Neil Conway <nconway@klamath.dyndns.org> writes:
> I also took the opportunity to remove a bunch of code in trigger.c
> that wasn't doing anything useful, AFAICS -- according to SQL99, SET
> CONSTRAINTS should only apply to the current transaction.
I would assume that Jan and Stephan put that code in for a reason.
"Rip out anything I can't find in SQL99" is not the way we do things
around here ...
regards, tom lane