9 Apr
2021
9 Apr
'21
5:08 p.m.
On Fri, Apr 9, 2021 at 2:35 AM Rich Morin <rdm(a)cfcl.com> wrote:
There's a story I've heard about the SUN-1
board that I'd love to have
confirmed, etc. Basically, it says that Stanford wrote a letter saying
that they didn't make any claims on Andy's work (because he was only an
undergraduate, so how important could it be, anyway...).
Sounds a bit far featured since he did his undergrad at CMU ;-)
What Andy designed would (the best I can tell) have started as the 3rd
generation processor board for the CMU's distributed front-end (n-th
generation of the CMU front end system). The CMU terminal front end was
originally a single 11/20 with a lot of ASLI's (asynchronous line
interfaces) in them with a parallel connection to the PDP-10s or other
larger hosts. The problem was it did not scale and when then Unix
machines began to replicate around campus, having a terminal on more than
one host was needed, so the distributed front-end was created (by Jim Teter
I think). But we started to allow the separated 11/20s to talk to each
other. But when the LSI-11s and the Alto's from Xerox appeared a more
network-based distributed front-end was built using an LSI-11 chassis. We
used an early TCP draft for the protocols and was my introduction to that
protocol, as I was part of the crew that switched it to be a Multbus based
system and then an Intel 8085 and a Xerox 3M interface, plus an n-port
serial board because the LSI-1 based systems were fairly expensive. I've
forgotten now, but we might have had a Z80 based processor at some point,
as Phil Karn I know had a Z80 C compiler we were using too and the 8085
stuff was assembler if I remember right.
IIRC recall somebody at Stanford (Bill Yeager ??maybe??) was doing
something similar to the LSI-11 system we were working. How much was
Stanford first before CMU I can not say. The 11/20 FE did predate it all,
but the two LSI-11s were sort of parallel efforts. I Also thought MIT was
doing something ChaosNet around the same time, Noah can fill you in more I
suspect.
About 2 years later, Andy built a simple 68000 processor [using SUDS -
which was what CMU used for Designs in those days] for the multibus version
of the DFE, and at some point, somebody (maybe Andy) switched it to an
Intel Ethernet board. None of the Multibus or LSI-11 based DFE's had an
MMU associated with them. Andy did take took his 68000 CPU design with him
to Stanford when he was a grad student and famously redid it adding an MMU
and a lot of other features [i.e. CMU board != Stanford Board].
By this time the CMU 'SPICE' proposal had appeared and the idea of the
"3M"
workstation was being batted around. The Stanford Univerity Network
Terminal was created that took his reimagined CPU, the raster display, and
other features (I think he was able to get the ethernet on the CPU board by
then) -- note it still is using a Multibus-I was the backplane and memory
was on a separate board.
Stanford licensed the SUN-1 design to a number of firms and while the IP
was generally available it was licensed. Cisco made their first router
with it, which had a basic architecture that is not unlike the CMU-DFE.
Imagin used them for their laser printers. VLSI Technologies would be
found (and later renamed SUN) to make them
ᐧ