I think we need to make the important distinction between three different issues - format, semantics and the distribution framework.
Personally I see PubSubHubbub as the distribution framework, independent of format or semantics. Just like you have different formats and semantics available on the web (HTML, Atom, JSON, etc), you could have different formats and semantics being distributed through the framework. The issue regarding formats and semantics could be a separate debate on its own. I would definitely like to see Github, Google Code and perhaps others like SourceForge etc all standardise on the same format and semantics for commits so you can process all of them using the same piece of code generically. But this has to be handled on a per-case basis. It would of course be impossible to anticipate every possible use case. :) The publisher could of course make multiple feeds available - one (or even more) for specialised formats+semantics and then a separate one for generic feed readers.
