Re: So what's null then if it's not nothing?

Jon Heggland wrote:
> In article <1133272410.483257.280890_at_o13g2000cwo.googlegroups.com>,
> boston103_at_hotmail.com says...
[...]
> > With respect to nulls, SQL is no more confused that Codd's original
> > work on which SQL's null treatment is based. There is no need to
> > attack SQL whose deficiencies are well known if all one wants to
> > discuss is the notion of null..
>
> Then consider my statement an attack on Codd, if SQL is faithful to his
> work. Alternatively, tell me how SQL is deficient with respect to Codd.

Most notably, SQL deviates from the 3vl with respect to EXIST.

>
> > > > Whether or not two values are considered distinct is irrelevant to the
> > > > null = null comparison.
> > >
> > > By SQL fiat, perhaps. But *should* it be? What gives SQL the right to
> > > redefine notions of equality and "distinctness" in this manner? Or never
> > > mind the right; does it make *sense*? Is it worth the price?
> >
> > null=null evaluates to unknown according to the 3vl logic rules.
>
> Indeed? Is NULL an inherent part of 3VL? Reference? I don't think Codd
> invented 3VL ...

In the Codd style 3vl it is. See the article I mentioned earlier. He uses w (omega) to denote NULL and his NULL is "value at present unknown".

He did not invent the 3VL, there had been Lukasiewicz before.

>
> > Again, SQL per se does not have much to do with the notion of null.
> > Please see Codd's article I mentioned earlier and comment on it rather
> > that criticizing SQL.
>
> Isn't it pretty much the same? Again, please tell me what you consider
> the deficiencies of SQL.

See above.

>
> As for Codd's article, I think it is a bad idea in itself to use 3VL.

I fully agree that nulls should be avoided as much as possible, however, if one uses them one should understand their behavior and the original motivation (by Codd). If one does not use nulls, then there is nothing to discuss really.

> Codd himself admits it produces seemingly paradoxical results and breaks
> what we expect to be tautologies. Also, to conflate the unknown truth
> value with null seems like a huge blunder. Finally, he discusses the
> RM's relationship to predicate logic, but fails completely to consider
> the effects of nulls on this relationship.
> --
> Jon
Received on Tue Nov 29 2005 - 17:16:32 CET