> yes, in theory, the SQL Server approach is more 'pure'
Not really. The SQL standard consistently defines all the nullable
constraints (CHECK, UNIQUE, FOREIGN KEY) in the negative. In other words,
the constraint doesn't permit violations (logical FALSE) but nor does it
require that the constraint be satisfied (logical TRUE). The UNKNOWN case is
not considered a violation of a constraint, which is why NULLs are
permitted. Unfortunately Microsoft disagrees ;-)
--
David Portas
SQL Server MVP
--
Received on Thu Oct 21 2004 - 17:49:08 CDT
