- stabilize rest of WebRTCPeer
 - use uid for the datachannel label
 - store a consistent WebRTC config in the dialer

copy-paste rendezvous works again, but with new interface allowing seamless recovery for the first time

to more sensible locations

maxedChan

Serene Han authored Mar 29, 2016 and Arlo Breault committed Apr 05, 2016

webRTCConn close as part of its reset, instead of depending on external deferred close after receiving reset channel, allows valid disconnections even before use

Serene Han authored Mar 23, 2016 and Arlo Breault committed Mar 27, 2016

replace webrtcRemote with webrtcRemotes map & indexing, client multiplexing remotes confirmed working (#31)

defer conn.Close for simplicity and remove unnecessary goroutines, improve error handling (close #12)

 * This will induce the tor client to establish a new SocksConn,
   which in turn will dial a new WebRTC connection.

   To be improved upon.

 * Part of #12

retry client offers immediately instead of reset, combine sendOffer and receiveAnswer into exchangeSDP

only flush datachannel with nonzero bytes, which prevents Tor from marking the SOCKS for close (#12)

remove webRTCConn SendLoop and simplify Write without additional channel, as net.Conn is already safe

clean up ice candidate log message.
still need to debug the copy loop break.

log traffic bytes only once every few seconds, along with OnMessage & datachannel.Send counts, to prevent flooded logs

Minimum viable webrtc pt now exists.
(close #1)