I've got a working implement ion a fix for this and a few other related issues here:
https://github.com/brozow/org.ops4j.pax.vaadin I've been a little slow committing it back and I'm sorry about that. We are using my implementation quite extensively in the OpenNMS project and it seems to work quite well. I would be happy to help get this one into shape but I don't want to step on toes since I've stepped away from it. Matt On Dec 7, 2012, at 12:55 PM, Scott Parkerson <sc...@parkerson.net> wrote: > [This is in response to a private email to Achim; I'm going to continue the > conversation on the OPS4J general list. --sgp] > > On Fri, Dec 7, 2012 at 11:07 AM, Achim Nierbeck <bcanh...@googlemail.com> > wrote: > > Regarding the Singleton approach, I'm not quite sure we are inline here. Pax > for Vaadin only registers a Servlet as a HttpService, Pax Web does the > connection to the Jetty HTTPServer world. Your suggestion would imply any > HTTPService implementation isn't multi-user capable, cause all are > Singletons. > > The problem is specific to the way Vaadin requires that a new instance of the > Application object has to be created for and stored with the session that > comes through the servlet engine. Otherwise you have the problem that was > described in PAXVAADIN-6, where you can load one of the samples, do some > stuff with it, and then move to another browser or computer and load the > sample and see the *exact* same application state! > > Afaik jetty had an issue to this and this was fixed in the past by pax-web. > > I did take a quick look at the code, and your suggestion in using the > Original Vaadin Servlet > https://github.com/sgp/org.ops4j.pax.vaadin/compare/master...PAXVAADIN-6#L1R62 > is surely gonna mess up with classloader issues. That's why the AppServlet > https://github.com/ops4j/org.ops4j.pax.vaadin/blob/master/pax-vaadin-service/src/main/java/org/ops4j/pax/vaadin/internal/servlet/VaadinApplicationServlet.java#L106 > does the same as the Orginal Vaadin Servlet. > > I didn't see any classloader issues, but then again, I haven't tried doing > anything fun like actually access something from another bundle within my > project yet. I'll revert those bits in my next commit and try to get to a > common ground. > > So I'm not sure about the current approach. We surely will need to discuss > this :) > > Neither was I. I felt that my approach was way to facile to be correct, which > is why I wanted to get in touch! > > Scott > > _______________________________________________ > general mailing list > general@lists.ops4j.org > http://lists.ops4j.org/mailman/listinfo/general
_______________________________________________ general mailing list general@lists.ops4j.org http://lists.ops4j.org/mailman/listinfo/general
