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

In light of the debate of the semantics of a null, well just rework the example:

1. The dog's colour is black
2. The dog's colour is ____.

to...

1. DOG has_colour BLACK.
2. DOG has_absent_property COLOUR.

Now I realise in an RDBMS to satisfy this sort of manipulation your going have to do a full decomposition down to binary predicates to handle this natively, and thats hardly going to be practical in terms of joins. However I see no reason why future DBMS could not handle these sort of situations in a more efficient manner. Nulls are after all information _about_ data (i.e. the data is absent), not values themselves, and as such have no place living where data itself does.The real world dictates that we will have to handle situations of absent data, so why not offer better support that suits both theory and practice.

So my question - would a system which recomposed tuples from fully decomposed (irreducible as opposed to binary) relations on request, but had the facility to specify "has_absent_property" be logistically feasible? Received on Fri Nov 18 2005 - 12:34:37 CET