Re: NULLs: theoretical problems?

"paul c" <toledobythesea_at_oohay.ac> wrote in message news:JXLui.45171$rX4.26997_at_pd7urf2no...
>
> (even though I'm not sure in "s{X} = t{X} implies s{Y} = t{Y}" whether
> "implies" stands for logical implication.)
>

Good catch. It seems that logical implication is not well defined for three-value logic. According to Lex de Haan and Toon Koppelaars' recent book:

 P | Q | P->Q

 T | U | U
 F | U | T
 U | T | T
 U | F | U
 U | U | U or T?

So possibly I have to modify what I said about no attribute with nulls satsifying any FD. Although some dependencies may be satisfied it won't be enough for 2NF or anything higher because {K}->{N} is never true where K is a key and N is an attribute which has nulls.

The fact that attributes with nulls don't satisfy join dependencies remains very solid and much more obvious.

-- 
David Portas