On Tue, 2007-01-30 at 23:45 -0600, Tony Caduto wrote:
> David Fetter wrote:
> >> That being said, I think it is a dumb feature. If you have data in
> >> one database, that requires access to another database within the
> >> same cluster. You designed your database incorrectly and should be
> >> using schemas.
> >>
> >>
>
> I would have to disagree, it's a feature that has been available on M$
> SQL server and the other commercial
> databases for years. It's hardly a dumb feature.
I think it's a more complex problem than most people think. For
instance, if I do:
/connect db3
begin;
set transaction isolation level serializable;
insert into mytable (id1, id2) (select a.id, b.name from
db1..schema.table a join db2..schema.table b);
what exactly DOES that transaction isolation level mean? We don't have
distributed transactions yet... I can see this creating as many
problems as it solves, at least in the short run. Especially if people
start doing updates based on cross db joins.
