Latency, in the context of live learning, is the end-to-end delay between the moment audio or video is captured at the sender and the moment it is rendered at the receiver; it is the single most important quality parameter for interactive instruction. Conversational interaction requires latency roughly at or below 150 ms as a ballpark target — beyond that, participants begin to talk over each other, pauses feel unnatural, and the class starts to resemble a satellite phone call rather than a classroom. WebRTC is designed specifically for this range, using DTLS/SRTP over UDP to avoid the head-of-line blocking that TCP introduces; dropped packets are concealed by the media engine rather than retransmitted. Latency in a WebRTC session has several additive components: capture and encoding delay at the sender, network transit time, jitter buffer time at the receiver (which trades extra delay for smoother playout), and decoding and render delay. The jitter buffer is a key dial: a larger buffer absorbs network jitter but adds latency, so adaptive jitter buffering algorithms attempt to keep the buffer as small as the current network conditions allow. ICE and TURN traversal add a fixed overhead when a direct peer path is not available. For screen sharing and the whiteboard, slightly higher latency (up to a few hundred milliseconds) is often acceptable; for spoken dialogue between instructor and learner, keeping end-to-end latency under 200 ms is the practical engineering target.
