I went to a session on the future of i-names this morning. Drummond Reed started off talking about what they are now. DNS names abstract IP numbers. URLs, based on DNS, typically point to specific locations. XRI provides an abstraction layer on top of the URL. i-names and i-numbers are synonyms. i-names provide a semantic identifier and i-numbers are a persistent identifier. i-numbers are never reassigned, but i-names might be.
Having a non-assignable identifier ensures that I can't lose my identity (and the rights that go with it). Any synonym in the XRI namespace resolves to the same i-number and all of them resolve to the same XRDS document.
i-names are transferable, but i-numbers are not. There was some discussion of why this is so, whether it's enforceable, and if its even good idea.
Future uses include ePayments, vendor relationship management, telephony (managing incoming calls), open reputation, community management, embedded authentication, "break glass" (e.g. emergency medical information), abstracting contact information (e.g. like 911, 311, 511, etc. for the Internet) and so on. Of course, this same list could be made in an OpenID sessions.
I asked Drummond what distinguished i-names from other identitfiers. The big answer seems to be about the permenance of the identifier. This does happen with URLs. Domains get sold. This is why I recommend owning your domain for your blog. At least then you decided to drop the identity.
There's always a lot of interest in XRI and i-names at these events. People seem to sense that there's something there, but I never seem to get real traction on what I build tomorrow to use them effectively. Discussion with Drummond convince me that we're making progress on that point and I look forward to benig able to play with them easily.