The feel of a LispM/List of running machines

John Casu John_Casu@NeTpower.com
Wed, 30 Apr 97 16:05:00 PDT


Wrt filesystems, you might want to take a look at the SHORE persistent   
object store,
developed at UW-Madison  (http://www.cs.wisc.edu)

I think this might provide the basis for the filesystem/persistent store   
functionality
you want, combined with the ability to interface to the external world   
via NFS.



 ----------
From:  DF[SMTP:DF@GLIDER.MACLAB.LONESTAR.ORG]
Sent:  Wednesday, April 30, 1997 8:36 AM
To:  cracauer; lispos
Subject:  The feel of a LispM/List of running machines


    Date: Wed, 30 Apr 1997 04:16 CDT
    From: Martin Cracauer <cracauer@cons.org>

    It seems we have a lot of old LispM users over here.

The idea of a Lisp OS we can run when we try to use inferior computers
attracts us like a lurid picture show attracts teenagers.

    I think it would be a good thing if you could write a short
    description of the LispM features you liked, what features turned out
    to be useful in practice and what could be left out. Don't think that
    is obvious, I bet even the former longtime LispM users have different
    opinions here. Not to speak of the Interlisp folks.

I think one of the foremost LispM features is the debugger system.  And
this also is an answer to something you asked in a different message re:
what can you do in a pure lisp system that you can't do with one running
under Unix?  With a VLM running under Unix, you are completely hosed if
something in the Unix system breaks (a disturbingly frequent event!).
On a pure lisp system, if something goes wrong anywhere in the OS, even
in very low level parts, you get a Debugger from which you can most
likely fix the problem and proceed, or at very minimum get a clear idea
of what happened.

Unix's file system is a total loss.  We need things like version
numbering, ACL's, two-level file deletion (delete/expunge), file-plists
for arbitrary user-defined properties, Don't-Delete flags, logical
pathnames, etc. Even more important than the above: robustness.  All of
these and more are indispensible LispM features.  (Yeah, I know my
earlier message derided the entire concept of "File systems", and I
believe it; but I also realize that what I advocate is in the future,
and we need to start simple).

I won't discuss all the hairy window system stuff, since CLIM handles
most of that adequately.

The who-line (mouse documentation, date/time, running process,
current syntax and package, process state, ->PROGRESS NOTES AREA<-, run
lights).

The ability to grab and mutilate -ANY- part of memory simply by knowing
its address -- so useful I defined a readmacro #@ to make it extra easy:
; grab the clock string from the wholine
#@17000014 ->
"Wed 30 Apr 7:46:05"
#<LOCATIVE 17000014>
12

I'm getting a little tired so I'll stop here.  I think I covered some of
the really critical stuff though.

LispM bugs:
LMFS has tons of really cool features, but in practice it's fairly slow.

*NO* internal security (not a bug for the LispM, but would be a bug for
an OS intended for broad consumption, and especially a bug for an OS
intended for multiple simultaneous users).

Draws 10 amps, dissipates ~3000 BTU, weighs 250 pounds.  (Approximate
figures for a 3640).

    We need to focus this project some more. The whole discussion about
    Unix-kernel based or pure LispOS shows that it is not clear what
    features of LispM we really want.

We want something that can both replace Unix in a relatively short
development time, and be the ultimate system on stock hardware.  These
are wildly diverging goals, but hopefully goal/1 will lead into goal/2
after the project has gained sufficient momentum.  Somehow we need to
find a way to develop the system in such a way that after goal/1 is
attained it won't mind having unix ripped out from under it and a real
OS put it its place.  Am I off-base here?

    Also, it is probably a good thing if everyone here gets the change to
    play with such a machine.

It would be a very good thing if everyone in the world had a chance to
play with a LispM.  Maybe in a few years they'll have a chance at the
next best thing.

    Maybe we should arrage a list of machines that can be seem and used?

My machines are available, for those in the North Texas area (or those
who don't mind a long drive).  There are still dozens of universities
that have LispM's laying around (usually in deserted labs).  I know MIT
still has half a dozen or so up at the AI Lab.  I think a fair number of
interested people could get away with just walking in and sitting down
at one for awhile (this is how I got my first exposure to a LispM: a TI
Explorer at SMU, where I was not a student; I spent weeks with it and
noone ever harassed me or even talked to me).

    To start, I'm in Hamburg/Germany and would welcome visitors to show
    them how dark the lights get when I turn on my 3600 :-)

I have a spare Cypher drive and a cooling turbine for a 3600, but no
3600 to put them in.  I also have a Hawley mouse with the old 3600-style
D-shell connector and a couple of wire-wrapped IO.2 boards.  If you are
interested in the tape drive or the mouse, mail me.  The turbine and
IO.2 boards are MINE ALL MINE! =)

    I also have manual sets I could lean to people and maybe organize   
sets
    to give away. Maybe someone could get through the hell of scanning
    these in. I already found them to be very OCR-friendly :-)

Hmm, you might want to get SMBX's approval for something like scanning
in manuals.  I've been wondering if they'd go for putting Release 6 into
the public domain (for non-commercial use), perhaps now is the time to
find out.

An alternative is to find the MIT LispM manuals and scan -those-
(assuming there aren't any hairy restrictions on them, which I think is
a reasonably safe assumption.  Anyone know for sure?)

Ok, now I'm getting quite tired, so I'm stopping for real this time.