Andrew,
> The requirement was, anyway, that we be able to read old versions of
> "archived" rows. IIRC there was an implementation choice, whether we would
> _never_ allow such rows to be SET READ WRITE or whether they'd be
> immediately upgraded to the present format on SET READ WRITE.
Well, in theory we need this capability for upgrade-in-place too. While that
project has kind of stalled for the moment, we'll pick it back up again soon.
--
Josh Berkus
PostgreSQL @ Sun
San Francisco