Network meeting
Summary of an October 1969 ARPANET network meeting and discussions.
Cerf, V.G., "Network meeting", RFC 21, DOI 10.17487/RFC0021, October 1969, <https://www.rfc-editor.org/info/rfc21>.
@misc{rfc21,
series = {Request for Comments},
number = 21,
howpublished = {RFC 21},
publisher = {RFC Editor},
doi = {10.17487/RFC0021},
url = {https://www.rfc-editor.org/info/rfc21},
author = {V.G. Cerf},
title = {{Network meeting}},
pagetotal = 2,
year = 1969,
month = oct,
} RFC text
Contents
Network Working Group V. Cerf
Request for Comments: 21 UCLA
October 17, 1969
At UCLA on October 10, there was a network meeting attended by:
SDC UCLA
John Kreznar Vint Cerf
Dick Linde Steve Crocker
Marty Bleier Jon Postel
Bob Long Michel Elie
UCSB
Ron Stoughton
Nancy O'Hara
George Gregg
Topics discussed:
1. Revisions to BBN memo 1822
2. Revisions to NWG/RFC 11
3. Transmission of multiple control messages
1. Changes to BBN Memo No. 1822 (underlined)
As informally communicated by Dave Wa
p. 11 "The IMP program can handle up to 63 active transmit links
and 63 active receive links at a time. If the Host attempts to
send a message on a new link when 63 active transmit links already
exist, a "Link Table Full" message will be sent from the IMP to
the Host, and the message will be discarded."
p. 11 "1. Any link that is not used for a period of 26 seconds
will have its entry automatically deleted by the IMP program."
[Cerf: How about deleting only if the transmit link table is full?
Crocker: No, because there is no other way for links to be deleted
so they would always tend to accumulate. Furthermore,
the table at one site may be full while another site may
not be.]
p. 13 "5 Regular with discard."
This allows IMP systems to generate traffic which never actually
reaches any Hosts since it will be discarded when it reaches the
top of the IMP-HOST queue in the destination Host's IMP. The
Network Measurement Center will make use of this feature.
p. 13 Message type 6 is no longer assigned, and message type 10
is really in octal so is actually type 8. Types 9-15 are unassigned.
p. 17 Type 10 is really type 8.
2. Revisions to NWG/RFC 11
This memo has been obsoleted by developments at UCLA and
discussions with other nodes. NWG/RFC 22 contains some of the
major changes. An updated version of NWG/RFC 11 is forthcoming.
3. George Gregg of UCSB will publish NWG/RFC 23 concerning the
transmission of multiple control messages over control links.
AI ELI5
On this page
§1Meeting Overview
On October 10, 1969, representatives from UCLA, SDC, and UCSB met to discuss early ARPANET protocol issues. The attendees included Vint Cerf and Steve Crocker from UCLA, John Kreznar from SDC, and Ron Stoughton from UCSB. The meeting covered three main topics: changes to the IMP-host specification, revisions to earlier RFCs, and a new proposal for transmitting multiple control messages.
Network engineers from three sites got together to talk about how the network should work.
§2IMP-Host Specification Changes
Changes to BBN Memo 1822 included clarifications on link table management. The IMP could handle up to 63 active transmit and receive links at a time. Links that were inactive for 26 seconds would have their entries automatically deleted. Discussion centered on the proper behavior: should links only be deleted when tables are full, or automatically to prevent accumulation? The meeting confirmed automatic deletion, as otherwise links would accumulate indefinitely.
They set limits on how many connections could be open at once and said old unused connections should be forgotten after 26 seconds.
§3Message Types and Updates
The memo also clarified message type assignments. Message type 6 was no longer assigned. Message type 10, which had been causing confusion, was clarified to really be type 8 in octal notation. Types 9 through 15 remained unassigned, reserving them for future use. RFC 11, the earlier protocol specification, was being superseded by newer developments at UCLA, with RFC 22 containing major changes. A fuller updated version of the protocol was planned.
They cleaned up which control message numbers were used and which were saved for later.
§4New Work
A new RFC 23 was to be published by George Gregg of UCSB on transmitting multiple control messages over control links. This follow-up work addressed the practical challenge of sending multiple control messages efficiently in a single packet.
A new note was being prepared about how to send several control messages at once.