> The interesting thing is that we do use Cached updates here
> and still run into this problem.
>
> We will try and see if there is any way that two updates can
> be done on the same blob record at the same
> time.

Yes, this is exactly why I thought it was a VERY BAD idea to restrict
updates because even when using the most transaction "light" mode of cached
updates you are still potentially going to hit the same record more than
once in an update batch.

This is totally aside from the record locking argument. Just because there
is a new record locking feature that apparently doesn't use dummy updates,
doesn't mean that they can eliminate the possibility of doing more than one
update to a record in a single transaction.

I'm concerned that there are some people working in the engine that don't
actually have a solid understanding of how the engine is being used...

Jason Wharton