[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: xmBlaster vs CORBA Events clarification
Brett Palmer wrote:
> 1. My experience with CORBA event services and JMS has been that once a
> publisher submits a message the message is gone. Therefore, only clients
> that are listening at the time of the publishing event will receive the
> With xlmBlaster it appears that clients can login after the event and still
> receive the message. Does the xmlBlaster server ever delete these messages
> or is it the responsibility of the publisher to delete messages?
The current implementation is as follows:
PtP messages are queued
(see requirements engine.qos.publish.destination.ForceQueuing.
Publish/Subscribe queueing and timouts are not
yet implemented but scheduled
(see engine.qos.subscribe.OfflineQueuing, note the status=OPEN)
Not the publisher is responsible to delete old messages, but
xmlBlaster will do it, in the way the QoS is set.
It is up to us, to implement the features we believe are
useful. If you or anybody else does disagree with some
requirements, please discuss them on this list.
You can browse the current requirements online (see
'XmlBlaster Reference Handbook')
I hope to implement this feature after xml-rpc support and
stdio access (simple xmlBlaster access for shell scripts and
all other interpreter languages using stdout and stdin)
Others are welcome to implement it as well :-)
> 2. I have two small Java clients programs that publish and subscribe
> respectfully. The publishing client submits 5 different messages to the
> queue, and the subscribing client listens to just one of these messages
> using an XPATH query. The publishing client continually runs in a loop
> submiting the 5 messages synchronously. When both programs are running
> simultaneously the subscribing client gets all of the other messages even
> though he is only listening to one type.
Is your XPath query correct?
Otherwise it would be a bug in xmlBlaster.
Please send us informations (your publish key and XPath subscription)
to check it.
> If I stop the publishing client
> and start the subscribing client, the subscribing client gets the correct
> message that has been sent.
> Does anyone know why this would happen?
ruff at swand.lake.de