Actually, Postgres doesn't have any great out-of-the-box solution for partitioning the database across machines. The usual suggestion is Slony, but that is no where near as robust and widely deployed as MySQL replication. The GPL licence for MySQL isn't really a problem for webapps anyway.
OTOH, Postgres does somewhat better than MySQL on a single box with multiple cores (it's fairly linear up to 8 CPU, which is much better than MySQL) - mainly because of the work Sun put into scaling it before they bought MySQL. (At least - that's according to some people from Sun who do a lot of performance work with both databases)
It's not as robust? Do you have a source for that? You're doubtless right about it not being as widely used.
I've always preferred the fact that Postgres tried to do things correctly. Most recently, I bumped into this with Mysql, and it reminded me why I get irritated when I use it:
Postgres does somewhat better than MySQL on a single box with multiple cores ... mainly because of the work Sun put into scaling it before they bought MySQL.
That's not really the case. Sun has some some benchmarking, but most of the work on Postgres SMP performance was done by others (mostly Tom Lane, who works for Red Hat, and various EnterpriseDB employees).
OTOH, Postgres does somewhat better than MySQL on a single box with multiple cores (it's fairly linear up to 8 CPU, which is much better than MySQL) - mainly because of the work Sun put into scaling it before they bought MySQL. (At least - that's according to some people from Sun who do a lot of performance work with both databases)