By Joel Snyder
Network World, 05/10/04
Original Article on Network World Web Site
SIP, the Session Initiation Protocol, is a control protocol for multimedia sessions. Although most SIP software and hardware is designed for the simple case of VoIP telephony, SIP is actually a generic protocol that can be used to create multimedia conferences with voice, video and other data streams (such as instant messaging-type applications). SIP doesn't actually transfer any of the voice or video data. Instead, SIP is used to set up a session, and another protocol (Real-time Transport Protocol, in the telephony case) is used to send the voice.
End points, also called User Agents, are the phones of the SIP world. Anything that sits at the other end of a SIP session is considered an end point. For example, a voice mail system acts as an end point. The most common SIP end point is going to be a hard phone: a box that looks a lot like the phone on your desk, but with an Ethernet port instead of a two-wire analog phone line coming out the back. Hard phones run SIP software, have IP addresses, and need a fairly hefty and complex configuration to survive. In addition to hard phones, two other important kinds of end points in SIP are analog telephone adapters (also called FXS gateways) and soft phones.
Soft phones are simply software versions of the SIP phone, typically designed to be installed on a PC, Macintosh or PDA device. With a soft phone and an inexpensive headset, you can turn your $3,000 laptop into the SIP version of a $10 phone.
An FXS gateway, or analog telephone adapter (often written as ATA), is a device that allows normal two-wire telephones to be connected to a SIP network. FXS stands for "foreign exchange station" and is an old telephony acronym used to describe what most of us consider a plain old telephone service line: something with two wires to which you connect a telephone. The FXS gateway is essentially a box that has an Ethernet port on one side, a two-wire analog telephone jack on the other, and SIP running in between. FXS gateways, such as the Cisco ATA-186 or the Multi-Tech MVP210, have become very popular in residential VoIP systems because they let you hook your existing analog phones to the digital SIP network.
SIP servers are systems that help phones talk to each other (and other end points). Technically, there is no such thing as a SIP server. Because SIP is a decentralized protocol, the traditional PBX has no direct VoIP equivalent. Phones can and do talk directly to each other for call control and voice traffic, and functions such as directory services and call control can be highly distributed. This makes it difficult to know what to call a system that does provide PBX-like SIP services, because a server might have a combination of registration services, call redirection, and call control functions. Where the exact function isn't important, the term "SIP server" has come into common use.
The two most common types of SIP servers are the registration server and the proxy server. A registration server receives and collates information about phones, helping to map from SIP addresses (such as an extension number or a SIP URL) out to IP addresses. The proxy server normally receives incoming and outgoing call requests on behalf of a phone. This lets the more sophisticated tasks, such as ringing multiple phones, dealing with DNS and Enum, or accounting, be pushed out to the proxy server, making the phone simpler, faster, easier to manage and less expensive. In many VoIP networks, the proxy server and registration server are the same system.