Re: Network Example: Sibling of Opposite Gender

On 2006-12-30, Lennart wrote:

> I did not have the time to discuss this earlier, but what you say is:
> as long as there are nulls in a schema it doesn't matter much
> (regarding normal forms), what we do in the schema, it is till not 1NF
> and hence, not in any other normal form as well.

At least as far as Codd goes, normal forms are orthogonal to nulls. The first are about reducing redundancy. The latter are about systematic treatment of missing and unknown information, which is known to be knowable, but still not known. Fitting such disparate concepts together is messy/productive, as always, but smearing the two concepts together is an even greater mistake.

> Clearly (IMO) there are better and worse schema's where null is
> present, agreed?

Yes. Nulls come about because of an essentially semantic concern. Such concerns force the relational/logic level, but they don't originate on it. If you only consider the relational level, you'll never understand why nulls exist. Hence, if you design schemas only considering the relational level and you're allowed to include nulls, you'll do it badly. That's how bad schemas with nulls come about. And at that level, you won't ever be able to appreciate nulls even when the higher, semantic concerns force someone to include them.

-- 
Sampo Syreeni, aka decoy - mailto:decoy_at_iki.fi, tel:+358-50-5756111
student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2