I took a first whack at an Atom extension for describing activity streams. The format described therein is the format expected (and, a few funky bugs notwithstanding, generated) by my experimental activity streams library for Perl.
There's definitely a lot of editorial cleanup to do, but I'm not doing that right now since I'm anticipating that this'll be changing quite a lot once some more folks throw their two cents in. Already there's discussion on the Activity Streams Google Group about using
atom:category for the verb and object type annotations in preference to custom elements; I intend to do some testing to see whether existing feed processing software acts well or acts badly with the various alternative serializations — including the one in my draft — since it's important that these feeds do something sensible when they turn up in traditional feed readers and other feed processing software, else there will be reluctance to add this stuff.
There's also discussion about alternatives to my proposal of nesting an
atom:entry-like structure inside the activity to describe the activity "object". A valid concern is that applications that consume and re-publish feeds are likely to drop unknown extension elements on the floor, so it would be good to find a way to behave well in this case.
I encourage folks who are interested in contributing to this specification, whether in the form of spec feedback or in the form of experimental implementations and testing, to join the discussion on the Google Group. I think Chris Messina will also be doing a talk about this topic at the Learning About the Open Stack for the Social Web event on Dec 19th.