Re: Report on ANSI X3H4 meeting
Peter Karp <pkarp@ai.sri.com>
Reply-To: cg@cs.umn.edu
Date: Mon, 3 Feb 92 10:06:41 PST
From: Peter Karp <pkarp@ai.sri.com>
To: sowa@watson.ibm.com
Cc: kr-advisory@ISI.EDU, srkb@ISI.EDU, interlingua@ISI.EDU, cg@cs.umn.edu,
mzemanko@note2.nsf.gov, jwinkler@nasamail.nasa.gov,
skperez@mcimail.com, jfulton@atc.boeing.com, mjones@atc.boeing.com
Subject: Re: Report on ANSI X3H4 meeting
In-reply-to: Your message of Sun, 2 Feb 92 22:15:00 EST
Message-id: <CMM.0.90.0.697140401.pkarp@rockaway.ai.sri.com>
I find all this talk about AI researchers who are hostile to the
notion of knowledge representation standards rather perplexing because
I do not believe that knowledge-interchange standards will in fact
inhibit research on knowledge representation in any way. Some
knowledge representation researchers seem to view a "standard" as
being a legislated imperative that everyone who represents knowledge
for any reason must use an officially sanctioned representation. My
belief is that standards are for use by applications builders who wish
to exchange the knowledge or data in their application with others,
and who therefore voluntarily subscribe to one of perhaps several
existing standards. We should expect that only the best understood
aspects of knowledge representation theory have been standardized, and
therefore that researchers in knowledge representation who are
exploring the frontiers of this field will have no reason or desire to
utilize a standard. Does anyone plan to legislate that such
researchers must use some standard representation? I would certainly
expect them to experiment with representations that are either
extensions to an accepted standard, or that are a quantum leap beyond
any known standard.
I think the database and networking communities provide good examples
here. Does either community believe that the widespread standards
that have been adopted for applications have seriously inhibited
research? Despite the widespread use of SQL within relational
systems, database researchers have developed a whole new breed of
object-oriented database systems that have nothing to do with SQL.
And look at the networking world, where we have a number of coexisting
"standards" such as X.25, TCP/IP, and DECNET. Admittedly, each has
has been sanctioned by organizations with different political power.
But network users often have the choice of adopting any of the three,
and researchers in networking and distributed systems are still free
to develop and experiment with new networking protocols on their own
local networks that have nothing to do with these standard protocols.
I believe that the knowledge-representation standards community must
make a stronger effort to convince knowledge-representation
researchers that they have nothing to fear from standards -- and even
better, that after 20 years of research, that knowledge representation
research has a chance to influence how applications are actually built
(as you may have guessed, I believe we should have had a KR standard
10 years ago).
Peter