Christopher Kings-Lynne <chriskl(at)familyhealth(dot)com(dot)au> writes:
> Incidentally, shouldn't the existing RI queries (eg. SELECT ... FOR
> SHARE) explicity specify operator(pg_catalog.=)?
Then they'd be guaranteed to fail for datatypes/operators created in
other schemas, rather than only at risk of failing. Don't see that
as an improvement really. I think we should fix it properly for 8.2
but not institute any half-measures beforehand.
regards, tom lane