What is SIP?
SIP — the Session Initiation Protocol — is the standard signaling protocol for establishing, managing, and terminating voice and multimedia calls over IP networks. Published by the IETF as RFC 3261 in June 2002, SIP replaced the earlier RFC 2543 specification.
SIP History
SIP was developed by the IETF MMUSIC working group. The original RFC 2543 was published in 1999 by Mark Handley, Henning Schulzrinne, Eve Schooler, and Jonathan Rosenberg. SIP borrows its request/response structure from HTTP and its address format (user@domain) from email, making it familiar to web developers and network engineers alike.
Signaling vs Media
SIP is signaling only — it sets up the call but does not carry the voice or video. Media travels over RTP (Real-time Transport Protocol), while session parameters (codecs, IP addresses, ports) are negotiated through SDP (Session Description Protocol) payloads embedded in SIP messages.
The SIP Trapezoid
The SIP Trapezoid is the canonical call flow from RFC 3261 Section 4. It shows two user agents (Alice and Bob) communicating through two domain proxy servers:
- Alice sends an INVITE through her outbound proxy
- The proxy forwards the INVITE to Bob's domain proxy
- Bob's proxy delivers the INVITE to Bob's phone
- Ringing and OK responses travel back through both proxies
- RTP media flows directly between Alice and Bob — not through the proxies
The trapezoidal shape comes from signaling following the proxy chain while media takes the direct path between endpoints.
Why This Matters
Understanding the trapezoid explains common VoIP issues: one-way audio occurs when RTP cannot traverse NAT even though SIP signaling works fine, SIP ALG on routers breaks the flow by rewriting headers, and a phone can register successfully but still fail to make calls if the media path is blocked.
Key RFCs
- RFC 3261 — SIP core specification
- RFC 4566 — SDP (Session Description Protocol)
- RFC 3550 — RTP (Real-time Transport Protocol)
- RFC 3263 — Locating SIP servers via DNS
- RFC 3264 — Offer/answer model with SDP