Mark,
This is why I choose to use the term "LOCATION" instead of "TABLESPACE"
. A "LOCATION" is a directory just like Postgresql has today. All the
patch would add is the ability to put object under different "LOCATION"
for the same database.
Jim
> Tom Lane wrote:
> >
> > "Jim Buttafuoco" <jim@buttafuoco.net> writes:
> > > I propose to add a default data location, index and temporary
locations
> > > to the pg_shadow table to allow a DBA to specify locations for
each
> > > user when they create databases, tables and indexes or need
temporary
> > > disk storage (either for temporary tables or sort files).
> >
> > Have you read any of the previous discussions about tablespaces?
> > This seems to be tablespaces with an off-the-cuff syntax. I'd
> > suggest taking a hard look at Oracle's tablespace facility and
> > seeing how closely we want to duplicate that.
>
> Sorry I missed the conversation about tablespaces. One of the reasons
I think
> Postgres is so usable is because it does not require the use of
tablespace
> files. If by tablespace, you mean to declare a directory on a device
as a
> tablespace, then cool. If you want to create tablespace "files" ala
Oracle, you
> are heading toward an administration nightmare. Don't get me wrong,
the ability
> to use a file as a tablespace would be kind of cool, i.e. you can
probably use
> raw devices, but please to not abandon the way postgres currently
works.
>
> On our Oracle server, we have run out of space on our tablespace files
and not
> known it was coming. I am the system architect, not the DBA, so I
don't have
> (nor want) direct control over the oracle database operation. Our
newbe DBA did
> not make the table correctly, so they did not grow. Alas he was laid
off, thus
> we were left trying to figure out what was happening.
>
> Postgres is easier to configure and get right. IMHO that is one of its
very
> important strengths. It is almost trivial to get a working SQL system
up and
> running which performs well.
>
>