Excerpts from Robert Haas's message of mié may 25 12:54:28 -0400 2011:
> I don't know. That's maybe better, but I'd be willing to wager that
> in some cases it will just slow down the rate at which we converge to
> a completely incorrect value, while in other cases it'll fail to
> update the data when it really has changed.
For regular tables I don't think there's a real problem because it'll
get fixed next time a full scan happens anyway. For toast tables, I
think the set of operations is limited enough that this is easy to prove
correct (or fixed so that it is) -- no HOT updates (in fact no updates
at all), etc.
BTW one thing we haven't fixed at all is how do HOT updates affect
vacuuming behavior ...
> I am wondering, though, why we're not just inserting a special-purpose
> hack for TOAST tables. Your email seems to indicate that regular
> tables are already handled well enough, and certainly if we only whack
> around the TOAST behavior it's much less likely to fry anything.
Well, having two paths one of which is uncommonly used means that it
will get all the bugs. As with the xl_commit WAL record comment from
Simon, I'd rather stick with having a single one.
--
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support