Re: theory and practice: ying and yang

Paul wrote:
> It seems to me that the null issue is all about whether unknown values
> should be handled by the relational engine (as in SQL, using 3-valued
> logic) or in the type engine, i.e. set up specifically for each domain/type.

Thanks for the concise and cogent definition - I never saw it spelled out quite that way.

SQL's problem, though, is an inconsistent and incoherent use of nulls - it's not exactly 3-valued, and doesn't mesh with any NVL theory I've read about. Read especially Date's criticisms of null in SQL aggregate functions.

• Eric