[TUHS] Anyone ever heard of teaching a case study of Initial Unix?

Clem Cole clemc at ccc.com
Tue Jul 9 03:04:42 AEST 2024


Al is right. Tom West led the Eagle project in Westboro, MA, which is
documented in the Soul of the New Machine. The Eagle project became the
32-bit MV/8000 Eclipse [see https://en.wikipedia.org/wiki/Data_General for
more details].   I knew a few of the HW guys, as I went to CMU with one of
them, and a couple of them came to do the Stellar CPU a few years later.
But I did not know the SW folks there like I know the DEC folks,
particularly since I never worked there.

That said, WRT to their later UNIX box, I did have access to the same at
Locus.  As we did a lot of work for DG, adding features and helping them
with POSIX/FIPS and SPEC1170 conformance — IIRC, we added the FS Switch for
NFS support.  It was probably the easiest UNIX kernel I have ever worked
with, with the shortest learning curve. I may remember that the User API
was based on SVR3/SVID, which means >>IIRC<<; it was based on Streams/TLI,
but ISTR is a user mode sockets layer for porting (but I'm not sure of that
-- it has been almost 30 years now).    The kernel itself was a scratch
rewrite with SMP in mind. The locking scheme was clean and simple and
worked well to order 32/64 processors in our tests.  Moreover, the kernel
was well-documented. I do not know for sure, but I remember being told by
some of the DG folks in NC that it was originally planned for the failed
Fountainhead system, which was canceled after West and the MA-based folks
delivered Eagle before the new system came from DG NC.

Also, I believe that DG had in response to BLISS a low-level system
language they called PL/N, but I don't know much about it/I never saw it -
I'm told that it was a >>very<< subset PL/1 syntax, but like PL/360 -
exposed a lot of hardware.  ISTR that they developed some MV series
compilers with it, but since Eagle was a 32-bit super set of Nova, AOS was
ported.  FWIW:. Since Multics had used PL/1 and supposedly Fountainhead was
heavily influenced by Multics (as was Pr1me, by the way), it would not
surprise me that PL/n has a PL/1 'flavor.'

I'd love to know for sure >>but my WAG<< is that that core work for
Fountainhead was reimplemented in C for their SMP 88000 box, and they added
a UNIX API layer to it.


ᐧ

On Mon, Jul 8, 2024 at 9:22 AM Larry McVoy <lm at mcvoy.com> wrote:

> On Mon, Jul 08, 2024 at 12:59:14AM -0600, arnold at skeeve.com wrote:
> > Brad Spencer <brad at anduin.eldar.org> wrote:
> >
> > > The later MV/xxxxx Supernova boxes could run Unix, I
> > > believe... (at least I remember the university running Unix on a MV
> > > series after I left).
> >
> > I think these were called "Eclipse", and the story of their
> > development is told in the famous book "The Soul of a New Machine".
> > For you youngsters out there, it's a great read.
> >
> > We had one at Georgia Tech, it ran a Unix emulation on top
> > of AOS (or whatever it was called). Later on DG ported Unix to
> > run on it native.
>
> I've heard, but never verified, that they did a really nice SMP Unix.
> If anyone has seen the code I'd like to hear what you thought.  The
> way it was described to me, it sounded like an SMP SunOS.
> --
> ---
> Larry McVoy           Retired to fishing
> http://www.mcvoy.com/lm/boat
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.tuhs.org/pipermail/tuhs/attachments/20240708/bb5512a3/attachment.htm>


More information about the TUHS mailing list