Re: role self-revocation

Greetings,

* David G. Johnston (david.g.johnston@gmail.com) wrote:
> On Thu, Mar 10, 2022 at 3:01 PM Robert Haas <robertmhaas@gmail.com> wrote:
> 
> > On Thu, Mar 10, 2022 at 4:00 PM David G. Johnston
> > <david.g.johnston@gmail.com> wrote:
> > > I dislike changing the documented behavior of CREATEROLE to the degree
> > suggested here.  However, there are three choices here, only one of which
> > can be chosen:
> > >
> > > 1. Leave CREATEROLE alone entirely
> > > 2. Make it so CREATEROLE cannot assign membership to the predefined
> > roles or superuser (inheritance included), but leave the rest alone.  This
> > would be the hard-coded version, not the role attribute one.
> > > 3. Make it so CREATEROLE can only assign membership to roles for which
> > it has been made an admin; as well as the other things mentioned
> > >
> > > Moving forward I'd prefer options 1 or 2, leaving the ability to
> > create/alter/drop a role to be vested via predefined roles.
> >
> > It sounds like you prefer a behavior where CREATEROLE gives power over
> > all non-superusers, but that seems pretty limiting to me.
> 
> Doh!  I edited out the part where I made clear I considered options 1 and 2
> as basically being done for a limited period of time while deprecating the
> CREATEROLE attribute altogether in favor of the fine-grained and predefined
> role based permission granting.  I don't want to nerf CREATEROLE as part of
> adding this new feature, instead leave it as close to status quo as
> reasonable so as not to mess up existing setups that make use of it.  We
> can note in the release notes and documentation that we consider CREATEROLE
> to be deprecated and that the new predefined role should be used to give a
> user the ability to create/alter/drop roles, etc...  DBAs should consider
> revoking CREATEROLE from their users and granting them proper memberships
> in the predefined roles and the groups those roles should be administering.

I disagree entirely with the idea that we should push the out however
many years it'd take to get through some deprecation period.  We are
absolutely terrible when it comes to that and what we're talking about
here, at this point anyway, is making changes that get us closer to what
the spec says.  I agree that we can't back-patch these changes, but I
don't think we need a deprecation period.  If we were just getting rid
of CREATEROLE, I don't think we'd have a deprecation period.  If we need
to get rid of CREATEROLE and introduce something new that more-or-less
means the same thing, to make it so that people's scripts break in a
more obvious way, maybe we can consider that, but I don't really think
that's actually the case here.  Such scripts as will break will still
break in a pretty clear way with a clear answer as to how to fix them
and I don't think there's some kind of data corruption or something that
would happen.

Thanks,

Stephen