> One thought I had about replication is base it off of the WAL files. As
> they are created, they can be forward into other servers (both on a LAN
> and WAN environment) and imported to mirror the originating database.
> (As opposed to having triggers that force writes into replicated servers
> on updates, inserts and deletes)
In that case, we would have a master / slave environment. Also, how do
we guarantee that the slaves get the info and are able to commit it? Next
problem: a wal has all tuples changed. though "delete from foo" is only
a few characters, the WAL would be however many rows are in the database
(that were deleted).
That said, however, we are looking at using some of the same methods
that are used to generate the WAL info to use for packaging info to go to
other hosts in the network (for pg-repl).
- Brandon
----------------------------------------------------------------------------c: 646-456-5455
h: 201-798-4983b. palmer, bpalmer@crimelabs.net pgp:crimelabs.net/bpalmer.pgp5