On Thu, Jun 12, 2008 at 12:14 AM, David Glazer <[EMAIL PROTECTED]> wrote: > I'm not familiar enough with the details to have an opinion on the best > implementation, but I do have some thoughts.on process for resolution: > > - if I understand right, there's an already-committed, partially complete, > abdera-based implementation for format=atom requests > - by the end of the week, we expect there to be a new, partially complete, > non-abdera-based implementation for format=json requests > - at that point we have four options: > a) shift all our weight to the abdera-based path, adding any missing > features and json support > b) shift all our weight to the new path, adding any missing features and > atom support > c) continue with two parallel implementations, finishing each independently > d) press reset and find a new approach
Here's my $0.02: I think option (a) is the best choice. Having implemented AtomPub once myself (ROME Propono), I don't the idea of trying to whip-up a new AtomPub implementation for XML or JSON when such a complete, compliant, pluggable and well-supported one already exists right here at Apache. One thing I like about the REST API implementation with Abdera is that it offers many additional ways to plugin to Shindig. For example, instead of creating a Person object model in Java and then allowing Shindig to use reflection to turn that into JSON data, I would rather generate the JSON data for people directly. With Abdera, I don't have to implement the PeopleService at all. Instead I can implement a People adapter and bypass the whole Person object model entirely. I think making this work with Abdera will make both Abdera and Shindig much stronger products. - Dave
