One word Jeff....CLASS

Thank you for the very clear article and all the testing done to make sure that what you claim is 100% correct. But I wouldn't be me if I had not at least some comments, so here they come:

1) In Figure 28 I personally found the switch from the 3 part assignment problem, solved by using 2 part assignment, straight to a "this is the correct way" which shows 3 parts again...confusing. In fact I think I still don't get the reason why one works and the other doesn't (and I would go 2 parts all the way because of this).

2) At the end when talking about a trigger method to keep a running total correct at all times, you state that it might not work well with batch updates. It is hard to argue with a 'may not work well' opinion, but I just want to state that I can think of a way that could in fact work well.

What about making some sort of temporary Delta/Offset table first for those records that changed and then use this as input in a batch update that corrects all depending records with the appropriate offset without using a pseudo cursor.

Its just an idea, I never tested it and cannot guarantee it runs into unforeseen problems either.