 1 new byte of the Time to
   Live field.

   The update for packet 2 is shown below. Note that this is an update
   to template 5, the length of update is 8 bits with a value of 1.  The
   new checksums and IP Identification Number are included, and the
   flags are set to indicate changes to the following fields: Time to
   Live, Add 8 bits to Sequence and Acknowledgement Numbers.  The new
   data is one byte following the header.

   Thinwire II would send this message over the line where it would be
   reassembled into the correct packet.

   Note: For purposes of synchronization, if three 0 length, template 0,
   type 0 packets are received, the next non-zero byte should be treated
   as a start of packet, and the template tables cleared.

  ____________________________________________________________________

     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
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |U|L|Template no| Len of change |   IP  Header  Checksum        |
   |1|0|0 0 0 1 0 1|0 0 0 0 0 0 0 1|0 1 1 1 0 1 1 1 0 0 0 1 0 1 0 0|
   |Y|N|     5     |       1       |           nnn                 |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |   IP Identification number    |      TCP  Checksum            |
   |0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0|0 0 0 0 0 1 0 0 1 0 1 1 0 0 0 0|
   |           2                   |           nnn                 |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  Bitfield     |  Time to Live |add to Seq no. | add to Ack Num|
   |0 0 1 0 1 0 0 0|0 0 0 0 0 0 0 1|0 0 0 0 0 0 0 1|0 0 0 0 0 0 0 1|
   |    T Ad8      |       1       |        1      |      1        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |    data       |
   |0 0 0 1 0 1 1 1|                                                
   |      "d"      |                                                
   +-+-+-+-+-+-+-+-+                                                

                      Packet 2. Thinwire II update

  ____________________________________________________________________







Farber & Delp & Conte                                          [Page 13]



RFC 914                                                   September 1984
Thinwire Protocol


Appendix C -- Tentative Specification for Thinwire III

   Thinwire III, as stated in the body of this paper, provides multiple
   virtual connections over a single physical connection.  As Thinwire
   III is based on a single point to point connection, much of the
   per/datagram information (routing and sequencing) of other transport
   systems can be eliminated.  In the steady state any bytes received by
   thinwire III are sent to the default higher level protocol
   connection.  There are escaped control sequences which allow the
   creation of additional connections, the switching of the default
   connection, the packetizing of datagrams, and the passing of
   information between the gateway and the personal computer.  The
   gateway and the personal computer manage a single full duplex stream,
   multiplexing control requests and streams of data through the use of
   embedded logical switches.

   The ascii character "z" (binary 01011011 ) is used as the escape
   character.  The byte following the "z" is interpreted to determine
   the command.  Table C-1 shows the general classes the  bytes (Request
   codes) can fall into.

   In order to transmit the character "z", two "z"'s are transmitted.
   The first is interpreted as an escape, the second as the lower case
   letter "z" to be transmitted to the default connection.  The letter z
   was chosen as the escape for its low occurrence in text and control
   data streams, because it should pass easily through any lower level
   protocols, and for its generally innocuous behavior.

   Descriptions of specifications of each of the Request codes are
   below.

   Starting with the range 0-31; these Request codes change the default
   connection. After a connection has been established, any characters
   which come across the line that are not part of a Request code
   sequence are transmitted to one of the connections.  To begin with
   this connection defaults to Zero, but when the "Switch Default
   Connection" command is received, characters are sent to the
   connection named in the request until a new request is received.
   Zero is a special diagnostic connection; anything received on
   connection number Zero should be echoed back to the sender on
   connection number One.  Anything received on connection number One
   should be placed on the diagnostic output of the receiving host.  Any
   other connection number indicates data which should be sent out the
   numbered connection.  If the numbered connection has not been opened,
   the data can be thrown away, and an Error Control Message returned to
   the sender.

   Escapes followed by numbers 32 through 255 are for new connections,
   requests for information, and error messages.  The escape will be


Farber & Delp & Conte                                          [Page 14]



RFC 914                                                   September 1984
Thinwire Protocol


   followed by a Request code, a one byte Request Sequence Number (so
   that the Reply to Request can be asynchronously associated with the
   Request), and the arguments for the specific request.  (The length of
   the argument field will be determined by the Request code.)  The
   format of the request will be as pictured below:

      "z" <Request Code> <Request Sequence Number> [ <Arguments> ... ]

   At this time the Request codes 32-63 are reserved.

