Re: In an RDBMS, what does "Data" mean?

Anthony W. Youngman wrote:

>> No, I think in this analogy Newton's model does correspond to a
>> specific database design. The possibility that the relational
>> theory itself is wrong corresponds to the possibility that algebra
>> is wrong.

We have to be careful here with ambiguities of language. When I said the "possibility that algebra is wrong" I meant the theory of algebra itself, not some model that could be set up using the language of algebra.

Newton's algebra was not wrong, but what we mean here is the model that Newton made using algebra as the meta-language wasn't wrong. Newton didn't invent *an* algebra, he invented a model using algebra.

It's very confusing when you've got algebra as a kind of meta-language for Newton's theory. But then also you have logic as a kind of meta-language for algebra itself. In common speech we use the word "algebra" to mean both the theory of algebra itself, and pieces of code written in algebra.

> So you cannot prove that relational theory is right or wrong from
> WITHIN the theory.

That depends what you mean by "right"(!). First-order logic is your meta-language for talking about your database. So you can show that it will be "complete" in the senses mentioned before, because you are kind of jumping "outside" the theory. What you can't do is show that's it's the best methodology for managing data.

> And we also have experiments to show that the axioms do (or don't)
> accurately describe reality. Einstein showed that Newton's axioms
> didn't describe reality, and replaced them by new axioms that did a
> better job. He didn't alter Newton's algebra at all - indeed, he used
> exactly the same algebra ...

OK, agreed.

Hmm, this stuff has the tendency to seem crystal-clear when you're writing it, but then turns to gobbledygook when you re-read it... :)

Paul. Received on Fri Jun 04 2004 - 19:27:43 CEST