On Sat, Jun 21, 2008 at 11:24:37PM +0200, Cyril SCETBON wrote:
> 
> Don't you think Slony should check if an earlier transaction exists but 
> also if relations locked (in that transaction) are concerned by the 
> replication ?

There's no way for Slony to know this.  I agree it would be nice, but it's
simply impossible.  This is a consequence of the way MVCC in Postgres works.

(I'm oversimplifying, note, because there are a few tricks in 8.3 and up
that have solved some of these problems.  But that won't help you.)

A