   The Request codes 64-127 are for stream server open requests.  For
   the purposes of compression, many of the common servers are assigned
   single byte codes.  See Table C-2.

   Request code 68 is to a connection to the default hostname server
   used by the gateway.  It takes 3 bytes for this request. It has the
   form:

      "z" < 68 > < Request Sequence Number >

   Request code 95 is to open any specified TCP Port at the specified
   address.  It takes 9 bytes for this request.  It has the form:

      "z" < 95 > < Request Sequence Number > < 4 bytes of IP address> <
      2 bytes of TCP Port >

   Request codes 96-127  are RESERVED for alternate transport protocols.

   The Request codes 128-191 are used for framing Datagrams and opening
   new Datagram connections.  The code 128 is the Start of Datagram
   code.  The format is:

      "z" <128> <Length of Datagram (2 bytes)> <Socket> Data ...

   As with the Stream opens, there are a number of assigned ports with
   codes for them.  They are listed in Table C-3.

   The Request Codes 192-254 are control, status and informational
   requests.  These are still under development, but will include:

      -flow control
      -get host/server/protocol by entry/name/number.
      -additional error messages
      -overall reset
      -open passive connection

   The Request Code 252 is the request to close a connection.  This
   Code, followed by the connection number, indicates that no more data



Farber & Delp & Conte                                          [Page 15]



RFC 914                                                   September 1984
Thinwire Protocol


   will be sent out this connection number.  A second request with the
   same connection number will indicate that no more data will be
   accepted on this connection.

   The Request Code 253 is the information request for a connection. The
   protocol, status, and port number of the connection should be
   returned. The format of this reply has yet to be specified.

   The Request code 254 is an error notification.  These are to be
   acknowledged with their Request Sequence Numbers.  Error codes are
   under development.

   The Request code 255 is the Reply to Request. The Request Sequence
   Number identifies the request being replied to.  The format is:

      "z" <255> <Request Sequence Number (in reply to)> <Length of reply
      (1 byte)> Reply...

   The Thinwire Drivers on each side will wait at their inbound sockets,
   and relay across the thinwire link
   character-by-character/packet-by-packet for the stream/datagram
   connections.

   Thinwire III is labeled as a tentative specification, because at this
   time, in order to publish this RFC in a timely fashion, several minor
   issues are still unresolved.  An example is the scheduling of serial
   line use. Short messages could be given priority over long packets,
   or priority schemes could be changed during the session, depending
   upon the interactive desire of the user.  Addition issues will be
   resolved in the future.





















Farber & Delp & Conte                                          [Page 16]



RFC 914                                                   September 1984
Thinwire Protocol


Appendix D -- Serial Line Interface Protocol (SLIP)

   Initial Specifications and Implementation Suggestions

   PHILOSOPHY

      The world is a dangerous place for bits.  Data transmission can be
      an time consuming business when one has to make sure that bits
      don't get lost, destroyed, or forgotten.  To reduce such problems,
      the Serial Line Interface Protocol (SLIP) maintains an attitude
      toward the world that includes both a mistrust of serial lines and
      a margin of laziness.  Examples of this approach include how SLIP
      recovers from errors and how SLIP handles the problem of
      resequencing (see PROTOCOL SPECIFICATIONS and IMPLEMENTATION
      SUGGESTIONS).

   THE MESSAGE FORMAT

      Both the Sender Task and the Receiver Task communicate using a
      standard message format and the Sender and Receiver Task of one
      machine's SLIP communicate using a shared buffer.  The message
      begins with a 1 byte Start of Header token (StH, 11111111) and is
      followed by a sequence number of four bits (SEQ) and an
      acknowledgement number of four bits (ACK).  Following the StH, SEQ
      and ACK, is a 5 bit length field which specifies the length of the
      data contained in the message. Following the length is a three bit
      field of flags.  The first bit is used to indicate that the a
      receive error has occurred, and the ACK is actually a repeat of
      the Last Acknowledged message (a LACK).  The second bit is used to
      indicate a Synchronize Sequence Numbers message (SSNM), and the
      third bit is used to indicate a Start of Control Message (SOCM);
      all three of these flags are explained below. Finally, at the end
      of the message is an exclusive-or checksum.  The message format is
      shown in figure D-1.

