Tom Lane wrote:
> Alvaro Herrera <alvherre@commandprompt.com> writes:
>> FWIW I don't think this idea is silly at all. It's so not-silly, in
>> fact, that we already have some access methods that do this if an index
>> cannot be recovered (I think at least GiST does it).
>
> Well, there's a difference between "rebuild the index when it can't be
> recovered" and "lose the index anytime the system burps". AFAICS what
> Glen is proposing is to not WAL-log index changes, and with that any
> crash no matter how minor would have to invalidate indexes.
Nooo...! This has nothing to do with WAL logging index changes. What I
propose would have no effect on an end user that continues to back up
indexes. It would give people the *option* to not back up indexes.
This is about disaster recovery and the backups required to recover
sanely, not about hiccups that cause only the last handful of
transactions to be redone.
Causing invalidation of indexes after a crash might be acceptable in
some settings, but not in mine, and that isn't what I'm after.
-Glen