rfc100.txt

<VC++网络游戏建摸与实现>源代码

Network Working Group                                             P. Karp
Request for Comments: XXXX                                          MITRE
NIC: 5761                                                26 February 1971


                  Categorization and Guide to NWG/RFCs

   The NWG/RFC Guide is an attempt to introduce some order into the
   NWG/RFC series, which now numbers 102.  The Guide categorizes the
   NWG/RFC notes, identifies topics under discussion and the relevant
   NWG/RFCs, and indicates whether the notes are current, obsolete, or
   superseded.

   A minimum subset of NWG/RFCs is identified.  This subset consists of
   the NWG/RFCs that one should read to quickly become familiar with the
   current status of topics.

   For historical reasons and for readers interested in tracing through
   the stages of development of a topic, a brief summary is given for
   each NWG/RFC relevant to a particular category.

   This initial Guide is being issued as a NWG/RFC since it establishes
   the basis for future releases.  So, please comment! Suggestions,
   criticism, corrections, etc., will be accepted for a period of
   approximately two weeks.  Be critical as I have not had to implement
   an NCP and probably have some misconceptions regarding various
   technical points.  An official version will be released on March 26.
   The Guide will then be a unique series of documents, separate from
   NWG/RFCs (as is the Document No. 1, No. 2 series).

   With regard to renumbering NWG/RFCs, I am inclined to keep she
   sequential numbering scheme presently employed.  The main reason for
   this position is that the current numbers have both historical and
   semantic significance.  For example, reference to "#33, #66, #83,
   etc." is a convenient shorthand (reminiscent of the old corny joke
   about joke #s) used extensively during meetings.  The list of
   "current status" NWG/RFC numbers should dispel any fear of
   maintaining stacks of NWG/RFCs for quick reference.  The subject is
   not closed, however, and I will entertain any objections,
   suggestions, etc.

GUIDE TO NETWORK WORKING GROUP/REQUEST FOR COMMENTS

   The NWG/RFC notes are partitioned into 9 categories, which in turn
   are divided into subcategories.  For each category the official
   document (if any), unresolved issues, and documents to be published
   are identified.




Karp                                                            [Page 1]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


   For each subcategory, relevant NWG/RFCs are listed and a brief
   description of the topics addressed in each note is given.

   The categories are again listed and the current NWG/RFCs identified
   (p. 23).  The NWG/RFCs in the list comprise the subset defining
   "current status".  Note that most of the documentation in the subset
   addresses topics in Category D - Subsystem Level Protocol, where at
   the present time most issues are unresolved.

   Finally, the NWG/RFCs are listed by number, with a reference to the
   relevant categories (p. 26).

A. ADMINISTRATIVE

A.1 Distribution list

   NWG/RFC #s: 3, 10, 16, 24, 27, 30, 37, 52, 69, 95

   The distribution list contains names, addresses, and phone numbers
   for recipients of NWG/RFCs.  The most recent list, NWG/RFC 95,
   designates the Technical Liaison as the recipient for each site and
   supersedes all other RFCs in this category.

A.2 Meeting announcements

   NWG/RFC #s: 35, 43, 45, 54, 75, 85, 87, 99

   General network working group meetings are held approximately every
   three months.  Special subcommittee meetings are held on an ad hoc
   basis.  All related NWG/RFCs are obsolete except 87, announcing a
   graphics meeting to be held at MIT in April and 99, announcing a
   general NWG meeting, Atlantic City, May 16-20.

A.3 Meeting minutes

   NWG/RFC #s: 21, 37, 63, 77, 82

   The meeting minutes present highlights of issues discussed at general
   NWG meetings and report definite decisions that are made.

   To be published: A NWG/RFC will be published by Dick Watson, SRI,
   reporting on the NWG meeting held at the University of Illinois,
   February 17-19.








Karp                                                            [Page 2]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


A.4 Guide to NWG/RFCs

   NWG/RFC #s: 84, 100

   The NWG/RFC Guide categorizes the NWG/RFC notes, identifies topics
   under discussion, the relevant NWG/RFCs, and denotes whether the
   notes are current, obsolete, or superseded.  Included in this
   category are lists of NWG/RFCs, ordered by number (as in 84) and/or
   by author.

