Fabien COELHO <coelho@cri.ensmp.fr> writes:
> Just call that a kludge as it means that the admin is expected to create
> as many dummy and unknown (if you have a custom dump file)
> tablespaces
There are any number of ways to find it out --- read the output of
"pg_restore -s", or just try the restore and observe the errors.
Besides which, we are talking here about the output of pg_dumpall,
which is currently always text.
> Then bad news, ISTM that altering the tablespace of an index, a sequence
> or a schema is not implemented.
Wrong, unnecessary, and trivial respectively.
I see this request as being exactly on a par with requests to make
pg_dumpall output restore into a different set of databases, or
into a different set of schemas than what was dumped from. Sure,
it would be convenient sometimes. But it's not *necessary* and it's
not something to be starting in on when we're already well into beta.
Could we have less straw-man-bashing and more discussion of the minimum
necessary solution for this problem? It's long past time to be gilding
the lily for 8.0. You can give it a new paint job in 8.1, if you like.
regards, tom lane