I've been watching the project-formerly-known-as-echo from the sidelines so far. My only contribution has been the observation that a new format can't avoid the politics because those are driven by economics rather than technology. That aside, I've been following the various discussions with interest. There have been some very smart suggestions made and some insightful observations on the nature of XML format design. However, there seems to be a real rush towards spitting out XML rather than concentrating on the usage model. I just read a posting by Kellan, which picks up on this too.
Since becoming test-infected I've been reluctant to dive straight into the code, instead thinking first about how clients of the code are expected to interact with it. I think this approach can be applied to XML development too, perhaps by specifying XPaths that should always supply the same information throughout the various iterations of the format. These become contracts between the format developers and the format consumers. It could work.