Alexander Korotkov <a.korotkov@postgrespro.ru> writes:
> Could we address both this problems by denying changing existing
> commutators and negator? ISTM that setting absent commutator and negator is
> quite enough for ALTER OPERATOR. User extensions could need setting of
> commutator and negator because they could add new operators which don't
> exist before. But it's rather uncommon to unset or change commutator or
> negator.
Note that this functionality is already covered, in that you can specify
the commutator/negator linkage when you create the second operator.
I'm not particularly convinced that we need to have it in ALTER OPERATOR.
regards, tom lane