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

x wrote:
> Prolog will let you store anything you want that is expressible in
> Prolog. If you want contradictory statements, you can have them.

Hmm that's strange. In a relational database I can't have both the proposition "Employee 123 is called John" and the constraint "no employee is called John" in my database. If I have one, it stops me adding the other.

I've got Prolog here and I'm playing with it trying to write a program that has two contradictory statements. It seems that the "not" predicate in Prolog works a bit strangely. Could you give me an example of a simple Prolog program that has two terms that contradict each other? Would in then follow that in that system anything is provable?

>>> I'm not sure about this. There must be exactly *ONE* real-world
>>> "interpretation" of the database.
>> 
>> Why must there? consider the database with one tuple like this: 
>> (1,2) There could be many real-world interpretations of this 
>> database surely? It's not inconceivable that two people who've 
>> never met have identical databases with totally different 
>> interpretations.

I'm using identical to mean "identical syntax, though possibly different semantics". (And maybe different relation, column names).

I think maybe we're talking at cross-purposes here: I'm thinking of a database from a purely syntactically viewpoint, you're thinking of it as a whole, semantics included.

I think that Godel's Completeness theorem says that if a statement about a database is semantically true *irrespective of which semantics you choose*, then it's provable purely syntactically using first order logic.

What database-related elements would you say correspond to the different parts of Godel's theorem (logic, theory, model, axioms)? If indeed you think it's applicable at all?

Paul. Received on Mon Jun 07 2004 - 16:53:07 CEST