RCode vs. CCode

• Mail folder: Interlingua Mail
• Next message: Robert MacGregor: "On the role of quotation in an RCode"
• Previous message: John McCarthy: "RCode vs CCode"

Message-id: <9009101633.AA29926@vaxa.isi.edu>
To: Interlingua@vaxa.isi.edu
Cc: hayes@parc.xerox.com
Reply-To: macgregor@venera.isi.edu
Subject: RCode vs. CCode
Date: Mon, 10 Sep 90 09:33:54 PDT
From: Robert MacGregor <macgreg@vaxa.isi.edu>

Pat has done us a great favor by raising the level of the Interlingua
discussion.  An important question before us is whether a CCode or an
RCode makes the more appropriate vehicle for translating between KR
systems.  Peter and Ramesh have already come out in favor of an RCode
approach, while Genesereth has declared for the CCode approach.

Claim 1: Implementing a CCode is beyond the abilities of current
inference technology.

Corollary: Since we can't build a CCode, the Interlingua should
implement an RCode.

Argument for Claim 1: Ramesh's discussion on interpreting the U.S.
Constitution is intended to illustrate the enormous amount of
inference necessary to emulate a CCode.  Peter and I in separate
earlier messages argued that instantiating high level language
constructs (an RCode approach) was preferable to mailing a bundle
of axioms describing one's own interpreter along with one's
knowledge base (a CCode approach).

Claim 2: The current Interlingua proposal is trying to be both
an RCode and a CCode.

Although Mike has recently declared for the CCode approach, he has on
numerous occasions argued that one ought to be able to execute an
Interlingua spec.  My impression is that EpiKit does exactly that
(modulo small differences in syntax).  Mike, I'd like to have your
impression on this, now that we have additional vocabulary for
discussing it.

Pat has noted that an emphasis on the use of quotation makes sense
for a CCode, but not for an RCode.  I completely agree.  If we
decide that an RCode approach is the correct model for an Interlingua,
then I would argue that the role of quotation in the Interlingua should
be limited strictly to that of a sentential belief operator, i.e.,
it should be an alternative to a propositional representation of
belief, and nothing more.

- Bob