I was talking with the team here today and we thought that it was a good idea to move the majority of the code from FeedServer to Abdera. We're not convinced that shutting down the FeedServer project is a good idea though. We think that keeping FeedServer alive as a packaged version of Abdera should be its role going forward. Our additions to Abdera like the Adapter Class have and will be moved to Abdera.
We plan on running a build on c.g.c that pulls from Abdera and generates specific runtime configurations. This could be a really easy to use, out-of-the-box server that has a few adapters chosen. This may turn out to be the reference implementation for the API server for Shindig or it may be a couple different feature sets. All changes to core functionality will be coded in the Abdera codebase. We assume this path will make it easier to be a downstream project, since we'll be depending on the jars and we can manage version dependencies. We have some tests that depend on jMock right now and we'd like those to stay in the FeedServer project until they can be ported to the Abdera trunk. The one thing that we're thinking we'd like to keep in feedserver is ongoing adapter implementation development. I know that James has said that he'd like Adapters to be contributed to Abdera, but I see a potential for clean separation. We're hoping, for example, that third parties would write adapters and not necessarily commit them to Abdera. Because we stabilize and simplify the interface, it's possible to treat an adapter implementation like a plugin that is packaged and distributed separately. It still may be worthwhile to have some basic ones, like the JDBC adapter in Abdera. please let me know what you think of this plan. davep
