PBS x SGE comparison?

[hanzl at noel.feld.cvut.cz]

Is there any comparison of PBS and SGE?

After reading tons of documents on both, I still feel to be a newbie
and unable to decide which way should I go (and take all my cluster
users with me).

I would really welcome any real-world experience - what can be done
with one of them and is hard with other, how do you feel about
source code quality of opensource versions, what do you think about
long-time prospects of both given structure of support it has now,
what real-word tasks made you to abandon one and use another etc.

I am about to port most of our speech recognizer training tools to one
of these - and would like to decide for few years to come. Please help
me to decide if you can.

Thanks

Vaclav