Re: GLOBAL vs LOCAL temp tables

Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> The real problem with current temp tables is the implementation. I see
> very quick growth of system catalogs with heavy use of temp
> tables(some hundred mega bytes per week on a busy system for
> example). To fix the system catalogs, we have to stop postmaster and
> have to do reindex. This is truly a pain.

I believe the btree compaction logic in CVS tip will fix this.
It would be nice to see in-the-field proof though.  Don't suppose you
want to run a test system with CVS tip?
        regards, tom lane