I remember one of the issues we hit with OSGI is that geode-core needs to be able to instantiate internal classes of other modules in order to deserialize them. Is that something that is going to be an issue with jboss modules? All of the documentation on classloaders for jboss modules is blank in their documentation - https://docs.jboss.org/author/ display/MODULES/Class+loaders.
Is using jboss modules going to break simple use cases of declaring a dependency on geode in maven or gradle and starting a geode client or peer embedded in an existing application? How are these modules actually bundled? -Dan On Thu, Mar 30, 2017 at 7:55 AM, Udo Kohlmeyer <ukohlme...@pivotal.io> wrote: > Hi William, > > I had looked at this. I liked the idea of JDK9's modular approach. With > it's GA data not too far off, it is a definite contender. > > --Udo > > > On 3/29/17 17:35, William Markito Oliveira wrote: > > Since the modularization effort may take a significant time and one of the >> big features of Java 9 is modularization, why not leverage the >> infrastructure of the JDK [1] for that ? This would not only bring the >> modularization but also make it 100% compatible with Java 9. >> >> It also provide solutions around class-loaders [2]. >> >> Food for thought... :) >> >> [1] http://openjdk.java.net/projects/jigsaw/spec/sotms/#defining-modules >> [2] http://openjdk.java.net/projects/jigsaw/spec/sotms/#class-loaders >> >> On Wed, Mar 29, 2017 at 3:29 PM, Jared Stewart <jstew...@pivotal.io> >> wrote: >> >> I have some questions about the ClassLoader isolation proposal. If the >>> class loaders of different modules are truly isolated, how can any module >>> load classes from a different module? And conversely, if different >>> modules >>> can load classes from each other, how can they truly be isolated? In >>> something like an application server, this seems much simpler, since the >>> wars that get deployed to an app server should never need to interact >>> with >>> each other. But how can this work with Geode modules? >>> >>> On Mar 29, 2017, at 10:00 AM, Udo Kohlmeyer <ukohlme...@pivotal.io> >>>> >>> wrote: >>> >>>> Hi there Guys, >>>> >>>> A more detailed proposal for a more modular Geode is available. Please >>>> >>> review and comment, either on this thread or on confluence. >>> >>>> https://cwiki.apache.org/confluence/display/GEODE/ >>>> >>> Geode+Modularization+-+An+approach >>> >>>> Udo >>>> >>>> >>> >