            ________________________________________________

 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+...
|     StH       |  SEQ  |  ACK  |  Length |Flags|...Data...
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+...
The maximum data length is 32 bytes.                0 1 2 3 4 5 6 7
This limits the vulnerability of receiver       ...-+-+-+-+-+-+-+-+-+-+
timeout errors occurring because of bit error .Data...|   Checksum    |
in the length field.                            ...-+-+-+-+-+-+-+-+-+-+

                    Figure D-1. SLIP Message Format

            ________________________________________________


Farber & Delp & Conte                                          [Page 17]



RFC 914                                                   September 1984
Thinwire Protocol


      The Sender, when idle but needing to acknowledge, will send out
      short messages of the same format as a regular message but with
      the SOCM flag set and the data field omitted.  ( This short
      message is called a SOCM, and is used instead of a zero length
      message to avoid the problem of continually ACK'ing ACK's ). The
      Sender Task, when originating a connection (see STARTING UP AND
      FINISHING OFF COMMUNICATIONS), will send out another short message
      but with the SSNM flag set and the data omitted.  This message (a
      SSNM) used for a TCP-style 3 way startup handshake.

   PROTOCOL SPECIFICATIONS and SUGGESTIONS

      The SLIP module, when called with data to send, prepends its
      header (SEE ABOVE) to the data, calculates a checksum and appends
      the checksum at the end.  (This creates a message.)  The message
      has a sequence number associated with it which represents the
      position of the message in the Sender SLIP's buffers.  The
      sequence number for the message can range from 0 to 15 and is
      returned in the ACK field of the other machine's Sender SLIP
      messages to acknowledge receipt.

      There are two scenarios for transmission.  In the first, both
      SLIP's will be transmitting to each other.  To send an
      acknowledgement, the Receiver SLIP uses the ACK field in its next
      outgoing message. To receive an acknowledgement, the Sender checks
      the ACK field of its Receiver's incoming messages.  In the second
      scenario, one SLIP may have no data to transmit for a long time.
      Then, as stated above, to acknowledge a received message, the
      Receiver has its Sender send out a short message, the SOCM (SEE
      ABOVE) which specifies the message it is acknowledging.  The SOCM
      includes a checksum of its total contents.  If there is a checksum
      error, THE SOCM IS IGNORED.

      When there is a checksum error on a received normal message, the
      Receiver asks its Sender to send out a SOCM with the LACK flag
      set, or set the LACK flag on its next message.  The Sender sends
      this flag ONCE then ceases to increment the acknowledgement number
      (the ACK) while the Receiver continues to check incoming messages
      for the sequence number of the message with a checksum error.
      (Note that it continues to react to the acknowledgement field in
      the incoming messages.) When it finds the needed message, it
      resumes accepting the data in new messages and increments the
      acknowledgement number transmitted accordingly.

      The sending SLIP must never send a message greater than four past
      the last message for which it has received an acknowledgement
      (effectively a window size of four). Under normal processing
      loads, a window size greater than four should not be needed, and
      this decreases the probability of random errors creating valid


Farber & Delp & Conte                                          [Page 18]



RFC 914                                                   September 1984
Thinwire Protocol


      acknowledgement or sequence numbers.  If the Sender has four
      unacknowledged messages outstanding, it will retransmit the old
      messages, starting from the oldest unacknowledged message.  If it
      receives an acknowledgement with the LACK flag set, it transmits
      the message following the LACK number and continues to transmit
      the messages from that one on.  Thus a LACK is a message asking
      the Sender to please the Receiver.  If the Sender times out on any
      message not logically greater than four past the last acknowledged
      message, it should retransmit the message that timed out and then
      continues to transmit messages following the timed out message.

