[ih] Internet-history Digest, Vol 69, Issue 11

[John Shoch]

My friends at the Computer History Museum always warn me about declaring
"firsts":
a.  you better get the facts absolutely correct, and
b.  you better be fanatically precise in defining the terms (every noun,
every adjective, etc.) -- and you better include the definitions.

The UCLA statement which started this exchange lacks both -- thus, almost
by definition, it is ambiguous and imprecise (and thus probably wrong).

Efforts at NPL and Sage are certainly worth looking at.

In addition, there was earlier work at SDC -- apparently in 1963 w/SRI, and
ca. 1966 with MIT Lincoln Labs.  I will make no judgement about any
"firsts" but let me bring to everyone's attention a couple of items:

--There is an interesting and comprehensive historical look at this Q-32
work in a paper by David Hemmendinger, published in 2016:
"Two Early Interactive Computer Network Experiments."
https://www.computer.org/csdl/magazine/an/2016/03/man2016030012/13rRUwciPgt
You can also access it at:
https://cs.union.edu/~hemmendd/History/network6.pdf

--In that paper he discusses an experiment in 1963 between SRI and SDC.  At
that time Lick had taken over ARPA/IPTO, and ARPA had taken over the Sage
Q-32 prototype that had been built for Sage at SDC.  Lick wanted SRI to
connect to the Q-32.  Doug Engelbart described the work at the History of
Workstations conference in 1986 [I spoke at the conference, and heard
Doug's talk, but 40 years later I do not remember these comments]:
https://dl.acm.org/doi/pdf/10.1145/61975.66918
"Lick moved very swiftly. By early 1963 we had a funded project. But,
whereas I had proposed using a local computer and building an interactive
workstation, Lick asked us instead to connect a display to the System
Development Corporation's (SDC's) AN/FSQ32 computer, on site in Santa
Monica, to do our experimenting under the Q32's projected new time-sharing
system. (Converting the Q32 to be a timeshared machine was SDC's IPTO
project.) Later that year, our project was modified to include an online
data link from Menlo Park to Santa Monica, with a CDC 160A minicomputer at
our end for a communication manager, supporting our small-display
workstation."

--Hemmendinger also discusses the more well-known work several years later,
ca. 1966, by Tom Marill (at CCA) and Larry Roberts (at MIT) to connect the
TX-2 to the Q-32 machine at SDC.

--There seems to have been a CCA Technical Report in mid-1966, but I have
never seen it;  Hemmendinger cites it as:
T. Marill, "A Cooperative Network of Time-Sharing Computers: Preliminary
Study," Technical Report No. 11, Computer Corporation of America,
Cambridge, Mass. (1966).
The preliminary study is also cited in a bibliography about the Arpanet:
https://apps.dtic.mil/sti/tr/pdf/ADA026900.pdf
Marill, T. A cooperative network of time-sharing computers: preliminary
study. Cambridge, Ma., Computer Corporation of America, I Jun 66. 53 p.
CCA-TR1-1 NIC 06458.  [Is this an SRI NIC identifier?]

--The Preliminary Report may have been a predecessor or an early draft or a
pre-print of a paper published later that year, to which Larry Roberts is
added as a co-author:
Thomas Marill and Lawrence G. Roberts, “Toward a cooperative network of
time-shared computers” in Proceedings of the AFIPS Fall Joint Computer
Conference, pp. 425-431, ACM, New York, NY (November, 1966).
https://dl.acm.org/doi/10.1145/1464291.1464336

--Some interesting highlights from the paper:
--They talk in passing about shipping programs from one machine to another,
but then focus only on providing remote terminal access -- from a terminal
on one computer, through a "network" to a program running on another
computer.
--An "elementary" model merely routes characters from a user's terminal
through to the local machine, and then out another terminal link to the
distant machine.  This requires no modifications at either end, but runs at
terminal speeds.
--They then expand the model:  "Thus, a possible alternative technique for
achieving increased data-rates without greatly increasing the burden on the
monitor would be to use high-rate data-only links, supplementing these by
low-rate command-plus-data channels over which communication to the remote
monitor could take place."  But this would require changes to the OS or
monitor.
--"The first step in that direction is the establishment of a message
protocol, by which we mean a uniform agreed-upon manner of exchanging
messages between two computers in the network."
--These are point-to-point messages, but can provide error control:  "The
primary reasons for considering the establishment of a message protocol are
the following: ... By formatting transmissions into messages, and including
a check-sum with each message, transmission errors can frequently be
detected. If detected, the messages can automatically be retransmitted in
accordance with the protocol."
--But this was (at the time the paper was written) still an experimental
work-in-progress:  "As will be seen below, work is proceeding on an
experimental network between the TX-2 computer at Lincoln Laboratory and
the Q-32 computer at System Development Corporation."
--"As soon as possible, a series of demonstrations and experiments will be
performed using the experimental network. The experience gained will be
reported at the conference."  [Was anyone at the 1966 Fall Joint?]

For more background on these early "networking" efforts I commend to you
the Hemmendinger paper from 2016.

John Shoch