I think I in that case would choose another design. For example, add a deleted column in the table and then have a scheduled task that would "physically" remove all rows in the database marked as deleted each day.
Could your solution perhaps work if you update the row through a query?
Anyway, I don't really agree with you regarding it as unwise to skip the updates. I see it as I'm working with data loaded from persistent storage and then finally persisting it again at commit, not as performing all updates directly in the persistent storage.
Someone else might have an answer to the last question.
[Message sent by forum member 'woel' (woel)]
http://forums.java.net/jive/thread.jspa?messageID=257029