Ian McCallion wrote:
>Philip Aston wrote:
>> The behaviour clearly requires a lot of ejbStore/ejbLoads to ensure
>> correct behaviour and consequently efficient implementations need very
>> tight intergration between the EJB server and database. There are
>> clear potential efficiency speed-ups if, for example, a container
>> knows it has exclusive access to a database.
>Disagree. Taking away Chip's examples, there is no need to do lots of
>ejbStore/ejbLoads. Correct behaviour is achieved if finders edit (i.e. add
and
>delete beans from) their results based on the beans that are already
involved in
>the transaction.
How should a container achieve the correct result in-memory, i.e. without
synchronization? In the case of BMP the logic of the finder methods is
expressed
(for example)in a sql statement. How should the container know if a dirty
bean
already involved in the transaction satisfies the sql condition?
I think with CMP an in-memory-strategy is possible to implement, but with
BMP there is no chance without synchronization. Or the specification must be
changed, for example there must be a mandatory boolean testXXX method for
each
findXXX method.
Dirk.
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
