Newsgroups: comp.lang.lisp
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!uhog.mit.edu!sgiblab!pacbell.com!amdahl!netcomsv!netcom.com!hbaker
From: hbaker@netcom.com (Henry G. Baker)
Subject: Re: working version of shrdlu
Message-ID: <hbakerCxvM1E.BAq@netcom.com>
Organization: nil
References: <37vao0$e5l@manuel.anu.edu.au>
Date: Tue, 18 Oct 1994 16:05:38 GMT
Lines: 46

In article <37vao0$e5l@manuel.anu.edu.au> Hugh Clapin <Hugh.Clapin@anu.edu.au> writes:
>	I am looking for a copy of Terry Winograd's shrdlu that is up and
>running on a reasonably accessible kind of machine (UNIX, Mac, PC etc). 
>The source code in old-fashioned maclisp is available by anonymous ftp,
>but I doubt I am enough of a lisp hacker to get it going on modern
>versions of lisp.  Terry Winograd mentioned that people have looked into
>getting it going:
>
>>There is no running version as far as I know.  The source code is
>available
>>via FTP on ftp://csli.stanford.edu/pub/shrdlu.tar.Z  If you're not on a
>>unix, we can figure out a way to get an uncomrpessed version.
>>
>>It is written in a dialect of Lisp (Maclisp) that was implemented on the
>>PDP6 and later PDP10.  I have had various inquiries from people
>interested
>>in converting it to a more modern LISP, but none of them have ever gotten
>>back to me with results.
>
>If anyone has any suggestions or information about the existence of a
>working version of shrdlu, the possibility or otherwise of making maclisp
>work (I have available to me Lucid Common Lisp and Austin Kyoto Common
>Lisp) or any other relevant help, please email me at: 

Shrdlu and Macsyma are nearly contemporary programs, so the work to make
shrdlu run is similar to that for Macsyma.  Getting shrdlu to run under
the pre-Common Lisp Berkeley Lisp should be relatively straightforward.
Getting shrdlu to run under Common Lisp would take more work, since a
whole bunch of things need to be declared as 'special' variables, and
a number of functions have changed names and/or calling sequences.

Shrdlu had a graphics front-end that would have to be mapped into some
standard graphics window system.  This part would take some work.

You should be aware that shrdlu never _really_ worked, in the same
sense that Macsyma did -- it was just a research prototype cobbled
together for a PhD thesis.  Therefore, substantial efforts would be
required to bulletproof it enough to make a museum-quality demo.

While you're at it, perhaps you would like to get Bill Woods' 'Lunar
Rocks' natural language program running, as well.  It used an early
version of what became Interlisp, and so would also require some work.

      Henry Baker
      Read ftp.netcom.com:/pub/hbaker/README for info on ftp-able papers.

