Re: Optimistic locking/Repeatable read

[Larry Blair]

I'm guessing you mean doing the updates in a separate transaction from the selects. That would reduce the window for issues, but changes from another transaction that occur during the updates transactions would still be an issue if the values used for the where clause come from a checkpoint created when the transaction is opened.

I read the page you referred to. It's a little less ambiguous than some of the other pages I've read and seems to indicate that the values used include changes from other transactions, but it says things like, "The snapshot of the database state applies to SELECT statements within a transaction, not necessarily to DML statements." Terms like "not necessarily" or "may" are not definitive.