"Ann W. Harrison" <aharrison@...> wrote:

>
> So, the question is whether we keep a long-standing, non-standard
> behavior. And, this time, my vote is no. I wouldn't make the
> change in 1.5, but would in FB2 and Vulcan.
>
> This is non-standard behavior in a SQL standard feature, which
> is a bug. It's not a priority 1 bug and there are workarounds,
> but it is a bug and we ought to fix it.

Agreed. And it's already fixed in my local copy of HEAD :-)

> 2) There are other issues around constraints, including
> their use of other tables, an whether or not they are
> invoked on delete. Here I agree with Dmitry that the
> standards does not disallow cross-table constraints and
> with everyone that constraint handling tends not to be
> as solid as it should be. But that's for another day.

OK. We also have a bugreport that check constraints should perform in the
system context, i.e. work as "dirty read" (like our PK and FK behave). I
don't have a clue how it corresponds to the SQL spec.


Dmitry