"scott.marlowe" <scott.marlowe@ihs.com> writes:
> On Fri, 27 Feb 2004, Tom Lane wrote:
>> In my mind, one of the main benefits of this work will be that we'll be
>> able to get *rid* of the initlocation stuff. It's a crock.
> OK, that's fine, but I keep thinking that a superuser should have to
> create the tablespace itself, and then tables can be assigned by users
> based on the rights assigned by the dba / superuser.
Yeah, we haven't yet gotten to the issue of permissions, but certainly
creating or deleting a tablespace has to be a superuser-only operation,
if only because you probably have also got some manual filesystem work
to do to set up the associated directory; and that has to be done as
root or postgres.
It might be a good idea to restrict connect/disconnect (if we use those
operations) to superusers as well.
regards, tom lane