Re: SUO: What the marketplace needs from us.
Jean-Luc,
I agree with you about the limitations of
relational databases. SQL is definitely *not*
my favorite logic-based notation, and the way
it has been implemented in RDBs leaves very
much to be desired.
> THIS is the problem, the bulk of the database
> is only *ground-level*, and even if stored
> queries and views are FOL and horn clauses
> the whole of the database is not homogenous
> and self-referencable. For instance you cannot
> have queries about the stored queries that
> would have such or such characteristics.
We very much need such facilities, and they are
not "blue-sky" theory. They are practical features
that a well designed system should support.
However, the purpose of my note to Eric was to point
out that the features which he considered practical
and which are currently planned for the so-called
"semantic web" are not even at the level of that
30-year-old DB language called SQL.
I completely agree with your following comment:
> Fairly common shortsightedness from the designers
> who are bent on solving the "practical problems of
> the day" while dismissing generality as "too complex".
> General solutions are of course a little harder
> to *think about* but also SIMPLER than ad hoc
> solutions and MUCH CHEAPER to implement (really!).
Certainly. And the QUEL language implemented on Ingres
was far and away more general and more easily extendable
than SQL. I blame many of the limitations of SQL on
the "human factors" people who did experiments on
secretaries and college students to see if they could
use it to ask simple queries. I have nothing against
such testing, but at least equal attention must be
paid to the theoretical foundations.
Ted Codd, by the way, was indeed very concerned about
the theory when he wrote his original papers, and he
was working down the hall from the people who brought
us SQL. However, Ted did not pay much attention to
SQL and let the designers go their own way. One reason
why QUEL was so much better than SQL is that it was
designed by people (i.e. Michael Stonebraker) who
understood both theory and practice. We need both.
My current concern is that the semantic webbers are
designing technology for the 21st century that does
not even come up to the level of the RDBs, which
were designed in the 1970s and are today running the
economy of the entire world. The semantic webbers
are ignorning both theory and practice.
John