This is neat, but I think if peercasting software which is "good enough" can be built on top of WebRTC data connections (mandatory encryption and forced TCP are at least two drawbacks) I think it will win in terms of public adoption if is inferior in other technical aspects simply because people will end up using it without knowing it.
That doesn't include discovery, tracking. It would have to be developed on top. It also necessitates a pretty much full implementation of the WebRTC client signaling, which is not too trivial.
BTW lately they are probably switching to use SCTP for data streaming. Encryption is still mandatory.
