RFC 60






Network Working Group                                           R. Kalin
Request for Comments: 60                                             MIT
                                                            13 July 1970


                       A Simplified NCP Protocol

Abstract



   This RFC defines a new NCP protocol that is simple enough to be
   implemented on a very small computer, yet can be extended for
   efficient operation on large timesharing machines. Because worst case
   storage requirements can be predicted, a conservative implementation
   can be freed of complicated resource allocation and storage control
   procedures. A general error recovery procedure is also defined.

Overview and Rational

   The central premise of this proposal is an insistence that all user-
   to-user connections be bi-directional. For those familiar with
   communication theory, this appears most reasonable. All communication
   requires a cyclical flow of information. To deny a simple association
   between a message and its reply makes protocol unnecessarily
   complicated and turns simple mechanisms of flow control into
   nightmares.

   It is proposed that a bi-directional connection, or duplex link, be
   identified by a pair of socket numbers, one for each end. This is
   half the number presently required. Associated with the connection
   are some number of "crates" or message containers. These crates
   travel back and forth over the link carrying network messages from
   one side to the other. Buffers are allocated at each end of the link
   to hold crates and the messages that they carry. Worst case buffer
   requirements are equal to the number of crates in circulation, or the
   "capacity" of the link.

Details

   A message buffer has four states which follow one another cyclically.
   They are:

   1) empty,

   2) filled with a message-laden crate to be unloaded,

   3) filled with an empty crate, and

   4) filled with a message-laden crate to be sent.



Kalin                                                           [Page 1]

RFC 60                  A Simplified NCP Protocol           13 July 1970


   Normally state transitions correspond to message arrival, message
   removal, message insertion and message transmission.

   For a process to be an NCP it must:

   1) be able to make initial contact with foreign hosts via the control
   link and, if necessary, delete user-to-user links left over from the
   previous system incarnation.

   2) be able to create user-to-user links.

   3) be able to interface users with these links.

   4) be able to delete user-to-user links.

   The first of the four functions shall not be discussed here except to
   point out that it contains critical races that can not be resolved
   without making assumptions about maximum message propagation delays.
   Since within the ARPA network, bounds on message turnaround time do
   not exist, the approach chosen must necessarily be tender. The other
   three functions are discussed first from the viewpoint of one
   interested in implementing a minimal NCP. Then extensions and
   improvements are proposed that are suitable for larger machines.

   Any NCP must be capable of creating a duplex link between a local
   user process and a remote one. The current protocol accomplishes this
   by queuing a potentially unbounded number of RFC's and waiting for
   the user to examine the queue to determine with whom he wishes to
   talk.  There is no guarantee that the user will ever look at the
   queue and there is no way to limit the size of the queue. The
   overflow error message suggested fails in the respect because it
   admits that the RFC will only be sent again. The picture need not be
   this bleak. The following network conversation demonstrates how
   connections can be made without using queues or relying on user
   process attention.

   Suppose that a local user process and a remote user process wish to
   establish a new connection. The remote process asks its NCP to listen
   for a connection request and gives it the socket identifier for its
   end. Optionally it can give both socket identifiers. The user process
   at the local end asks its NCP to send a request for a duplex link
   (RFDL). It specifies both socket identifiers of the proposed link.
   The local NCP sends a RFDL over the control link with the following
   format:

   RFDL <my socket> <your socket> <max number buffers> <spare>





Kalin                                                           [Page 2]

RFC 60                  A Simplified NCP Protocol           13 July 1970


   The third argument is normally supplied by the local NCP and
   indicates the maximum number of buffers the NCP will consider
   allocating to this duplex link. If buffers are in user storage the
   count may be given by the user in a call made to the NCP.

   The RFDL is received at the remote host and the remote NCP compares
   <my socket> and <your socket> against the socket identifiers supplied
   by unmatched listens issued to it. For listens in which just a single
   identifier was given only <your socket> must match. If both socket
   identifiers were given, they both must match. If a match is found an
   acknowledgement message with the following format is sent back by the
   NCP:

   ACDL <your socket> <my socket> <number buffers> <spare>

   The <number buffers> parameter is equal to the smaller of <max number
   buffers> as specified in the RFDL and the number of message buffers
   agreeable to the remote NCP. If no match is found the error message
   returned is an ACDL in which <number buffers> equals zero. Note that
   the RFDL mechanism is similar to a RFC mechanism in which the bound
   on queue size is one and connection acceptance is done entirely by
   the NCP.

   The two varieties of a listen correspond to two modes of channel
   operation. The single parameter variety, as typified by a LOGIN
   process, is to be used by programs that will "talk with anyone who
   happens to dial their number". Screening of contacts for
   appropriateness is left to the user process. The double parameter
   listen is used by user programs who know with whom they will
   communicate and do not wish to be bothered by random RFDL's from
   other sources. Given the way in which socket name space is
   partitioned, it is impossible to get a matching RFDL from any process
   but the one intended.

