On Wed, 2008-03-12 at 17:55 +0100, Andreas Kostyrka wrote:
> Some questions/hints:
> 
> 1.) What versions of PG are you running? Some version of PG (I think it
> was 8.2, but I'd had to check that) has become much more exact in
> processing UTF-8. If you are replicating from an PG version that is
> older than the version on the slaves this could be it.

Just to clarify, 8.2 was a little weird regarding encoding. It was more
strict in some contexts, but there were still ways to get invalid byte
sequences inserted, and after the bad data was there, you couldn't COPY
the data (dump/restore wouldn't work). That could also explain his
problems.

8.3 became more completely strict, so there is now no way to get the
data in there in the first place.

I'm not entirely sure this is his problem, however, because he said he
could dump/restore. Maybe he's just got different server encodings?

Regards,
	Jeff Davis