RTP
/ are-tee-PEE /
UDP is a wonderfully thin envelope — it gets a packet to the right program and not much more. For voice and video that is almost enough, but two things are missing: the receiver needs to know what order the packets go in, and exactly when each chunk of sound or picture should be played. RTP, the Real-time Transport Protocol, adds precisely those two pieces of information on top of UDP. Think of it as stamping each media packet with a track number and a clock reading so the far end can reassemble the performance in the right order and at the right pace.
Concretely, RTP adds a small header to each media packet carrying, among other fields, a sequence number and a timestamp. The sequence number increments by one per packet, so the receiver can put packets back in order and notice gaps (a missing number means a lost packet — the signal to invoke concealment). The timestamp records when that media was sampled, which is what lets the playout buffer schedule each packet's playout on a steady clock regardless of jittery arrival. RTP itself adds no reliability, no encryption, and no quality guarantee — it is a framing and timing layer, almost always paired with a control companion called RTCP, which periodically reports statistics like loss rate and jitter back to the sender so it can adapt (for example, lower the bitrate).
Why it matters: RTP is the workhorse format under nearly all Internet voice and video — VoIP, video conferencing, live streaming. A common misconception is that RTP 'provides quality of service.' It does not: it gives the receiver the timing and ordering information needed to cope, but the bits still ride best-effort UDP and can be delayed or lost. RTP describes how to label real-time media; it does not make the network treat that media any better. Encryption is added separately (for example, SRTP, the secure profile of RTP).
A VoIP sender emits packets with sequence numbers 100, 101, 102, 103 and timestamps spaced 20 milliseconds apart. The receiver gets 100, 102, 103 — sequence 101 is missing, so it knows exactly one packet was lost and conceals it, while the timestamps still tell it when to play 102 and 103.
RTP's sequence number reveals loss; its timestamp drives smooth playout.
RTP does not reserve bandwidth, guarantee delivery, or by itself encrypt anything. It only labels media with ordering and timing; the network underneath is still best-effort, and security comes from a separate secure profile.