On Mon, 29 Mar 2004, Tom Lane wrote:
> "scott.marlowe" <scott.marlowe@ihs.com> writes:
> > since the purpose of the pg_hba.conf file is to ensure that you never
> > manage to lock yourself out of your database, might it make sense to have
> > a pg_hba table in each database that can be / will be / should be(???)
> > overidden by the pg_hba.conf file,
>
> I don't think we want user authentication driven off of actual tables.
> That would mean paying *all* the costs of backend launch before we could
> reject an invalid connection request.
>
> It might be possible to do something with a flat file as an intermediary
> between the postmaster and the tables that are the master data. We
> already do this for pg_shadow passwords, and I've been thinking of
> proposing that we add a flat file for the database name -> OID mapping
> so we could get rid of the horrid hack that is GetRawDatabaseInfo().
> Per-database flat files would be a bit messy though.
Actually, I had thought of pg_hba as being a global table, not a per
database one. That would mean only one flat file, wouldn't it?
And while we're at it, maybe we should have a setting somewhere should
someone execute the famous "update pg_shadow set usesuper = false" that
someone did a while back to be able to force an account to be a superuser
account. In postgresql.conf or something like it. While it's another
problem, it falls under the same "keeping people from locking themselves
out" thread.
---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match
