This article original from Http://blog.csdn.net/voipmaker reprint annotated source.
The WebRTC-Peer part uses the ice framework, the ICE framework includes the Stun,turn, and one problem with the actual use of WebRTC to develop audio and video applications is that calls are built very slowly because the ice process takes too much time for the client to communicate with the stun server before initiating the call The server reflexive address and the relay address, together with the local host address and port constructs class three ice candidate, puts the three classes of candidate into the SDP attribute (a=) and completes this action before actually initiating the SDP offer request, the recipient Using the same process, both sides of the complete collection of the other side of the SDP before the start of peer detection, the middle of time wasted in all candidate are acquired before sending, so in order to speed up the call setup time, the connectivity detection time ahead, the program called Trickle Ice, The idea is that the client side collects candidate side to send to each other, for example the local candidate does not need through stun obtains directly can initiate, this reduces the time that the connectivity detection completes, in order to increase the call to connect the speed, WEBRTC also introduced the different type media (audio, Video) multiplexing, which is the use of a UDP port, which is represented in the SDP as A=group:bundle Audio Video
Indicates that the client supports the multiplexing of audio and video RTP.
WEBRTC series of topics Trickle ice