Re: Proposal: 6NF

"Brian Selzer" <brian_at_selzer-software.com> wrote in message news:hwoVg.7861$TV3.6237_at_newssvr21.news.prodigy.com...

> I agree that decomposition into separate relations still has its place.
I
> just don't think that nulls should be dismissed arbitrarily.
>
> After further consideration (prompted by Bob's harangue), I think that the
> results of some of the operators enumerated above are not sensible. Ø does
> not belong to any numeric domain, and you can't add apples and oranges,
but
> on the other hand, there is only one empty set, so Ø = Ø should be TRUE.
>

I've been following your discussion on this subject, and I largely agree with the major points you've been making. (Unusual, since we've disagreed in the past).

I also agree with the above, that nulls should not be dismissed arbitrarily. In another subthread, I'm trying to develop the theme one point at a time. Where I actaully get to depends on the responses I get.

But here's where I anticipate getting to: as a theoretical tool, nulls are unnecessary. A system without nulls can be just as expressive as one with nulls. As a practical matter nulls are just about indispensable. A system built with no accomodation for nulls just isn't going to work as practically as one that allows for nulls.

Where I'm very uncertain is whether admitting nulls inescapably leads to 3VL. I think not. But I'm not there yet.

But on nulls and the empty set. The empty set is very clearly a value. Null is very clearly not a value. Using the empty set in place of a null is very clearly the road to confusion.

There is one point I'm confused on: what is the domain of the empty set? does it even have a domain? To me, the empty set of character strings is not "the same thing" as the empty set of integers. But I may be thinking like a computer person and not like a mathematician. Received on Fri Oct 06 2006 - 14:59:15 CEST