Preserving a Multics System

From: Douglas Quebbeman <dhquebbeman_at_theestopinalgroup.com>
Date: Sat Oct 21 11:13:53 2000

> > > So is anyone archiving the software for this Mulitcs system?
> >
> > An archive of Multics exists at MIT. If my recent personal
> > experiences with 9 track magtape can be considered representative
> > and not anecdotal, then I think the tapes will still be readable
> > even ten years from now.
> >
> What kind of 9-Track tape drives can read these tapes though? IIRC, the
> drives we had on the DPS-8's were pretty much bog-standard 9-Tracks (I
> remember writting a tape to be read on a Sun box).

Well, these tapes were written by Multics tape_archive command
(the granddaddy of tar). Back in '93, someone at MIT used what
I presume was tar on some Un*x to restore the archive to the MIT
Andrew File System. However, tar seems not to have handled the
Multics symbolic links occuring on the tape before the objects
to which they pointed, so not everything got restored.

In particular, the portion of the archive that should have
included the running binaries was not restored. This would
mean recompiling the system and generating the boot image,
but there's a worse problem...

The most serious problem with the archive, though, was that MIT
never had the source to the later Multics PL/I compiler, though
they did at one time have the source to the earlier one. But of
course the existing Multics source relies on the later compiler.

So, assuming the restored archive contains all the pieces of the
OS at the time the snapshot was taken, you'd still need to get
or reimplement a PL/I compiler for the Level 68 architecture
(a plain DPS-8 is a Level 61, IIRC, while the DPS-8/M is a
level 68 machine, the original implementation of which was for
the Honeywell 6180 machine).

Ultimately, for a proper preservation, permission would have
to be secured from Bull to run the OS, have the source, etc.
What exists at MIT is valuable for some research on how a
Multics was put together, but as a basis for a museum system
would leave many many man-years of reprogramming left to be
done. As a programmer, for me, that would be the fun part.

But I intend on contacting TPTB at MIT to see whether I
could donate the labor to attempt a second, more proper
restoration of the archive, by using GNU TAR and making
whatever changes need to be made to it so that it does the
job right. Current plans are to contact them this time
next year.

-dq
Received on Sat Oct 21 2000 - 11:13:53 BST