>>> On Mon, Jan 28, 2008 at 10:36 AM, in message <3001(dot)1201538162(at)sss(dot)pgh(dot)pa(dot)us>,
Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> in general pg_dump's charter is to reproduce
> the state of the database as best it can, not to "improve" it.
Seems that I've often seen it recommended as a way to eliminate bloat.
It seems like there are some practical use cases where it would be
a pain to have to do a CLUSTER right on the heels of having used
pg_dump to psql.
This does seem like the right way to do it where a user really wants
to maintain the physical sequence; my biggest concern is that
CLUSTER is sometimes used to eliminate bloat, and there is no real
interest in maintaining that sequence later. I'd bet that people
generally do not alter the table to remove the clustered index
choice, so this option could be rather painful somewhere
downstream, when the sequence has become pretty random.
Maybe it would make sense if it was not the default, and the issues
were properly documented under the description of the option?
-Kevin