GOMEZ Henri wrote: >>You can start with my doc, there are two references to >>JavaGroups in there >>http://www.filip.net/tomcat/tomcat-javagroups.html >>http://www.javagroups.com/ >> >> >>>- Does it use multicast or others broadcast techniques ? >>> >>Virtual Synchrony and Probabilistic broadcasting on top of IP >>Multicasting. >>see the docs >> > > Excellent stuff, something which could have its room in > jakarta !!!
JavaGroups is cool since it is pure, multiplatform Java, although (from what I know) it cannot fall back to TCP when multicast isn't available, and the license could be "better".. :) If you liked that, also have a look at Spread: http://www.spread.org/ The messaging engine (daemon) isn't written in Java, but it's very fast and efficient, and runs on all the popular OSs. The Java clients connect to it via TCP sockets, so the clients can be pure-Java. The License is similar to the Apache license, and unlike JavaGroups the engine does know to fall back to using TCP unicast when multicast is not available. It's been around for quite some time, too. What I'd like to do at some point is take Filip Hanik's TC4 session replication code (looking nice!) and make it switchable to use either Spread or JavaGroups, or other communication mechanisms for keeping the session data in sync. Pluggable messaging back-ends.. FWIW, I'd like to see the in-memory session replication code as part of TC4 itself, with a pluggable messaging layer API that allows a separate messaging system to be used. But, if people decide that it's better suited to j-t-c, then that's okay (not quite as good, IMO), but I'd still like it to have a pluggable messaging layer API. Cheers. -- Jason Brittain <[EMAIL PROTECTED]> 650-228-2644 CollabNet http://www.collab.net -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