A.5 Policies

   NWG/RFC #s: 18, 24, 25, 27, 30, 37, 41, 48, 53, 54, 72, 73, 77, 82,
               102

   NWG/RFCs categorized as policy contain official stands on issues
   i.e., the position taken by S. Crocker, NWG Chairman.  The issues
   covered are varied.

   In particular:

   77 and 82 discuss meeting policy.

   72, 73, 77, and 82 discuss the decision to delay making changes to
   the Host/Host protocol in order to first gain experience with the
   network.  A committee to propose specific changes has been formed.

   37 discusses changes to the Host/Host protocol and the schedule for
   introducing modifications.

   53 sets forth the mechanism for establishing and modifying the
   official Host/Host protocol.

   54 presents the initial official protocol.

   48 presents some suggestions for policy on some outstanding issues.

   41 requests the tagging of IMP-IMP teletype messages.

   Documentation conventions for NWG/RFCs are given in 24, 27, and 30.

   25 and 18 designate uses for particular link numbers. 25 has been
   superseded by 37 and 48. 18 is obsolete.

   102 discusses the issuing of Document #2, in lieu of the official
   modification procedure outlined in 53.





Karp                                                            [Page 3]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


B. HOST/IMP PROTOCOL (LEVEL 1)

   Official document: BBN Memo No. 1822 (latest revision - February
   1971)

   Unresolved issues: Location of first byte of data in a message.

   To be published: Document No. 2 will be written by S. Crocker and
   will, among other things, resolve the first byte location issue.

B.1 General Topics

   NWG/RFC #s: 17, 17a, 19, 21, 33, 36, 37, 38, 46, 47, 102

   In particular:

   17 raised several questions regarding HOST/IMP protocol.  In 17a,BBN
   responds to the questions.

   19 proposes that the hosts control the ordering of IMP/Host traffic
   rather than getting messages delivered in the order received by the
   IMP.  This proposal is counter to BBN's position, specifically
   expressed in 47; that is, buffering is a Host rather than an IMP
   function.  The purpose of buffering in the IMP is to handle surges of
   traffic, thus IMP buffers should be empty.  NWG/RFC 19 is obsolete.

   21 discusses changes to BBN Memo No. 1822.  The remarks are obsolete.

   33 contains a general description of the interface between a host and
   the IMP.  NWG/RFC 47 comments on NWG/RFC 33.

   The use of RFNMs (type 10 and type 5 messages) to control flow is
   discussed in NWG/RFCs 36, 37 and 46.  The official position in "cease
   on link" (i.e., discontinue the mechanism) is presented in 102 and
   renders obsolete the remarks in 36, 37, and 46.

   38 discusses the changes to message format that would be necessary if
   multiplexing connections over links was allowed.













Karp                                                            [Page 4]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


B.2 Marking/Padding

   NWG/RFC #s: 44, 48, 49, 50, 54, 64, 65, 67, 70, 102

   In particular:

   102 presents the decision of the Host/Host protocol committee to
   abandon the marking convention and to ignore padding.  The issue of
   whether to have the first data byte begin after 72 bits of header or
   to use double physical transmission (NWG/RFC #s 65, 67) is discussed.

   The former official position is expressed in 54: "All regular
   messages consist of a 32 bit leader, marking, text, and padding.
   Marking is a (possibly null) sequence of zeros followed by a 1;
   padding is a 1 followed by a (possibly null) sequence of zeros."

   Several proposals to eliminate marking have been made. 64 suggests a
   hardware modification to eliminate marking/padding by adding
   appropriate counters to Host/IMP interfaces. 65 suggests breaking
   regular messages into two messages. 67 supports 65. 72 and 73 suggest
   that such changes be postponed until sufficient experience with the
   network is gained.

   44 introduces the notion of double padding and presents two
   alternative approaches when a message does not end on a Host word
   boundary:

      a) The host provides padding in addition to the IMPS ("double
         padding")

      b) The host shifts messages to end on a word boundary.

   48 explains double padding in more detail and discusses the pros and
   cons.  A suggestion is made to use marking to adjust the word
   baundary (alternative b).  NWG/RFCs 49 and 50 are concurrences with
   48.

   70 presents a method to handle the stripping of padding from a
   message.

   All NWG/RFCs in this category have been superseded by 102.

