T sources and manual

[Jonathan A Rees <JAR@AI.AI.MIT.EDU>]

    Date: Wed, 8 Jun 88 16:33:56 -0200
    From: "Anton Hartl" <unido!tumult!hartl@uunet.UU.NET>

    1. Is it possible to build a T system without having a running one?

Is it possible to build, say, a GNU C compiler or PCC without a having a
running C compiler?

       We recently obtained the *sources* for T3.1 from wheaties.ai.mit.edu,
       and realized that it is not possible to build a T system solely from
       the sources.  So we had to get the vax-specific image to get a
       running T system.  Is this a bug or a feature?

Feature.  It means we didn't have to waste time writing an interpreter
in some low-level language like C.  We knew a native code compiler was
going to be needed in any case, so we bootstrapped using that.

       Where did the implementors of T start?

We started from PDP-10 Maclisp.  The T-in-Maclisp emulation probably
only exists now on DUMPER tapes in my office (and possibly Norman
Adams's), since Yale has discarded its TOPS-20 machines.  And it
wouldn't do you much good, since it only ran old versions of the old T
compiler (2.3?) which as far as I know were discarded long ago.  The
compiler for T 3.0 is a total rewrite.  So the whole process would
probably take you four or five iterations, even if you had all the
hardware and sources.

       I assume there must be some kind of T interpreter written in some 
       high level language like C (and NOT T); this interpreter could be
       ported more easily to different kinds of CPUs as a starting point
       for the complete T system.  
       Is it possible to make this interpreter available?

No such beast; I agree that the situation is lamentable.  I have often
considered doing a T emulation for MIT Scheme or Common Lisp, but never
felt motivated enough.  It probably wouldn't be too hard (although some
people at Cognitive Systems, Inc.  tried and couldn't get reasonable
performance).  T in MIT Scheme (as a macro package plus runtime library)
would be particularly useful since the dialects are very compatible and
MIT Scheme has an interpreter written in C.

    2. The second problem concerns the T manual.  
       The manuals we obtained are almost all in scribe format, a format
       we unfortunately are not able to deal with.  Are there T manuals in
       TeX format?  We tried the scribe2latex translator,
       but it catches only 90% of scribe directives.  Would you
       like to wade through 330KB of text to fix the remaining 10%?
       Perhaps the T manual of the distribution itself could be
       transformed to TeX format, since TeX is probably more likely
       available than scribe.

Any volunteers?  This would be a service.  I don't know of anyone who
has the time.

Sorry... you get what you pay for...