      The following describes a partial implementation of SLIP.  System
      dependent subjects like buffer management, timer handling and
      calling conventions are discussed.

      The SLIP implementation is subdivided into four modules and two
      sets of input/output interfaces.  The four modules are: The Sender
      Task, The Receiver Task, the buffer Manager, and SLIPTIME (the
      timer). The two interfaces are to the higher protocol and to the
      lower protocol (the UARTian, an interrupt driven device driver for
      the serial lines).

   OPERATIONS OF THE SENDER TASK

      The Sender Task takes a relatively noncomplex approach to
      transmitting.  It sends message zero, sets a timer (using the
      SLIPTIME Task) on the message, and proceeds to send and set timers
      for messages one, two, and three.  When the Receiver Task tells
      the Sender Task that a message has been acknowledged, the Sender
      Task then clears the timer for that message, and marks it
      acknowledged.  When the Sender Task has finished sending a
      message, it checks several conditions to decide what to do next.
      It first checks to see if a LACK has been received. If it has then
      it clears all the timers, and begins retransmitting messages
      (updating the acknowledgement field and checksum) starting from
      the one after the LACK'ed message.  If there is not a LACK waiting
      for the Sender Task, it checks to see if any messages have timed
      out.  If a message has timed out, the Sender Task again will clear
      the timers and begin retransmitting from the message number which
      timed out.  If neither of these conditions are true, the Sender
      Task checks to see if, because it has looped back to retransmit,
      it has any previously formulated messages to send.  If so, it send
      the first of these messages. If it does not have previously
      formulated messages, it checks to see if it is more than three
      past the last acknowledged message.  If so, it restarts from the
      message after the last acknowledged message.  If none of these are
      true, then it checks to see if there is more data waiting to be
      transmitted.  If there is more data available, it forms the
      largest packet it can, and begins to transmit it.  If there is no


Farber & Delp & Conte                                          [Page 19]



RFC 914                                                   September 1984
Thinwire Protocol


      more data to transmit, it checks to see if it needs to acknowledge
      a message received from the other side.  If so then it sends a
      SOCM.  If none of the above conditions create work for the Sender
      Task, the task suspends itself.

      Note that the Sender Task uses the Receiver Task to find out about
      acknowledgements and the Receiver Task uses the Sender Task to
      send acknowledgements to the other SLIP on the other side (via the
      ACK field in the Sender Task's message). The two tasks on one
      machine communicate through a small buffer. Because
      acknowledgements need to be passed back to the Sender Task
      quickly, the Receiver Task can wake up the Sender Task (unblock
      it).

   OPERATIONS OF THE RECEIVER TASK

      The Receiver Task checks the checksums of the messages coming into
      it.  When it gets a checksum error, it tells the Sender Task to
      mark the next acknowledgement as a LACK.  It then throws away all
      messages coming into it that don't match the message it wants and
      continues to acknowledge with the last ACK until it gets the
      message it wants.  As a checksum error could be the result of a
      crashed packet, and the StH character can occur within the packet,
      when a checksum error does occur, the recovery includes scanning
      forward from the last StH character for the next StH character
      then attempting to verify a packet beginning from it.  A valid
      message includes a valid checksum, and sequence and
      acknowledgement numbers within the active window of numbers.  This
      eliminates the need for the resequencing of messages, because the
      Receiver Task throws away anything that would make information in
      its buffers out of sequence.

   OPERATIONS OF SLIPTIME

      The timer task will maintain and update a table of timers for each
      request.  Its functions should be called with the timer length and
      the sequence number to associate with the timer.  Its functions
      can also be called with a request to delete a timer.  An
      interrupt-driven mechanism is used to update the running timers
      and to wake up the Sender when an alarm goes off.











Farber & Delp & Conte                                          [Page 20]



RFC 914                                                   September 1984
Thinwire Protocol


   THE INPUT AND OUTPUT INTERFACES

      To force SLIP to do something, the higher protocol should create a
      buffer and then call SLIP, passing it a pointer to the buffer.
      SLIP will then read the buffer and begin sending it.  The call to
      SLIP will return the number of bytes written, negative number
      indicates to the caller that SLIP could not do the request.  Exact
      error numbers will be assigned in the future.  To ask SLIP to
      receive something, one would call SLIP and SLIP would immediately
      return the number of bytes received or a negative number for an
      error (nothing ready to receive, for example).

