On Thu, 2009-09-17 at 17:44 -0400, Tom Lane wrote:
> Dimitri Fontaine <dfontaine@hi-media.com> writes:
> > I don't see any reason why not breaking the user visible behavior of
> > tuples CTID between any two major releases,
>
> > Am I completely wet here?
>
> Completely. This is a user-visible behavior that we have encouraged
> people to rely on, and for which there is no easy substitute.
Agreed. I investigated that avenue as a possible implementation approach
when designing HOT and I didn't find anything worth taking away.
I'm very much in favour of a higher-level solution to compacting a
table, as has been discussed for the batch update utility. That avoids
most of the low-level yuck that VACUUM FULL imposes upon itself and
everyone around it. If we want to move forward long term we need to keep
the internals as clean as possible. Hot Standby would never have been
possible without that principle having already been applied across the
other subsystems.
-- Simon Riggs www.2ndQuadrant.com
