[TUHS] Sandy Fraser on Spider and Unix, 1975

Paul Ruizendaal pnr at planet.nl
Fri Nov 10 21:11:03 AEST 2017


I happened to come across a 1975 report from the University of Warwick
which includes a section on the state of computer networking.
(https://stacks.stanford.edu/file/druid:rq704hx4375/rq704hx4375.pdf)

It contains a section that appears to be a summary of a chat with Sandy
Fraser about Spider (pp. 66-69). It has some information on Spider network
software and Unix that is new to me, and I find amazing. I had not expected
some of this stuff to exist in 1975.

Below some of the noteworthy paragraphs:

[quote]

Spider is an experimental packet switched data communications system that
provides 64 full-duplex asynchronous channels to each connected terminal
(= host computer). The maximum data-rate per channel is 500K bits/sec. Error
control by packet retransmission is automatic and transparent to users.
Terminals are connected to a central switching computer by carrier transmission
loops operating at 1.544 Mb/s, each of which serves several terminals. The
interface between the transmission loop and a terminal contains a stored program
microcomputer. Cooperation between microcomputers and the central switching
computer effects traffic control, error control, and various other functions
related to network operation and maintenance.

The current system contains just one loop with the switching computer (TEMPO I),
four PDP-11/45 computers, two Honeywell 516 computers, two DDP 224 computers,
and one each of Honeywell 6070, PDP-8 and PDP-11/20. In fact many of these are
connected in turn to other items of digital equipment.

Spider has been running since 1972 and recent work has shifted away from the
design and construction of the network itself to developing user services to
support other research activities at Bell Labs. A major example of this has
been to construct a free-standing file store (extracted in fact from UNIX [88])
and connect it to the network. This is available as a service to any user of
the network.

[...]

The ring is used in different ways by the various computers connected to it.
The filestore has already been mentioned. Two computers use this for conventional
back-up, and access it on a file-by-file basis.

Two other machines - dedicated to laboratory experiments - access it on a
block-within-file basis. To help with program development for these dedicated
machines, the UNIX system (available on yet more computers) is used during
"unavailable" periods for editing and program preparation. The user then leaves
his programs in the filestore ready to load when he next gains access to the
dedicated machine.

Two other "dedicated" machines provide the user interface of UNIX, but lack
peripherals and a UNIX kernel! In place of both is a small amount of software
that transmits all calls on the UNIX system to a full UNIX system elsewhere!
The ring system with its filestore also provides a convenient buffering service.

Finally Fraser tells of the time where one of the PDP-11 machines was delivered
sans discs. A small alteration to a UNIX system diverted all disc transfer
requests to the filestore, where a suitable amount of disc was made available.
The system ran a full swapping version of UNIX at about a quarter of the
normal speed.

[/quote]





More information about the TUHS mailing list