On Sun, Sep 05, 2004 at 10:08:54AM -0400, Jan Wieck wrote:
> >slave can be configured to read from the 3 locations 1, 2 and 3 and
> >simply aggregate the data received.
> 
> This sort of configuration is not planned, and I think the concepts 
> behind Slony-I will give you a really hard time to even incorporate it 
> on a theoretical base.

This is true, although you could probably do it by using different
tables in each source database and a somewhat complicated series of
views and rules to make it look like you're using the same table name
in each.  So, for instance, you have a table called replicated_table. 
The table is actually a view of replicated_table_loc_1 and inserts
happen in location 1 via a rule which actually puts the data in
replicated_table_loc_1.

On the aggregate system, replicated_table is a view of the union of
replicated_table_loc_1, replicated_table_loc_2, &c.  You'd have to
write some tricky SQL to make the views perform acceptably fast, but
I think it could work.  

A

-- 
Andrew Sullivan  | ajs at crankycanuck.ca
In the future this spectacle of the middle classes shocking the avant-
garde will probably become the textbook definition of Postmodernism. 
                --Brad Holland