Thomas Lockhart <lockhart@fourpalms.org> writes:
> I do have a concern about how to implement some of the SET commands if
> we *do* respect transactional semantics. For example, SET TIME ZONE
> saves the current value of an environment variable (if available), and
> would need *at least* a "before transaction" and "after transaction
> started" pair of values.
I intended for guc.c to manage this bookkeeping, thus freeing individual
modules from worrying about it. That would require us to transpose the
last few special-cased SET variables into generic GUC variables, but
I consider that a Good Thing anyway.
regards, tom lane