From: Marshall <>
Date: Sat, 5 Jan 2008 14:04:28 -0800 (PST)
Message-ID: <>

On Jan 5, 7:29 am, "David Cressey" <> wrote:


> The "not applicable" case is the one that can be obviated by normalization.
> That may the the reason that c.d.t. regulars treat it as unworthy of
> discussion.

Well, either case, and any other reason that one might use nulls, can be obviated by some approach or another that doesn't use nulls. We could just use booleans for example.

> The "not available" case is more philosophical. Even with full
> normalization there can be facts that the database doesn't know, and knows
> that it doesn't know. There can also be facts that the database doesn't
> know, and doesn't know that it doesn't know. In my mind, this case has more
> to do with the nature of knowledge itself than with the theory of databases.
> But, for some reason, it never fails to attract the same discussion every
> time the topic comes up.

I have noticed that there are certain things that people (different people
at different times) want to move from user-level considerations into system-level considerations. Missing/inapplicable is one often one such.
Time seems to be another such, and occasionally one sees default values being addressed this way. The motivation seems to be the desire to get something for nothing. The question (IMHO) is whether such a move makes the system work any easier. It often seems to be the case that the short term situation gets better and the long term situation gets worse.

Marshall Received on Sat Jan 05 2008 - 23:04:28 CET

Original text of this message