Re: Which normal form is this violating?

> The reason that you had to compare the two tables is because you
> didn't know whether they represented the same predicate or not. Keep
> in mind that the database cannot in general know if two tables are
> logically the same or not.

So, if I have two tables with the same predicate (which, if I followed the thread correctly we *are* allowed - albeit in contradiction to Date's principle of Orthogonal Design), then if I rename the first table to be the same as the second, should a good DBMS in fact merge the two tables into one / return an error if the two tables state any contradictory facts ? Sound like a neat feature to me.

But then again, what about a view that was a union of the two tables? Would I receive a run time error if the tables ever came to state contradictory facts, or would the view creation have a side effect of constraining the tables to not get in such a state?

All sort of interesting in data integration scenarios and makes me wonder about what role 'names' have in the Model. I.e. I question Date's statement that "encoding meaning into names - of relvars or anything else - violates the Information Principle". My answer being, well what is in the system catalog if not names of relvars, attributes and types?

Regards
Paul Vernon Received on Sun May 12 2002 - 15:41:38 CEST