Understanding SIP and RTP

To understand the following subsections, you should understand some things about SIP and RTP. First, SIP is a signaling protocol that controls the call with methods such as INVITE, BYE, and CANCEL. The SIP protocol includes in the INVITE request information about the session (audio/video/text) using a protocol called SDP (Session Description Protocol). The information contained in the SDP describes one or more media flows configured between two user agents.

A SIP proxy never participates in the media flow, thus it is media agnostic. In other words it supports calls with whichever media are specified by UA and gateways. However, sometimes a B2BUA (back to back user agent) such as mediaproxy can be installed at the same server to treat RTP audio (that is, NAT traversal mechanism). The SDP protocol works in an Offer/Answer model. The SDP offer is embedded in the INVITE request and the answer in the 200 OK response.

Example: Excerpt from Ethereal:

The packet described in the preceding figure is an INVITE request. The INVITE request embeds the SDP packet that describes the session information. We can see there the INVITE generated on an eyeBeam phone. It is offering to use the Codec G.729 at the UDP port 8558 for Audio (I conceal the IP address for security reasons). The attribute rtpmap:101 telephone-event/8000 describes the DTMF forward method being used (RFC2833). The other device, in this case a gateway, answers the offer in the 200 OK reply.