Ann,

> You would have been able to insert id=20 even if the constraint
> were present. The standard says that immediate constraints should
> be evaluated after the action, not before, so a standard compliant
> system would reject your update. Firebird evaluates constraints
> before the action, not after.

Again, please. Do you say that Firebird checks the constraint before
the INSERT when the table is perfectly consistent, then it allows
INSERT without checking the values, etc. making the database
inconsistent (e.g. any update or insert that does not fix the problem
will fail even when data are perfectly correct)?

> I'm not saying that's good, just that it might once have been
> defensible.

Would you or Helen write a document describing how constraints work?
Or maybe it is already in Helen's book?

Roman