I'm fairly neutral about OpenID, and I think the spec itself is fairly explicit about not attempting to solve the problems listed herein. That said, there's a ton of real-world, right-now problems to tackle here.

"allows users to share their own personal Attention Profile in much the same way that OPML allows the exchange of reading lists between News Readers"

a Rails-centric approach to OpenID implementation

not entirely sure i'm ready to subscribe to this

I kind of like this.

everyone wants to outsource their auth