For large instance, the size of a cluster infos event is quiet large
(1MB) and is sent every second. If these events are queued, it can takes
a lot of RAM. The time limit ensure that such events don't stay in queue
if they are not processed quickly, as they are all the same, one in the
queue is enough.

OO-2041: upgrade hibernate and infinispan, update mapping to use TableGenerator with legacy-hilo optimizer

OO-1615,OO-1616: remove dependency to Spring ORM, remove hook between OpenOLAT and Spring Transaction, make the dependency path to DB from LockManager clearer, refactor LockManager

OO-1635,OO-722: refactor the ajax by iframe to use XHR, push the current business path to the browser via HTML5 push state, implements back button support via HTML5 History API...

OO-1550: remove agency without listeners, do all the synchronization of the listeners with the infocenter map

OO-1443: upgrade JUnit ( long overdue ), remove deprecated junit.* packages, fix some issues with order dependent tests, fix an issue with empty ClusterDoInSync

OO-1424: remove subscriptions if a member is removed from a group or a course, or a group is removed from the relations to a course

OO-756: check the key of the owner of lockentry instead of user name as the user name can be changed after the user be deleted

OO-592: remove tx manager annotations scanner, move configurations in resources/serviceconfig to java/... where they are needed, use import instead of the *Context.xml which allow deployment on weblogic and some others small fixes