Re: Granting SET and ALTER SYSTE privileges for GUCs

Mark Dilger <mark(dot)dilger(at)enterprisedb(dot)com> writes:
> On Nov 16, 2021, at 7:28 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> True; as long as the expectation is that entries will exist for only
>> a tiny subset of GUCs, it's probably fine.

> I understand that bloating a frequently used catalog can be pretty
> harmful to performance. I wasn't aware that the size of an infrequently
> used catalog was critical.

My concern is not about performance, it's about the difficulty of
maintaining a catalog that expects to be a more-or-less exhaustive
list of GUCs. I think you need to get rid of that expectation.
In the analogy to ALTER DATABASE/USER SET, we don't expect that
pg_db_role_setting catalog entries will exist for all, or even
very many, GUCs. Also, the fact that pg_db_role_setting entries
aren't tied very hard to the actual existence of a GUC is a good
thing from the maintenance and upgrade standpoint.

BTW, if we did create such a catalog, there would need to be
pg_dump and pg_upgrade support for its entries, and you'd have to
think about (e.g.) whether pg_upgrade would attempt to preserve
the same OIDs. I don't see any indication that the patch has
addressed that infrastructure ... which is probably just as well,
since it's work that I'd be wanting to reject. (Hm, but actually,
doesn't pg_dump need work anyway to dump this new type of GRANT?)

regards, tom lane