Re: The Big War: Symbolics againts Unix boxes

[Brad Miller <miller@CS.ROCHESTER.EDU>]

OK, more (indirect) symbolics bashing, but with a caveat: I don't want
anyone to think that I don't like symbolics machines. I have one on my desk.
I'm getting a UXL400 soon. I think they are far and away the best machines
to develop lisp code on. Truely a Rolls Royce. But a Rolls Royce in a world
where most people only want chevys, and (worse) one in which their edge (ie.
their environment's superiority to lesser makes) is steadily eroding and I
DON'T SEE SYMBOLICS DOING ANYTHING ABOUT THIS. In fact the meeting notes of
15 Dec. indicate Symbolics has no intention of improving their environment
beyond whatever they provide in 8.0. Terrific. There are no possible
improvements to be made in the way developers write lisp code. Sorry, I
don't believe it.

Worse, I'm the one who gets hammered on here in this predominantly UNIX
environment, having to justify expendatures at every turn. I'd love to hear
good arguments for continuing to specify Symbolics. For a long time, it
really was the only thing I could build and maintain our KR system on, so it
was easy. That's no longer true. So what quantitative arguments are left?
Not speed. Not cost. Productivity is a very slippery thing, which I've yet
to see any concrete scientific data on. E.g. j. random luser being able to
write GNU from scratch on a lispm in the same amount of time RMS can do it
on a 4/330. (which would be a pretty impressive argument if true :-).
Claiming *I'm* more productive on one works to some extent, but that's
religion, not evidence. (Luckily they want to keep me reasonably happy :-).

So, the arguments I use in this forum are no more or less than those I have
to face (or expect to face) every time I have to say yes, we really should
continue paying x thousand a month to maintain our 40s, or to upgrade or
whatever. If anyone can supply me with stuff I can take to the bank and use
to continue to spec and keep Symbolics machines on my desk and others, it
would be greatly appreciated. But the report from this front is: I'm running
out of ammo. Maybe that's Symbolic's marketing fault, but somebody needs to
help.

    Date: Wed, 17 Jan 90 14:15:55 N
    From: baechler%elma.epfl.ch@CUNYVM.CUNY.EDU

      There are a lot of things which make the difference. The simplest one is
    that you lose your whole environment each time you log out.

Don't log out, or disk-save an image of the stuff you load. Booting my
symbolics and reading in all the files I hack into the editor takes over an
hour, and I have to reboot pretty often (because I'm using PCL which isn't
all that robust :-( ); with luck this will improve in 8.0. The point is that
yes, having an (login independent) environment is very nice, and yes, too
bad UNIX vendors don't realize their workstations are single-user machines,
and yes, this is a nice point in Symbolics favor, but how much money is it
worth? Given nobody is forcing you to log out of your UNIX machine. (if they
are: compare apples with apples: someone ELSE logging into your symbolics
may well need to reboot. One hacker's environement is another luser's
nightmare.

    each morning takes time. Allegro itself is ok, but when used with GNU, the
    whole set becomes really slow: a lot of time is spent paging. When, in
    addition, X11 is used, it becomes a real catastrophe.

Forget using an 8meg machine; you really need 16-24 meg. (8 meg is barely
enough to use X, but you could X to a 24 meg machine and run allegro
there). Conversely, my 3620 running Zmacs and an application is a bit
painful too. Probably because I only have 2mw on my 20. (addressed by
ordering a uxl400 with 20mb :-).

    All the tools integrated in the composer are also
    really primitive, when compared with Symbolics.

True, but this is improving; more importantly Franz and Lucid both are
working hard to make their products more like a lispm; Symbolics isn't
really doing anything to stay ahead. (sigh. I really wish they would. I
really would like to be able to justify continuing to buy and maintain
symbolics machines in our lab, but the arguments get harder every year).

    The availability of the sources
    in Lisp machines is also very useful for debugging purposes because it
    frequently helped me to explain why my code was false.

Franz will sell you the sources to Allegro.

      People here also had problems when developping software using dynamic
    windows but it however remains that they are really convenient and that
    you can rapidly build a user interface as long as you do not want to use
    their tricky capabilities. When compared to X11, I would much prefer
    to use program frameworks and dynamic windows than trying to program
    X11.

Absolutely!!! BUT once CLIM is supported under Allegro (march 90 last I
heard) you can use that; you'd probably want to port to it on your Symbolics
once it's available anyway (from DW).

      Thes UNIX boxes are also not so cheap because you must add a lot of memory,
    big internal disks and a disk server to make them usable.

True, but you don't have to do that to every machine; you can have a few
mega machines with the memory and disk to run one or more lisps, and let
users with cheap sparcstations X to them. The other consideration is system
administration too. If you already have UNIX boxes there is no additional
staff overhead. If you are an educational institution, SUN, DEC and others
typically heavily discount hardware, MUCH more than Symbolics does too, so
that drives the equations a bit as well.