> That has caused a "certain amount of grief" around here, so I'd be
> agreeable in at least general terms to this.
>
> This approach has the further merit that doing a pg_dump on a subscriber
> node gets results that aren't "hacked up."

The biggest crevate I see to this methood is if you were to DROP SCHEMA 
clustername CASCADE; on a subscriber you would loose all of your original 
triggers, rules and constraints. Though I don't see this as an big issue 
unless you are to lazy to properly failover/promote a subscriber to becoming 
an origin, in which case I'm inclined to thing you are SOL anyway..  Another 
minor point is having a form of double data validation  incase there is a bug 
in the Origin version of PostgreSQL, but not in the subscribers version.



-- 
Darcy Buskermolen
Wavefire Technologies Corp.
ph: 250.717.0200
fx:  250.763.1759
http://www.wavefire.com