Hi Hugo

No need to be sorry about the self-promotion when it leads to an interesting blog entry with a useful external reference. The "much ado" paper contains several excellent examples of Date's ability to think unclearly. I wasn't aware this was available on the web.

I think we will probably continue to diagree about NULLs; Date's reductio ad infinitum argument about a system with two or more nulls is blatant nonsense. I can see how the argument that the I-NULL has semantic content outside the domain originally designated for the values can be made, but it can equally be made for the A-NULL and indeed for the single NULL of the 1979 paper (after all, the model has it explicitly added into each domain because the original domain contains no such value), so if one accepts the argument that this is a violation of atomicity or of 1NF one ends up unable to accept NULLs at all. So I'm happy with 2 distinct nulls. Of course I'm also happy (slightly less happy - but only slightly) with a single NULL, as long of course as it's recognised that it is not a value in for example the domain of integers but a marker indicating the absence of a value. I think it's possible that Codd went too far when he introduced a 4-valued logic to go with his 2 NULLs: there's nothing in a 2 NULL system that can't be handled easily in 3-valued logic (as Codd himself seems to say in the 1993 paper you referenced).

Anyway, thanks for the comment; while we may disagree, I'm perfectly happy to see people disagree with me as long as (like you, but unlike Date on the topic of NULLs or some people who have taken him too seriously) they have coherent and respectable arguments for the positions they take.