Alban Hertroys <haramrae(at)gmail(dot)com> writes:
> On 18 Jul 2012, at 5:08, Tom Lane wrote:
>> I wonder whether we could improve this by postponing the no-shell-types
>> check from creation to function runtime.
> I don't suppose it'd be possible to treat it as a deferred constraint? Then the check would be moved to the end of the transaction.
You mean, after we've already crashed, or allowed a security breach to
happen? Doesn't sound very helpful. In any case, my concern is that
there not be any added overhead, not about moving it around.
regards, tom lane