Unix History Diagram --- AOS quirks

[User Rdkeys Robert D. Keys]

> > A couple of lesser known BSDish oddities from Big Blue....
> > 
> > Add IBM's AOS.  That was a straight 4.3BSD port dating from 1987 or 1988.
> > This was the only official port for the RT-PC hardware.
> 
> I have a paper about this port which claims that it's in fact tahoe or so,
> and has an independent implementation of mmap().  I'll try to dig it up --
> 
> I think it was in one of the old Waite Group books.

Please do!  I would like to see that.  I did see at one time Tahoe
mentioned, but I did not understand how the IBM ports were related
through that.  There are so few folks around that know anything about
these IBM critters, even on the old RT newsfeed.  Most of the stuff
has become dumpster fodder, sadly, although I had the good fortune
this past week to resurrect two RT's from the dumpster, and get one
up by combining sufficient parts to get it to boot.

What specifically would one look for to exactly differentiate
a vanilla 4.3, from a Tahoe, from a Reno, from a 4.4, from a 4.4Lite,
on non-standard hardware?  The books get somewhat obscure on this
unless running VAXen or HP300's or such.  The RT is a little bit
non-standard.  I was thinking it was a 68000ish machine in IBM's
wrappers, but others have said it was distinctly different from a
68000 based line.  Also, I can't find anyone that was in on the
AOS project enough to know from whence it was originally derived.
The dating seems to be 1987 or 1988.  Was Tahoe around then?
Salus suggests straight 4.3 was June of 1986, and Tahoe was June
of 1988 (Salus, p. 165).  Anyone around CSRG then that remembers
when IBM got what code?  Salus does not mention any IBM AOS stuff,
only the mainframe stuff.  AOS seems to be mostly a sleeper, almost
forgotten in time.

.....

> > Add IBM's unoffical ``4.4Lite'' port.  That was a somewhere between 4.4
> > and the real Lite dating from 1994 or 1995.  It was apparently done by
> > IBM or IBM contractors, with source trees from two development streams
> > combined together to resolve developmental divergences.
> 
> Someone here had a tape of this, yes?

There is a Finnish repository that has some of it relating to the 4.3
port, and one or two other RTish archives.  Try something like jumo.luti.fi,
or jumi.luto.fi, or something like that, but I don't have the url exactly,
and can't find the stick'em note where I ran across it.

Yes, there was an IBM'er that said he had some original tapes.  I was
hoping he would check with someone at IBM to see what the status was.
There was a group at Carnegie-Mellon that had some machines with AOS,
but I don't have any pointers to anyone up there, for sure.  I would
hope the old RT boxes and AOS would fall under the Antique Unix umbrella,
and thus, be amenable to the PUPS archives scope of things.  But, I
am only a newbie voice in the crowd.....

>From what I have found out, there were two install tapes and two boot
floppies.

The main boot floppy is the sautils disk (stand alone utilities).
That then loads a miniroot floppy that does the scripted install.
The scripts don't work unless you have the original tapes, and
the orignal hardware configuration which was a pair of 70mb esdi
drives.  The installation needs to be done manually, instead, if
the hardware differs from that.  It should be a straight 4.3 style
restore process.  I guess they expected only to have dual 70mb
esdi drives in the old RT tower machines, as AOS platforms.

The first tape is the combined root and user dumps to hd0a and hd0g.
The second tape is the source tree dump to hd1g.

> The AOS releases I saw all came with source.  I wish IBM would donate the
> bits they wrote -- much information on ROMP processor bugs, etc. simply
> doesn't exist anywhere else.

That is how I understand it.  There were some manuals for it, but
noone seems to know anything about those anymore, and what info I
have is more sketchy than for sure.  There was also some other machine
called an ``Academic Machine'' that was a siamesed ROMP processor
on a Model 60 PS/2 MCA bus machine.  I have not exactly understood
how that thing actually worked, and noone on the net seems to have
one, although there are two ROMP boards that are reputed to still
exist that plug into the MCA Model 60 PS/2 box.  Apparently the
Model 60 was the terminal/disk IO system and the ROMP board ran
the BSD.

> Patches to fix this circulated widely -- I recall one _very_ late night
> in Bill Cattey's office at Athena waiting for someone in Stockholm to
> send us a patch so we could make some tapes I needed the next morning.

The thing will tar/dump to the tape, but won't retension or erase
correctly.  If you know what that patch was, I would be interested
in it, for sure.  I assume it has to do with something like twiddling
the right hardware ports with the right bit patterns, maybe, to run
the retension and erase functions on the hardware.  I am curious,
though, and wonder if something like the 4.3 mt, which does work,
would work correctly in the Lite suite.  I was of the impression
that there was some binary compatibility between 4.3 and 4.4/4.4-Lite,
but I am not sure.

.....

> > The 4.4Lite is fair to good, but still missing working tape I/O.
> > Code size about 300 megs binary.  It needs 16M ram and 300mb HD.
> > Bloat seems to have set in on this one, since the whole system is
> > well over 1 gig in size.  It barely will run on two 300mb HD.
> > The login says it is 4.4Lite and not straight 4.4.  There is no
> > indication of how pure ``Lite'' it really is.
> 
> Wow, I wonder if it really is "Lite".  Again... I wish IBM would free
> the relevant bits, we've wanted for a long time to make NetBSD run on
> these beasts.  One major obstacle is that nobody at IBM seems to even
> know where the "official" sources are, or who would have authority to
> turn them over.

I have no idea how pure or impure the code is.  I came along so late
in the song and dance act that I don't know enough of the internals
to compare, yet.  So much to learn.....  Years ago I bounced this
off our IBM rep, but went to AIX on the PS/2, instead of the RT BSD.
I did not know very much then, nor now....(:+{{.....

> > I dunno anything about how these originated developmentally, but
> > the AOS seems to be vanilla 4.3BSD and all else may have developed
> > from that, possibly after the RT line became back-burner stuff.
> > I would be very interested in any history from anyone on the list
> > that was around IBM at the time on these.
> 
> Me too.

Who all on the list were IBM'ers in that era that might still remember
enough of this to fill us in?  The history is half the fun, and sure
makes the perspective on the rest more interesting and well rounded.

Anyone on the list actually playing around and running an RT?
I am beginning to feel very lowendian that I am not on a PDP11, VAX,
HP300 or such.....{:+{{...   but, a VAXStation 3500 just appeared
in surplus.... maybe the bidding force will be with me.

Bob Keys


Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.1/8.9.1) id GAA20122
	for pups-liszt; Wed, 16 Dec 1998 06:15:11 +1100 (EST)