I disagree at least a little bit, the fact that there's support on the browser side is much much more important than support on the server. I can write a server of arbitrary complexity given enough time, the scope of the problem is now one that I can fix.

I also found the following in my research yesterday, which looks like a very promising set of abstractions for WebRTC data channels.

https://github.com/geckosio/geckos.io