Use unordered datachannel in WebRTC about weron

[NNdroid]

I think this should not be feasible. First of all, IP data packets are usually sent in fragments under tun/tap. This will cause the MTU of the tunnel to be a reasonable value, otherwise communication cannot be carried out. This will not happen when using strong parity flow. The problem is, it can be viewed as streams, not packets.

[ccaapton]

Stream over stream is just another version of tcp over tcp, which is very problematic. The user should able able to choose a smaller mtu, or rely on tcp mss discovery mechanism.

And why do we need strong parity on the first place?

[pojntfx]

I'm definitely open to taking another look at this once the transition to an asynchronous RPC protocol is done (see https://github.com/pojntfx/panrpc), but with the current RPC system this would not be possible.