Tom Lane writes:
> Right. In SQL92 constraint names have to be unique within the table's
> schema. Postgres allows two different tables to have similarly-named
> constraints, and that difference is the root of the issue.
But that should not prevent us from assigning an explicit schema to each
constraint, as we in fact currently do. This issue is a bit more tricky
than it seems. For example, constraints may also belong to a domain, so
even if we allowed SET CONSTRAINTS a.b.c it is still not clear that "b" is
a table.
--
Peter Eisentraut peter_e@gmx.net