Re[2]: [Hibernate-devel] caching algorithm

[Doug C. <de...@fl...>]

>> Hmm. The before-select timestamp is the local time? If the database
>> uses versioning (or caching) internally, the select could return stuff
>> as old as the start of the transaction... which could be earlier than
>> after-update even when before-select isn't.

> aaahhhh of course. interesting. So we should use a timestamp from before
> the start of the session?

That would be better. I think the best you can do with local
timestamps is to record the commit time of the writer's transaction,
and see that it preceeds the begin time of the reader's transaction.

e