   Message buffers for the connection are allocated in the remote host
   before it sends the ACDL and in the local host at the time the ACDL
   is received. The number of buffers at each end is equal to the
   <number buffers> parameter in the ACDL. The state of all remote
   buffers is "empty" and of all local buffers "filled with empty
   crate". After buffers are allocated the local user process is
   notified that it is able to start sending messages.

   The type of interface presented by the NCP between the user process
   and the newly created duplex link is a decision local to that host. A
   simple but complete interface would provide two calls to be made to
   the NCP. GETMESSAGE would return the next message from the link
   complete with marking, text and padding. PUTMESSAGE would take a




Kalin                                                           [Page 3]

RFC 60                  A Simplified NCP Protocol           13 July 1970


   message, marking and text only, and buffer it for transmission. The
   obvious logical errors would be reported.

   We suggest that message alignment be left to the user. On most
   machines it is a simple but time consuming operation. If done in the
   NCP there is no guarantee that the user will not have to readjust it
   himself. It is usually not possible to know a priori whether the text
   portion should be right adjusted to a word boundary, left adjusted to
   a word boundary, aligned to the end of the last message, or
   fragmented in some exotic way.

   Within this protocol message boundaries are used to provide storage
   allocation information. If not required by the user this information
   can be forgotten and the user interface can be made to appear as a
   bit stream. Though welcomed by purists, such a strategy may produce
   complications when attempting to synchronize both ends of a link.

   Links are deleted by removing empty crates from them and reclaiming
   the buffers allocated to the crates removed. Only buffers with crates
   in can be reclaimed; empty buffers must remain available to receive
   messages that may arrive. When no crates are left, no buffers remain,
   and the socket identifiers can be forgotten. When empty crates are
   removed, a decrement size message is sent to the foreign NCP to allow
   it to reduce its buffer allocation:

   DEC <my socket> <your socket> <number of buffers dropped>

   A reply is solicited from the foreign NCP to affirm the deletions or
   to complain of an error. Possible errors include "no such link" and
   "impossible number of buffers dropped".

   The option to close a link can be given to a user process by
   providing either of two system calls. NOMOREOUTPUT declares that no
   more messages will be sent by the local user process. All local
   buffers for the link that contain empty crates are reclaimed by the
   NCP. DEC messages are sent to the foreign NCP. As crates are emptied,
   via GETMESSAGE calls, their buffers are reclaimed too. As an
   alternative, the call KILLMESSAGE can be implemented. This call can
   be used in place of a PUTMESSAGE. Instead of filling an empty crate
   with a message to be sent, KILLMESSAGE will cause the crate to be
   reclaimed and a DEC control message sent.

   In situations where the user process has died, or for some other
   reason can not close the link, more drastic measures must be taken.
   For these situations, the ABEND control message is defined:

   ABEND <my socket> <your socket>




Kalin                                                           [Page 4]

RFC 60                  A Simplified NCP Protocol           13 July 1970


   After sending an ABEND the issuing NCP starts to close the link. All
   buffers containing input are destroyed. A DEC is issued for these and
   the previously empty buffers. If messages arrive on the link, they
   are destroyed and a DEC is issued. Any ABEND received for the link is
   ignored.

   When the remote NCP receives the ABEND, it stops sending messages
   over the link and refuses new messages from the user process at its
   end.  Empty buffers are reclaimed. Pending output messages are
   destroyed and their buffers reclaimed. Input messages are fed to the
   user process as long as it will accept them. Buffers are reclaimed as
   input is accepted. DEC's are issued to cover all buffers reclaimed.
   When the user process will take no more input, input messages are
   destroyed and their buffers reclaimed. Eventually all buffers will be
   reclaimed at both ends of the link. At such time the connection can
   be considered closed and the socket numbers used can be reassigned
   without ambiguity.

   Under this proposed protocol the above four functions constitute all
   that must be part of a network NCP. If buffers are allocated only
   when free ones exist there can be no "overflow" errors nor is there
   any need to place further constraints on message flow. For any user
   message that arrives buffer room is guaranteed. All control messages
   can be processed without requiring additional storage to be
   allocated.  Attempts by a user process to issue too many listens can
   be thwarted by local control procedures.

   Inefficiencies in storage will result when the number of outstanding
   connections gets large. One price of coding simplicity is a fifty
   percent utilization of buffer space. On large hosts it may prove
   advantageous to implement some of the following NCP extensions. With
   more complicated flow control procedures, it becomes possible for an
   NCP to allocate more buffer space than actually exists and still not
   to get into trouble. Other extensions provide message compression,
   improved throughput and user transparent error recovery.

   Because some extensions require the cooperation of foreign hosts and
   assume that they have implemented more than the minimal NCP it is
   proposed that an inquiry control message be used to find out what
   extensions the foreign host has implemented. The response to an INQ
   will be a control message defining a host profile. If an "undefined
   error" message is returned, the foreign host is assumed to have only
   a minimal NCP.

