Re: NULLs

On Jan 8, 7:50 am, Hugo Kornelis
<h..._at_perFact.REMOVETHIS.info.INVALID> wrote:

> And finally, I think that the closed world assumption will be a hard
> sell for businesses. I found this definition "Closed world assumption:
> if you cannot prove P or ~P from a knowledge base KB, add ~P to the
> knowledge base KB." athttp://cs.wwc.edu/~aabyan/Logic/CWA.html. If I
> understand this correctly, this implies that, if a knowledge base holds
> information about a person with ID XP55303, but no information
> whatsoever about the birthdate of this person XP55303, this person does
> not have a birthdate at all and hence can not exist.
>
> Most businesses will (unless legally prohibited) gladly do business with
> persons who refuse to state their date of birth - as long as they can
> wave a valid credit card. Telling a customer she doesn't exist is bad
> for business. Ergo, most businesses choose the Open World assumption.

The above formal definition of the CWA above is compatible with missing information when the intensional definition is stated appropriately. Eg

P = it is known to HR department that Alice is 25 years old.

If P cannot be proven from the DB, then we deduce ~P.

IMO the intensional definitions (which are outside the RM formalism) should always accommodate the need to support missing information where required, and therefore the RM/RA which gives us a logical inference system can *always* assume the CWA. There are many advantages. We can stay strictly in a 2VL, we avoid the problems with tautologies, we can interpret relation difference in terms of logical negation etc.

I won't argue with your desire to use NULL as long as you choose Zaniolo's interpretation which has a very simple correspondence to a 2VL. I have the feeling that NULLs are practical in tables with many null-able attributes, because the "proper" approach using greater normalisation would be painful, particularly if the primary key isn't particularly simple or stable. Received on Tue Jan 08 2008 - 01:58:00 CET