Before WHIP, every WebRTC ingest service had its own signalling protocol — WebSocket schemas, REST quirks, vendor SDKs. An OBS plugin written for one service could not work with another without code changes. WHIP standardises the simplest possible signalling: the encoder POSTs an SDP offer to a known URL, the server replies with an SDP answer and a session URL, and media flows over the resulting WebRTC PeerConnection. PATCH and DELETE on the session URL manage ICE candidates and teardown.
RFC 9725 was published in March 2025 after several years as a draft (draft-ietf-wish-whip). By 2026 WHIP is supported by OBS (built-in since 30.x in 2023), Cloudflare Stream, Mux, Dolby.io, AWS MediaLive, Twitch, and most managed live platforms. It does for WebRTC ingest what RTMP did for Flash ingest — a single common surface area that every encoder vendor can target. The killer combination is OBS pushing H.264/Opus over WHIP to a managed service that converts to LL-HLS for distribution.
WHIP gives sub-second contribution latency where RTMP gives 2–5 seconds, with no port headaches (all signalling on HTTPS/443, media on standard WebRTC UDP). The cost is that WHIP requires WebRTC-capable encoders and ingest servers, both of which are heavier to operate than RTMP. The trend through 2026 is WHIP for new low-latency workflows, RTMP for everything that already works and has no latency requirement.