      SLIP, when it wants to talk to the underworld of the serial
      interface, will do much the same thing only through a buffer
      written to by the UARTian (for received data) and read from by the
      UARTian (for sent data).

   OPERATIONS OF THE BUFFER/WINDOW MANAGER

      The Manager tends a continuous, circular buffer for the Sender
      Task in which data to be sent (from the downcalling protocol) is
      stored.  This buffer is called the INPUT-DATA BUFFER (IDBuff).
      The Manager also manages a SENDER TASK'S OUTPUT-DATA BUFFER
      (SODBuff), which is its output buffer to the UARTian.

      The IDBuff has associated with it some parameters.  These
      parameters include: START OF MEMORY (SOM), the start of memory
      reserved for the IDBuff; END OF MEMORY (EOM), the end of memory
      reserved; START OF DATA (SOD), the beginning of the used portion
      of the IDBuff; and END OF DATA (EOD), the end of data in the
      IDBuff.  The SOM and EOM are constants whereas the SOD and EOD are
      variables.

      The SODBuff is composed of four buffers for four outbound messages
      (less the checksum).  The buffers can be freed up to be
      overwritten when the message that they contain is acknowledged by
      the SLIP on the other side of the line.  When a message is in the
      SODBuff, it has associated with it a sequence number (which is the
      message's sequence number).  The Sender Task can reference the
      data in the SODBuff and reference acknowledgements via this
      sequence number.

      When the application has data to be transmitted, it is placed in
      the IDBuff by the application using functions from the Manager and
      the EOD is incremented.  If the data the application wants to send
      won't fit in the buffer, no data is written, and the application
      can either sleep, or continue to attempt to write data until the




Farber & Delp & Conte                                          [Page 21]



RFC 914                                                   September 1984
Thinwire Protocol


      data will fit. The Sender Task calls a Manager function to fill a
      message slot in the SODBuff.  The Sender Task then sends its
      message from the SODBuff.

      The Manager also maintains a buffer set for the Receiver Task. The
      buffers are similar to those of the Sender Task.  There is a
      CHECKSUMMED OUTPUT-DATA BUFFER (CODBuff), which is the final
      output from SLIP that the higher level protocol may read.  The
      CODBuff is also controlled by the four parameters START OF MEMORY,
      END OF MEMORY, START OF DATA, and END OF DATA (SOM, EOM, SOD, and
      EOD).

      There is also an inbound circular buffer the analog of the
      SODBuff, called the RECEIVER TASK'S INPUT-DATA BUFFER (RIDBuff).

      When the UARTian gets data, it places the data in the RIDBuff.
      After this, the Receiver Task checksums the data.  If the checksum
      is good and the Receiver Task opts to acknowledge the message, it
      moves the data to the CODBuff, increments EOD, and frees up space
      in the RIDBuff.  The higher level application can then take data
      off on the CODBuff, incrementing SOD as it does so.

   STARTING UP AND FINISHING OFF COMMUNICATIONS

      The problem is that the SLIP's on either side need to know (and
      keep knowing) the sequence number of the other SLIP.  The easiest
      way to solve most of these problems is to have the SLIP check the
      Request to Send and Clear to Send Lines to see if the other SLIP
      is active. On startup, or if it has reason to believe the other
      side has died, the SLIP assumes: all connections are closed, no
      data from any connection has been sent, and both its SEQ and the
      SEQ of the other SLIP are zero.  To start up a connection, the
      instigating SLIP sends a SSNM with its starting sequence number in
      it.  The receiving SLIP acknowledges this SSNM and replies with
      its starting sequence number (combined into one message).  Then
      the sending SLIP acknowledges the receiving SLIP's starting
      sequence number and the transmission commences.  This is the three
      way handshake taken from TCP, After which data transmission can
      begin.












Farber & Delp & Conte                                          [Page 22]