Good point. With the PKs already isolated, it shouldn't be difficult to get a subset of any rows that didn't insert.

It's probably paranoia on my part but with SQL it seems anything that can go wrong usually does -- in spades. For most of my clients, I have no direct access to the data once the app goes into production. The DBAs are physically remote, always inaccessible and often surly so I try to make the application "self-correcting" if possible. It's easier than waiting a week for a phone call from someone who won't understand what the problem is anyway <g>.

Thanks for the suggestions David.