The PleiadesGeo software I've been developing serves up Atom feeds, and now it seems we can suck such feeds back in. I asked Stephan Richter at the Plone Conference whether Plone would benefit from standardizing on Atom. His take was that Atom is too generic, and that the applications developed by Lovely Systems needed richer and more precise expression. I'm still interested in seeing what we can accomplish using more generic structures.
I get to work on something closely related today. PleiadesGeo already serializes content to KML for display in Google Earth. Now we're working on opening up the inbound lanes.