Thanks Aleksey, that explains it. Unfortunately it means that extended functionality of SQL DELETE and UPDATE is pretty much unuseful without FLOCK and should be avoided in general purpose apps. I for one would avoid it.

I do have one more question on this. In my experimenting, I set REPROCESS to, say, 3 sec and did *not* receive an error when the record in the "filter scope but not in the join scope" was locked by another VFP instance. This is *not* how locking usually works, as it would normally report an error when it "really wants" to lock a record and can't. Is this by design? Or am I seeing things?

--Andrew