Newsgroups: comp.lang.scheme
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornell!murthy
From: murthy@cs.cornell.edu (Chet Murthy)
Subject: Re: United we stand... (was: GNU Extension Language Plans)
Message-ID: <1994Oct25.204549.1480@cs.cornell.edu>
Organization: Cornell Univ. CS Dept, Ithaca NY 14853
References: <KAELIN.94Oct24122409@binki.bridge.com> <SCHWARTZ.94Oct25124554@roke.cse.psu.edu>
Date: Tue, 25 Oct 1994 20:45:49 GMT
Lines: 34

schwartz@roke.cse.psu.edu (Scott Schwartz) writes:

>kaelin@bridge.com (Kaelin Colclasure) writes:
>   In addition, Scheme-48 has some characteristics which would make it a
>   nicer base for GEL (hey, that's not bad...) than SCM.

>It also has the serious disadvantage that it is humungous.  On a
>sparc, requesting minimal heap, it consumes about 4M of data space.
>In a shell or lightweight extension language that kind of bloat is
>simply absurd.  You might as well use emacs.

>For comparison, perl5 runs in about 500K, SCM in about the same.  

SCM also has a copying GC.  That's 2x right there.  On the other hand,
the technology they use for their virtual machine is space-age -- they
generate the virtual machine from scheme code written in a subset of
Scheme called PreScheme.

As a result, there is also a mark-and-sweep-GC version of scheme-48,
which is used on (*shriek*) standalone robots (!!!)

Yes -- you read that right -- standalone robots.  With (if I remember
right) less than a MB of memory, total.

It's pretty neat stuff.  Those of you who haven't checked it out
should look at the paper:

           Program Mobile Robots in Scheme

at the scheme-48 WWW page:

    http://martigny.ai.mit.edu/~jar/s48.html Tue Aug 23 20:44:02 1994

--chet--
