Interesting, though I'd like to see more detail - what a deadlock looks like in the error log and how to decipher it would be good. Ultimately you cannot avoid deadlocks entirely. Applications MUST trap for them and resubmit the query.

Andy