> Fortunately, I think the change schema default tablespace is unusual
> enough that we can wait for 8.1 where people will really want the
> ability to move schemas for already-created objects too.
Ok. I disagree;-) Or one has to consider that tablespace is just
"experimental" and should not be used but for simple tests. If so, maybe a
warning should be output everytime the feature is used...
ISTM that most people don't need tablespace at all. So it is ok for them.
However, those who might consider using tablespace because they need it
will have somehow a lot of data. Their large amount of data is likely to
be precious to them. If they are messed up because they cannot move them
around as they need it, they won't be happy. Saying things like "you can
just pg_dump/pg_restore" or "edit pg_dump output by hand" won't make them
very happy either, esp if the amount of data is huge, and it may well be
the case if they use tablespace.
It seems to me that the current implementation is not completed. For
instance, do you expect someone to be able to remove a tablespace? At the
time it is not always possible, because all objects cannot be moved away
from a tablespace (namely sequence, as it is considered useless to move
them, what looks true from a performance perspective, but not from an
administrator perspective).
So it seems to me that tablespace handling must be looked at very
carefully from a "big data" user perspective and their need.
Well, as usual, I may be quite wrong;-)
Have a nice day, thanks for your answer,
--
Fabien Coelho - coelho@cri.ensmp.fr