Further to the resolver discussion; I've removed the use of PROPERTY_AUTO_START since this is already available in the deployment layer as DeploymentUnit.getRequiredStage() Any processing done in OSGi, should respect what other apis choose for this flag, although AFAIK nobody else uses it? As far as I understand it, the purpose is to not "start" (resolve) bundles unless they are needed by classloading or something in persistent storage says they should be started. The setting of the RequiredStage is already done by the OSGiBundleStateDeployer when the bundle is installed, but there is currently no persistent state, so there is a TODO there. I've still yet to understand what all this extra processing is actually doing. http://www.jboss.org/index.html?module=bb&op=viewtopic&t=160839 I don't mean the choice of wrapper versus extension which is just a choice of elegance and maintainability. NOTE: This is very different from the lazy start feature of OSGi where we should leave the deployment in the CLASSLOADER/RESOLVED stage until somebody loads a class. What is happening here is things are being moved to the INSTALLED/ACTIVE stage out of the normal sequence. View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4265118#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...