Re: Commits 8de72b and 5457a1 (COPY FREEZE)

* Robert Haas (robertmhaas@gmail.com) wrote:
> You know, I hadn't been taking that option terribly seriously, but
> maybe we ought to reconsider it.  It would certainly be simpler, and
> as you point out, it's not really any worse from an MVCC point of view
> than anything else we do.  Moreover, it would make this available to
> clients like pg_dump without further hackery.

I really don't agree with this notion that the behavior of TRUNCATE, a
top-level, seperately permissioned command, makes it OK to introduce
other busted behavior in existing commands.

> I think the current behavior, where we treat FREEZE as a hint, is just
> awful.

I agree that it's pretty grotty, but I had assumed it was at least
deterministic, ala TRUNCATE/COPY and WAL...  If it isn't, then this
certainly gets really ugly really quickly.

I don't think that means we should go ahead and try to always optimize
it though- even when it isn't explicit, there will be an expectation
that it's going to work when all the 'right' conditions are met.  I know
that's certainly how I feel about TRUNCATE/COPY and WAL'ing.
Thanks,
    Stephen