27 Mar
2020
27 Mar
'20
10:21 p.m.
On Fri, Mar 27, 2020 at 11:47 AM Robert Armstrong <bob at jfcl.com> wrote:
But that's another curious thing - there ARE
drivers included with RT11
for the DEUNA, DEQNA and the Pro series Ethernet NI. What was the point of
those? Did DEC think people would write their own networking software?
Absolutely. The project was completed after I graduated, but I helped to
design and develop the protocol for it. CMU's Mellon Institute built a
system originally for the Pittsburgh Press, to automate their newspaper
distribution system (they had had huge problems with theft). It was so
successful, it got used in about a number of large papers in the US and
Canada like the Philly Inquirer, Globe here in Boston, Chicago Sun-Times,
one in Montreal and maybe Quebec (I think it may have been as many at 25
installations). The system went through a couple of HW and SW
generations/refinements.
The original version for the Press had each station had an LSI-11 running
RT-11 doing some realtime control of the sorters and bundles - that sent
messages back to another mini over the network which was doing overall
control/scheduling but did not have to be RT. Originally, it used a
custom network solution over a CMU designed fiber-board, but eventually,
DEC released the ethernet stuff which was a lot cheaper. BTW: The original
worry at the time we were designing it was if something like ethernet could
work on a factory floor due to electrical noise *et al*. CMU was running
3M Xerox ethernet at the time to the campus-wide Vaxen and PDP-11s. But
getting the Xerox boards and transceivers was not easy.
The protocol was custom and similar to TCP without IP under the covers
(actually we borrowed from some earlier TCP work we did from the CMU front
end, but did have IP in it -- IIRC Don Gregg and Tron McConnell did the
heavy lifting in the end - Tron on the HW and Don on the SW). I'm pretty
sure that at one point the Mellon folks refined it later to swap out the
LSI's and switched to Multibus based systems for most of the later installs
in the early 1980s. I should ask him, but I think Tron ended up using a
Stanford University Network (a.k.a. SUN) board. But when we spec'ed it in
78 and did the original prototype work in the lab, it was on LSIs.
Much of the user-level app code was written in DEC FORTRAN, although the
system support stuff was assembler and C. But, IIRC, it was a customer
requirement as the Press did not want to use that funky C thing -- they
knew they could get FORTRAN programmers for later maintenance.