Jeff Janes <jeff(dot)janes(at)gmail(dot)com> writes:
> I thought superusers bypassed permissions checks, but was surprised where
> in this case they do not:
This amounts to asking for superuser permissions to propagate into
SECURITY DEFINER functions called by the superuser, which strikes me
as an utterly horrid idea.
(RI triggers aren't actually implemented with the SECURITY DEFINER
mechanism, but they act like they are.)
regards, tom lane