Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.3 4.3bsd-beta 6/6/85; site ucbvax.BERKELEY.EDU
Path: utzoo!decvax!bellcore!ulysses!ucbvax!B.CS.UIUC.EDU!mklein
From: mklein@B.CS.UIUC.EDU (Mark Klein)
Newsgroups: mod.ai
Subject: TMS Variables
Message-ID: <8605202118.AA17499@b.CS.UIUC.EDU>
Date: Tue, 20-May-86 17:18:54 EDT
Article-I.D.: b.8605202118.AA17499
Posted: Tue May 20 17:18:54 1986
Date-Received: Thu, 22-May-86 00:44:09 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The ARPA Internet
Lines: 14
Approved: ailist@sri-ai.arpa

I have been trying to decide whether or not to change some code I have (a 
pattern-directed inference engine, justification-, logic-, and assumption-
based TMS's) to handle variables in the data. Variables do seem to add
more flexibility (specifically, universal and existential quantification, I
guess), but have some problems as well - for example, one can no longer
use open-coded unification to match up assertions to rule triggers (I think).
What I'd like to know is whether it is easy or hard to have variables
in data for each kind of TMS (it seems to be simple for
the justification-based TMS)? Is there any advantage to having variables in 
data that you can't get simply by writing the appropriate rules (i.e. in
terms of efficiency, expressiveness, ability to control inference, etc)?

			Thanks,
			   Mark Klein