IP: more on DCS discussed

[Dave Farber <dave () farber net>]

Frank was a grad student at UCI during the DCS effort. Djf

------ Forwarded Message
From: "Frank Heinrich" <frh100 () earthlink net>
Date: Sun, 12 May 2002 15:45:05 -0700
To: "Karl Auerbach" <karl () cavebear com>
Cc: <dave () farber net>
Subject: Re: DCS discussed

Karl,

thanks for copying me on this...It was fun   and its easy to get
nostalgic....(even for debuggin code in hex through  front panel lights and
switches).

Dave,

As I recall, the hardware part ("Token Ring") was operational early in the
70's (probably 73, but maybe late 72??, as  we were publishing  theoretical
design papers in 72).   I left UCI in early 75, and the hardware things had
been working solid for quite a while, and the software was also at some
level of functioning long before I left.

I can't remember if it was summer of 73 or 74 that Larry and I got the
distributed kernel and basic user processes going (I still remember the late
night in the basement when we got rid of the "last bug "  that was crashing
the kernel on a single CPU, and agonized over whether we should go home with
that victory in hand or recompile and try it on multiple processors.  We
decided to go for it, and within 20 minutes, we were up (albeit limping) on
two CPUs, dramatically validating the architecture and design.  Few things
since have been as big a "high").

So basic DCS OS software was working in 74, and maybe earlier, although it
continued to evolve through the mid-70's.

as we are "clarifying " the records, you might want to also mention that DCS
had what may have been the first of what we would call today a microkernel,
and that components of the OS outside the microkernel could be distributed,
as well as "user" processes.  Also, the architecture Larry and I
implemented for some system and user process interactions was the first
instance I saw of something that I later recognized as Remote Procedure Call
(its possible we just weren't paying enough attention to someone else doing
it and had "re-invented" it.  But who else was actually *doing* distributed
processing as cooperating, message-linked distributed processes?).

and as a further point of reference, the basic seeds of the strict
message-based, layered OS concepts were influenced by some theoretical work
of Per Brinch Hansen, although as I recall, that work was not a distributed
architecture.  I remember bringing a paper of his to the attention of the
project, and arguing for its adoption as our basic architecture for the DCS
OS (it wasn't much of an arugment, the fit was so obvious). But that was the
seed that pressed us to a strict message model, not just between machines
but even within the OS, and that led to the microkernel and distributable OS
components.  It was an obvious fit with the process-based addresing and
communication that had been the cornestone of the basic DCS concept.  We
were simply carrying it one step deeper into the internals of what ran on
each computer.



--frank

------ End of Forwarded Message

For archives see:
http://www.interesting-people.org/archives/interesting-people/