Date: Wednesday, 25-Jul-84 19:40:53-BST
From: O'Keefe HPS (on ERCC DEC-10)
Subject: Prolog built in predicates

I recently had occasion to port the Prolog library from C-Prolog
to PopLog.  In 8 solid hours of hacking I didn't quite manage to
finish the conversion, but if I'd had EMACS instead of EDT I
would have.  It is unfortunate that any conversion was needed at
all; the library had already been converted from DEC-10 Prolog
to C-Prolog, and I thought most of the DEC-10 specific stuff had
either been knocked out entirely or identified.  What I'd forgotten
was that I'd introduced a fair bit of C-Prolog specific stuff in
the interests of efficiency.  One of these days I'll learn not to
worry about efficiency...  However, the C-Prolog version of the
library does run under Bill Clocksin's Prolog-X system.

I encountered three kinds of problems.
1) Pointless syntactic differences, such as the fact that
   /* start of commented out code:
   ......
   /* end of commented out code */
   caused Poplog to throw up its hands in horror and refuse
   to look at the rest of the file.
   I've now got a way around that: the DEC-10 parser and a
   suitable tokeniser are in the library.

2) Differences I was expecting, such as commands like "help"
   already being defined as something else, some predicates
   already being defined, some predicates the library uses
   not being defined, VMS file names being absurdly short so
   that library file names had to be changed, ...
   Some things are inherently system dependent, and the way
   to cope with this would have been for me to take more care
   to isolate system dependencies in the first place.

3) Differences in the built in predicates that I was not
   expecting and that did not add to the power of the language.
   I shouldn't *have* to cope with this problem.

There are a number of Prolog systems that claim to be close
enough to DEC-10 Prolog to be useful.  However, because their
authors have not had access to a clear description of what the
built in predicates are supposed to do, such fundamental things
as functor/3 often differ from DEC-10 Prolog/C-Prolog/Prolog-X
enough so that the Prolog library can't be used.  (By the way,
I'm not saying that Poplog is unusually incompatible.  On the
contrary, it is unusually DEC-10 compatible, and is getting
better all the time.)  The differences generally don't do a
thing to help anyone at all.

So I've written a 100+ page document describing most of the
predicates the Prolog library needs.  It does NOT specify
DEC-10 Prolog, indeed DEC-10 Prolog falls quite a way short
of this "proposed standard" in many respects.  It describes
an upwards compatible extension, including strings and complex
numbers as optional data types.  I have tried to give design
principles which determine what the built in predicates should
do rather than to design each one on its own.

The document isn't complete yet.  It doesn't cover the data base
or I/O, though it does touch on them.  However, I shan't have
any opportunity to do any more work on it this year, so I thought
I'd solicit comments on the current draft.

The document is now at SU-SCORE in <PROLOG>PLSTD.MSS (I think)
in SCRIBE format.  It refers to a number of the library files,
including some new ones written so that DEC-10 Prolog could
run code assuming the new "standard".

Please do read it if you can and send me your comments,
either to

        OKeefe.R.A.%EDXA@UCL-CS
                  ^ don't forget the third full stop!
or to
        Richard A. O'Keefe,
           Department of Artificial Intelligence,
              University of Edinburgh,
                 8 Hope Park Square,
                    Meadow Lane,
                       Edinburgh EH8 9NW,
                          SCOTLAND.

If you want a hard copy, I'm afraid you'll have to wait until it
comes out as a DAI Working Paper, which shouldn't be too long.
I'm not the person to write to, you can find out who to write to
for DAI reports by looking in almost any recent issue of SIGART.