Robert Haas <robertmhaas@gmail.com> writes:
> One thing that's bugging me a bit about this whole line of attack is
> that, in the first instance, the whole goal here is to support
> inheritance hierarchies that mix ordinary tables with foreign tables.
> If you have a table with children some of which are inherited and
> others of which are not inherited, you're very likely going to want
> your constraints enforced for real on the children that are tables and
> assumed true on the children that are foreign tables, and none of what
> we're talking about here gets us to that, because we normally want the
> constraints to be identical throughout the inheritance hierarchy.
There's a nearby thread that's addressing this same question, in which
I make the case (again) that the right thing for postgres_fdw constraints
is that they're just assumed true. So I'm not sure why this conversation
is proposing to implement a lot of mechanism to do something different
from that.
regards, tom lane