Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!cbatt!ihnp4!inuxc!pur-ee!uiucdcs!uiucdcsp!forbus
From: forbus@uiucdcsp.UUCP
Newsgroups: comp.lang.lisp
Subject: Re: LISP for IBM compatible PC's
Message-ID: <80300002@uiucdcsp>
Date: Wed, 28-Jan-87 12:41:00 EST
Article-I.D.: uiucdcsp.80300002
Posted: Wed Jan 28 12:41:00 1987
Date-Received: Fri, 30-Jan-87 02:40:25 EST
References: <1048@sfsup.UUCP>
Lines: 44
Nf-ID: #R:sfsup.UUCP:1048:uiucdcsp:80300002:000:2409
Nf-From: uiucdcsp.cs.uiuc.edu!forbus    Jan 28 11:41:00 1987



>The February issue of AI Expert begins a three-part wrap-up of microcomputer
>LISPs; both the series and I really see no alternative to the Gold Hill
>products if you want a quality development environment where the editor
>really supports the process. Caveats: dynamic scoping in the PC version's
>interpreter, one-dimensional limit on arrays, limited floating-point on the
>PC version.

If it isn't lexically scoped it isn't Common Lisp.  GEMACS, on both GCLISP
and GCLISP-LM, is a complete dog.  Anyone who has used a serious lisp
machine will slowly go mad if forced to use it for any length of time.
Frankly, I would not recommend running GCLisp on a PC or XT at all.  It's so
slow it is frustrating.

If you aren't going to use Common Lisp, then I think TI's PC-SCHEME is
clearly the implementation of choice.  Their editor, while no speed demon,
is better than Gold Hill's.  Split-screen mode (one half lisp, one half
editor) is a real win.  Their compiler works on the PC, and works well
(including tail recursion optimization).  Bignums are suported.
Performance, given the machine limitations, is quite good:  TI-Scheme on an
XT is more fun to use than GCLisp-LM on an AT.  At $100, TI has really done
the world a terrific favor.

>The AT-specific version, though, is a thing of beauty that is
>making our Symbolics users' eyes bug out.
>						Cheers, Peter C.
>/* End of text from uiucdcsp:comp.lang.lisp */

This is a joke, right?  I've used Symbolics machines for years, and I've
also developed over 200 pages of code for instructional purposes to run on
our lab of 32 AT's running GCLisp-LM.  While GCLisp-LM is fine for many
instructional purposes, I could not recommend it for serious research or
development.  Compare:  you can put 16MB of RAM on your AT, or you can put
16MB of RAM on your Symbolics machine AND have 200MB of virtual memory.
Guess which computer can run larger programs and horse around more data?
Not to mention the relative quality of the programming environments...

GCLisp-LM is a fine product, even though it has numerous bugs and glitches.
But people are fooling themselves if they think an AT is sufficient for most
AI work.  You may wonder why Gold Hill is way overdue on their PC version
with lexical scoping, and is spending all their time on the 386 version.  If
I were them I'd do the same thing -- the 386 is closer to being a machine
that can run Lisp well.