On Sep 30, 2010, at 4:33 PM, jule wrote: > > Hi Alex, > > you´re right, reloading after restart will get it back to work. Thanks for > your explanation. Is there a schedule for the patch you mentioned?
Our plan (at my work) is to get https://issues.apache.org/activemq/browse/AMQ-2948 accepted (to fix the 'only 1 ajax client per session' problem), and then submit a new patch to fix the 'GET & POST in different sessions' problem. The solutions to both of those issues require changes to amq.js, so the fixes need to be applied serially. > > For others facing the same problem: Meanwhile I found out that appending a > "?ClientId=YourClientID" to the GET and POST urls (for either REST-API or > AJAX-API [uses REST-API]) will prevent this problem from showing up. I guess > this will bypass the automatic id assignment by the broker (via cookie). It > seems to work very reliably now. Supplying a 'clientId' URL parameter is supported in the REST interface, but it's not (yet) supported in the ajax interface (MessageListenerServlet). I (and others at my job) have been working on a patch to fix this. I think we're nearly there, but I'm hoping to get some feedback on the patch I attached to AMQ-2948. If you supply a ClientID url parameter in an ajax request, do you see different results? It looks to me like the current ajax code (in MessageListenerServlet) will just ignore this parameter, but I'm interested if you've seen other results. What kind of test did you do? > > ActiveMQ is so fast that I can use it for a real-time application (an > interactive whiteboard). The broker in the middle allows my clients > (mobiles) behind NAT/PAT to be reachable (push via ActiveMQ HTTP long-poll). > I am very content with this now. Yes, we're getting good performance as well. alex