C. HOST/HOST PROTOCOL (LEVEL 2)

   Host/Host protocol specifies the procedures by which connections for
   inter-Host interprocess communication over the network are
   established, maintained, and terminated.  The software which
   implements the protocol within each Host is called the Network



Karp                                                            [Page 5]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


   Control Program (NCP).  The topics included in this category are
   connection establishment and termination, flow control, interrupt
   handling, error control and status testing, dynamic reconnection, and
   the relationship between connections and links.

   Official documents: Document No. 1 by S. Crocker, 3 August 1970, with
   modifications presented in NWG/RFC 102.

   Unresolved issues: Length of control messages
                      Location in message of first byte of data
                      Flow control algorithm
                      Socket identification format

   To be published: Document No. 2 will be written by S. Crocker and
   will resolve the first three issues.  A NWG/RFC will be written by J.
   Heafner, in collaboration with E. Meyer and G. Grossman. presenting
   the pros and cons on alternative proposals for socket number
   identification.

C.1 Host/Host Protocol Proposals

   NWG/RFC #s: 9, 11, 22, 33, 36, 37, 38, 39, 40, 44, 46, 48, 49, 50,
               54, 57, 59, 60, 61, 62, 65, 68, 93, 102

   The official Host/Host protocol presented in Document No. 1 is based
   on the proposals, discussions, acceptance, and rejection of ideas in
   the above list of NWG/RFCs, up to and including 59.

   In particular:

   9, 11, and 22 represent an early attempt at a Host/Host protocol. 11
   supersedes 9 and 22 contains some modifications to control message
   formats presented in 11.  The protocol was not considered powerful
   enough because it didn't provide for inter-host communication without
   logging in.  This protocol was thrown out as a result of a network
   meeting in December 1969.

   33 is the basis for the current protocol.  It was presented at the
   SJCC, 1970.

   36 is a modification of 33.  It discusses connection establishment
   without switching, flow control, and introduces the idea of
   reconnection.  Control commands are summarized. 36 was distributed at
   a Network meeting in March 1970.







Karp                                                            [Page 6]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


   37 presents the reaction to 36 and presents ideas on reconnection
   flow control and decoupling of links and connections.  Provisions of
   error detection, status testing, experimentation and expansions are
   discussed.

   38, 39, 40, 44, 49 and 50 are comments written in response to the
   meeting. 46 is also a comment but in the form of a rewrite of 33. 46
   introduces the notion of interrupts, INT, and ECO for status testing.

   47 concerns the philosophy behind the notion of a link.

   48 summarizes the issues discussed in the above NWG/RFCs.

   54 is the initial official protocol submitted for criticism,
   comments, etc.  It introduces a new mechanism for flow control in
   which the receiving host allocates buffer space and notifies the
   sending host of the space available.

   57 and 59 comment on 54.

   Document No. 1 differs from NWG/RFC 54 as follows: commands GVB and
   RET have been added for flow control and error condition codes have
   been added to ERR.  NWG/RFC 102 presents some modifications to
   Document No. 1: fixed lengths are specified for ECO, ERP, and ERR; a
   new pair of commands RST and RRP (suggested in 57) are added.

   60, 61, and 62 propose new Host/Host protocols, quite different from
   the current official protocol. 62 supersedes 61. 60 and 62 are worth
   considering for possible implementation in future protocols.
   Hopefully, more documents of a similar nature will be generated as
   experience is gained with the current protocol.

   NWG/RFCs 65 and 68 comment on Document No. 1.

   93 points out an ambiguity in Document No. 1 regarding the
   requirement of a message data type in the message sent from server
   socket 1.  The ambiguity is resolved by 102 which eliminates message
   data type from level 2 protocol.

C.2 NCPs (Description, Structure, Techniques)

   NWG/RFC #s: 9, 11, 22, 23, 33, 36, 44, 46, 48, 55, 70, 71, 74, 89

   This category includes RFCs which give details of system calls, table
   structures, implementation techniques, etc.






Karp                                                            [Page 7]

RFC 100           Categorization & Guide to NWG/RFC's   26 February 1971


   In particular:

   NWG/RFCs 9, 11, and 22 are obsolete

   23 is a general statement on sending or receiving multiple control
   messages in a single communication.

   33 discusses the system calls used for interaction between the NCP
   and a user process.

   36 describes a possible implementation giving table structures and
   their interrelationships.

   44 lists the system calls that SDC feels should operate, includes
   spec. of calls to NCP.

   NWG/RFC 48 presents Postel's and Crocker's view on the environment in
   which a host time-sharing system operates, suggests some system