Need for standards

sowa@watson.ibm.com
Message-id: <199201100523.AA11679@venera.isi.edu>
Date: Fri, 10 Jan 92 00:20:41 EST
From: sowa@watson.ibm.com
To: INTERLINGUA@isi.edu, KR-ADVISORY@isi.edu, SRKB@isi.edu, CG@cs.umn.edu
Cc: JONESM@atc.boeing.com, SPEYER@mcc.com, TYLER@cme.nist.gov
Subject: Need for standards

Ken Forbus commented on the insufficient evidence for good standards
for knowledge representation:

> I guess what bothers me in this whole discussion is that all of the
> complaints and arguments against SQL2 being a reasonable idea could be
> leveled at the interlingua and KR standards effort.  Based on the small
> number of experiments in large-scale KR (e.g., CYC) I doubt that we
> have enough evidence to comfortably set standards that will actually
> be useful over a long period.

The main problems with SQL2, as my ISO correspondent pointed out, are
the structural weaknesses of SQL-1.  And those weaknesses did not result
>From any lack of evidence:  Ted Codd's original proposals were far
better structured than SQL; various trial implementations (including
one that my correspondent helped to develop) had query languages that
were better than SQL; and any number of people in the industry could
have (and did!) design better languages than SQL.  But the people who
had better systems did not actively participate in the discussions that
led to the adoption of SQL as a standard.

Yet despite the weaknesses of SQL, a mediocre standard is better
than no standard at all.  The standardization of SQL has provided an
enormous benefit in allowing mainframes, minicomputers, workstations,
and personal computers to share data.  What made SQL succeed is not
the fact that ANSI and ISO blessed it, but the fact that the industry
desperately needed a standard, and SQL was the only thing available.
The main reason why we are left with such a mediocre standard is that
the people who knew better didn't participate (that's the same reason
why we end up with mediocre politicians).

I would like to see some of the best researchers get involved in
in the standards efforts because I see an enormous demand building
up for standards for information interchange -- for CASE tools,
for engineering and manufacturing data exchange, for communications,
for OO-systems, for semantic databases, for knowledge bases, etc.
Standards are being promoted whose theoretical foundation is dubious
or nonexistent.  If we continue to do our research while ignoring what's
going on around us, we may be forced to live with or interface to some
languages that are even more distasteful than SQL.

I realize that we do not know enough today to design a perfect standard.
But we will never know enough.  If we don't do our best, other people
may do something a lot worse.

John Sowa