"Martijn Tonies" <m.tonies@stripped> wrote on 16/03/2006 11:02:32:
> Well, the question still is if you should store "unknown" at all ;)
>
> Not according to Date: you should store what is known. See the remarks
> about the "true propositions", from which relational databases are
derived
> (but you probably know that).
As someone totally unread in the theory of databases, that seems unduly
puritanical. I assume that what Date would propose is that you have
another table (related by master key) in which, if you do not know
something, you do not enter it. But this means that if you have 10
different pieces of potentially but not necessarily available information
about a single master record (e.g. a person), you have to do a 10-way join
in order to retrieve all the information about them. Replacing a
theoretically ugly null flag with a 10 way join strikes me, as an engineer
rather than a theoretician, the wrong side of the elegance/practicality
trade-off.
Alec
