Re: allow granting CLUSTER, REFRESH MATERIALIZED VIEW, and REINDEX

On Wed, 14 Dec 2022 at 14:47, Jeff Davis <pgsql(at)j-davis(dot)com> wrote:

Furthermore, MAINTAIN privileges on the partitioned table do not grant
> the ability to create new partitions. There's a comment in tablecmds.c
> alluding to a possible "UNDER" privilege:
>
> /*
> * We should have an UNDER permission flag for this, but for now,
> * demand that creator of a child table own the parent.
> */
>
> Perhaps there's something we want to do there, but it's a different use
> case than the MAINTAIN privilege, so I don't see a reason it should be
> grouped. Also, there's a bit of weirdness to think about in cases where
> another user creates (and owns) a partition of your table (currently
> this is only possible if the other user is a superuser).
>

I strongly agree. MAINTAIN is for actions that leave the schema the same.
Conceptually, running MAINTAIN shouldn't affect the result of pg_dump. That
may not be strictly true, but adding a table is definitely not something
that MAINTAIN should allow.

Is there a firm decision on the issue of changing the cluster index of a
table? Re-clustering a table on the same index is clearly something that
should be granted by MAINTAIN as I imagine it, but changing the cluster
index, strictly speaking, changes the schema and could be considered
outside of the scope of what should be allowed. On the other hand, I can
see simplicity in having CLUSTER check the same permissions whether or not
the cluster index is being updated.