At 01:34 PM 1/14/00 +0900, Hiroshi Inoue wrote:
>I agree with Tatsuo though I prefer
> create table t1 (i int) tablespace foo;
>.
>Isn't it preferable to encapsulate the table location and storage type ?
Encapsulation is almost always preferable, but I don't think Tatsuo was
saying otherwise, merely pointing out a clever trick that I hadn't
though of (I didn't realize that "/foo" would be rooted rather than
just put under PGDATA, in fact I stay away from quoted non-SQL non-"normal
language" identifiers altogether, being somewhat interested in portability
of my application code).
And it is a clever trick...obvious to an insider, but not to me.
>At first,a tablespace would only correspond to a directory and it won't
>be so difficult to implment. But we would gain a lot with the feature.
>In the future,the tablespace may be changed to mean real(??)
>tablespace.
I think the decision to keep the mechanism for providing separate
storage mangers was a good one (if I understand correctly that there
was once consideration of removing it). Even if it is never used in
mainstream Postgres, some specialty application may use it, one of
the potentials that comes from open source software.
- Don Baccus, Portland OR <dhogaza@pacifier.com> Nature photos, on-line guides, Pacific Northwest Rare Bird Alert
Serviceand other goodies at http://donb.photo.net.