On Wed, Sep 9, 2020 at 2:37 PM Michael Paquier <michael@paquier.xyz> wrote:
>
> Another thing that was itching me is the introduction of 3 GUCs with
> one new category for the sake of two SQL functions. For VACUUM we
> have many things relying on the GUC delays, with autovacuum and manual
> vacuum. Perhaps it would make sense to have these if we have some day
> a background worker doing checksum verifications, still that could
> perfectly be in contrib/, and that would be kind of hard to tune as
> well. The patch enabling checksums on-the-fly could also be a reason
> good enough. Another thing we could consider is to pass down those
> parameters as function arguments, at the cost of not being able to
> reload them.
I'm not terribly happy with adding that for now, but it's quite clear
that there'll eventually be a lot of new stuff added that will benefit
from either the category or the GUC. FTR once we reach an agreement
on how to do this check (I'm wondering if it'll stay an SQL function
or become a plain backend command, in which case GUCs would be
mandatory), I'll also be happy to work on a background worker to help
people running the check regularly. So in my opinion it's better to
add them now so we won't have to change the sql function definition
later when other facilities will rely on them.
