> Well, we could aim for something far more simple than the JSK? Something > with real simple exports. I've got a Exporter class somewhere, you only give > it the component name and the reference to the service implementation and it > works. I also got a ReggieStarter class lying around, which starts reggie > from within the program with minimal config. If we provide a global instance > of a Configuration object that produces a minimal config, would that be > enough to get people started?
There were lots and lots of efforts at "quick start" (open source) libraries for Jini, but these were all (to my knowledge) run as outside projects, often by 1+ persons; none seemed to gather momentum. I suggest that if we are going to offer a "quick start" we cast the net far and wide and see what is already available that we can learn from, and possibly get contributions from. Personally, I think some practical goals for what could be put into core would include annotations for declaring services and optional IOC/DI-based configuration, at least in service definition, in particular offering a means to use Spring/Guice/JSR 299 for configuration. On the client side, I could see value in a simple utility above the SDM, which builds on the caching already available and allows for transparent balancing across multiple service instances of a particular type and/or failover if a service instance disappears or has problems. For both service and client, some JMX beans would come in handy for monitoring the current state of the system. For new users, I think offering a couple of Maven archetypes would be useful, as it is a convenient way to define and share a project skeleton with everything downloaded and installed for new users. The major Java IDEs all support reading Maven POM files directly, which _really_ makes it easy to open up and get started with a foreign codebase. Interested in where this discussion goes. Patrick
