2008/5/28 TLFox <[EMAIL PROTECTED]>:

> It strikes me that the JMS client must be sending the persistent non
> transacted messages non blocking. I.e. it is sending the next one without
> waiting for confirmation from the server that the last one has been
> persisted (which involves a network round trip).

Yes, you are correct.

> If JMS semantics are to be followed, then the JMS client must wait for such
> confirmation before sending the next one (in the same way as it must wait
> for commit to hit the server and a response to be received).

No, the JMS specification does not mandate this.

However I agree this is annoying (it is a result of the behaviour of
the underlying AMQP protocol version implemented by Qpid M2, since
publish is not a synchronous operation). I believe that the later
version of the protocol can support proper guaranteed delivery without
requiring transactions - maybe someone who worked on the M3 Java
client can let us know whether it supports both modes?

RG

Reply via email to