KB interchange standards

John McCarthy <jmc@sail.stanford.edu>
Date: Tue, 26 Nov 91 09:39:21 -0800
From: John McCarthy <jmc@sail.stanford.edu>
Message-id: <9111261739.AA08702@SAIL.Stanford.EDU>
To: ginsberg@t.stanford.edu
Cc: JFULTON@atc.boeing.com, TI@darpa.mil, friedland@ptolemy.arc.nasa.gov,
        gio@darpa.mil, interlingua@isi.edu, kr-advisory@isi.edu,
        neches@isi.edu, ohlander@isi.edu, ramesh@isi.edu
In-Reply-To: Matthew L. Ginsberg's message of Tue, 26 Nov 91 08:56:48 PST <9111261656.AA00689@t.Stanford.EDU>
Subject: KB interchange standards
Reply-To: jmc@cs.stanford.edu
I also believe that languages for knowledge interchange will not
settle down soon.  However, I don't object to temporary standards.
In fact I think they will advance the field.

Here's a technical idea, or the beginning of one, intended to make
upgrading the standard easier.  Any expression in KIF should
allow encapsulation with a version number.  Thus if
(at McCarthy Stanford) is a legal expression, then
(KIF-1993 (at McCarthy Stanford)) should also be legal.  In
KIF-1993, the two expressions will  have the same meaning.
However, KIF-1995 may give (at McCarthy Stanford) a different
meaning.  Programs that use KIF should often dispatch on
the KIF version label even when there is only one meaning
in all current versions.

If a text has left out version labels, it will be easy to put
them back in for certain classes of expressions, e.g. all
occurrences of (at ?x ?y) can be replaced by (KIF-1993 (at ?x ?y)).

Sorry for bothering people if this has already been provided for.