   A simple extension is to define a control message that will replace
   user RFNM's. A user RFNM is a null text message sent, for example, as
   a reply when a file is transferred via a duplex link. They are
   inefficient since they tie up an entry in the IMP's link assignment



Kalin                                                           [Page 5]

RFC 60                  A Simplified NCP Protocol           13 July 1970


   table and degrade network throughput. A more efficient solution is to
   send a special message over the control link. In this way one short
   message can replace several user messages.

   URFNM <my socket> <your socket> <number of user RFNM's>

   Because the control link is concurrent with the return side of the
   user link, URFNM's can not be substituted for user RFNM's when there
   are other messages to be sent on the return link. Otherwise ordering
   will be lost and with it user transparency.

   Throughput can also be increased with a mechanism to add additional
   crates on a duplex link. This might be at user instigation or be a
   decision of the NCP.

   INC <my socket> <your socket> <number buffers desired>

   The foreign host replies to an increase request by returning an INCR.

   INCR <my socket> <your socket> <number buffers to be added>

   If the foreign NCP is unable to meet the additional buffer demand,
   <number buffers to be added> will be less than <number buffers
   desired> and possibly zero. The initial state of all local buffers
   added is "filled with empty crate" and of all foreign buffers
   "empty".

   The spare argument in the RFDL and ACDL could be used to declare the
   maximum sized message that will actually be sent in that direction. A
   perceptive NCP could observe this information and allocate smaller
   buffers. A lesser NCP could ignore it and always assume maximum
   length messages. For example, if the field were zero then only user
   RFNM's would be sent. A smart NCP would allocate no storage at all.

   If the NCP retains a copy of each user message sent over the network
   until a reply is returned, an automatic error recovery procedure can
   be implemented. Because the capacity of the link is always known, an
   NCP can determine whether there are messages in transit. This is done
   by first sending a STOP message to the foreign NCP:

   STOP <my socket> <your socket>

   The STOP message tells the foreign NCP to temporarily stop
   transmitting messages over the selected link. Unlike CEASE-ON-LINK
   there is no guarantee as to how many messages will be sent before the
   STOP takes effect. The local NCP then sends a link inquiry message:

   LINQ <my socket> <your socket>



Kalin                                                           [Page 6]

RFC 60                  A Simplified NCP Protocol           13 July 1970


   The reply gives the number of crates at the foreign end of the link.
   The LINQ message is repeated until this number plus the number of
   local crates equals the capacity of the link. At this time no
   messages are in transit and the two ends of the link have been
   synchronized.  Messages can now be identified relative to the
   synchronization point.  Thus the local NCP can send a control message
   asking, for example, that the third to last message be retransmitted.
   The foreign NCP is able to identify which message this is and to
   retransmit it. Once all errors have been recovered a START control
   message, similar in format to the STOP, is sent to the foreign NCP
   and normal operation continues. The entire recovery procedure can be
   transparent to both user processes.

   It is expected that the larger hosts will not adhere strictly to the
   worst case storage allocation requirements. Rather they will allocate
   more buffers than they have and reply on statistics to keep them out
   of trouble most of the time. Such conduct is perfectly permissible as
   long as it is transparent to foreign hosts. The protocol allows an
   NCP to lie about storage allocation as long as he is not caught. In
   situations where detection appears imminent some of the following
   control mechanisms will need to be applied. They are listed in
   increasing order of power.

   a) Do not send out any user RFNM's or other short messages. There is
   a good chance that they will be replaced by longer messages that will
   strain buffer capacity even more.

   b) Try not to accept any new messages from the IMP. Block local
   processes attempting to issue messages.

   c) Issue DEC's to free up buffer space. Do not allocate more than one
   buffer to RFDL's and refuse INC's.

   d) Fake errors in messages waiting for local user action. Do this
   only if the host that sent it has implemented error recovery. This
   will free buffer space and allow you to recover later. This final
   measure is admittedly a last resort, but it should be powerful enough
   to control any emergency.

   It is the hope of the author that the above protocol presents an
   attractive alternative to that proposed by RFC 54 and its additions.
   Although it appears at a late date, it should not be more than a
   minor jolt to implementation efforts. It is simple enough to be
   implemented quickly. If adopted, a majority of the present sites
   could be talking intelligently with one another by the end of the
   summer.





Kalin                                                           [Page 7]

RFC 60                  A Simplified NCP Protocol           13 July 1970


References

   [1] Crocker, S.D., Postel, J., Newkirk, J. and Kraley, M., "Official
   protocol proffering", RFC 54, June 1970.

Author's Address



   Richard Kalin
   MIT Lincoln Laboratory







         [ This RFC was put into machine readable form for entry ]
           [ into the online RFC archives by Ian Redfern 4/97 ]

































Kalin                                                           [Page 8]