Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!think.com!paperboy!osf.org!coren
From: coren@osf.org (Robert Coren)
Newsgroups: comp.unix.wizards
Subject: Re: Why not Multics? (was Re: BSD tty security, part 3: How to Fix It)
Message-ID: <21579@paperboy.OSF.ORG>
Date: 1 May 91 14:03:53 GMT
References: <542@trux.UUCP> <1991Apr30.142053.2313@sctc.com> <3096@cirrusl.UUCP> <Z}F_HC=@uzi-9mm.fulcrum.bt.co.uk>
Sender: news@OSF.ORG
Organization: Open Software Foundation
Lines: 63

In article <Z}F_HC=@uzi-9mm.fulcrum.bt.co.uk>, igb@fulcrum.bt.co.uk (Ian G Batten) writes:

[a pretty accurate description of why Multics was never a major
commercial success.]

As a member of the Multics development team for many years, I can give
a bit of an "inside" view. Multics was, in fact, regarded by most of
Honeywell's marketing as competition for GCOS (which had a much larger
customer base), and, especially in the early years (1973-78 or
thereabouts) a prospective customer had to push pretty hard to get
someone to sell him one.

|> It was always said that the Multics and GCOS people in Phoenix barely
|> spoke to each other, and GCOS had the ear of the management.  It always
|> needed hacked hardware (ie DPS8/M rather than DPS8).

Yes, each of the various machines that ran Multics at one time or
another was a "modified" version of the then-current GCOS machine, and
it often took a while to get the modifications shaken out. There was a
project (or at least a "study") started in about 1983 to design and
build hardware specifically targeted for Multics, but its funding was
cut before anything real happened.

|> > More specifically, where can we buy Multics to run on our favorite
|> > hardware?  Why can't we buy it?
|> 
|> I know that someone (who I suppose should be named by those who know the
|> current position rather than me) tried to buy Multics from Honeywell. I
|> never heard quite what the outcome was, or what he was planning to do
|> about hardware.

There were a couple of attempts to interest Honeywell (and then
Honeywell Bull --> Bull HN) in a deal whereby an entrepreneur would
take it off their hands and concentrate on selling just Multics, but
nothing ever came of it. Honeywell was never willing to completely let
go. [It's not entirely clear that such an effort would have been
commerically viable in any case.]  There are a few Multics systems
left in the world, and Bull has contracted with a company in Calgary
named ACTC (Advanced Computing Technology Centre?) to provide
maintenance (I'm a little hazy on the details of this deal). I don't
think you can buy it from Bull anymore (they've stopped making
DPS8/Ms); maybe you can pick up a used one at a "distress sale" :-).

|> I heard theories as to how much magic would be needed to make it run on a
|> modified 386 platform, but there would be an awful lot of FIXED BIN(25)
|> declarations to change.
 
I assume that's a typo for "fixed bin(35)"; likewise "fixed bin(17)",
"bit(18)", "bit(36)", etc. Multics ran on 36-bit hardware, and in fact
was not designed to be portable; there were a number of explicit and
hidden assumptions about the capabilities of the hardware. Making a
portable version of Multics would have been a *big* job. Even simply
porting it to some specific 32-bit machine would have been a major
undertaking.

|> 
|> Sad day when I stopped using Multics.  Sad day.
|> 

Indeed. But, as it turns out, there *is* life after Multics (something
I might not have believed 7 or 8 years ago).
	Robert