Andres Freund <andres(at)anarazel(dot)de> writes:
> Given that this is just SQL level, I don't see why we'd need a special
> type of language here. You can just use DO etc.
I'd make that point against the whole proposal. There's nothing here that
can't be done with current_setting() + set_config(). I'm pretty dubious
about layering extra functionality into such a fundamental utility command
as SET; and the fact that we've gone twenty-odd years without similar
previous proposals doesn't speak well for it being really useful.
regards, tom lane