Although two SIP devices can talk directly to each other, they generally will use an intermediary system that acts as a SIP proxy.  Note the SIP proxy only participates in the SIP messages—once the call is set up, the phones send their voice traffic directly to each other without involving the proxy.  SIP proxies are very helpful in offloading tasks and simplifying implementation of end station telephones.  As an example, a SIP phone might want to make a call to another phone at extension 4094.  Although the phone could have some magical way of translating that 4094 into an IP address or location, it typically will simply send its call request to its own SIP proxy.  The job of the SIP proxy is to 20ERR_DNS_FAIL know what “4094” really means: Is that a phone?  An auto-attendant system?  Perhaps several phones, 20not all to be rung at once?

We normally think in telephony of phones as having numeric addresses.  In SIP, an end station has a SIP URI (a form of URL) that identifies it and is used in the SIP protocol.  Because phones generally have numeric keypads, the phone is responsible for translating what you dial (such as extension 4094) into a SIP URL (such as sip:4094@sip.ilabs.interop.net). You can learn more about how SIP URIs, traditional telephone numbers, DNS, and IP addresses all interact in our white paper on 0ADNS “ENUM.”

The diagram below shows a SIP dialog involving two parties (Alice and ; margin-right: 2 Bob) and their SIP proxy servers, the Atlanta and Biloxi.  In this case, the SIP messages have been heavily abbreviated to show the flow of traffic.

Although the diagram here shows that the proxies do not participate in the SIP protocol once Alice acknowledges that Bob has picked up the phone, not every call will work that way.  A proxy may elect to “stay in the middle” of the conversation even after the 20ERR_DNS_FAIL&body=CacheHost call is connected to provide some mid-call features, such as conferencing services, 20exist. or accounting.  Note that even if the proxy is in the middle of 0AErrPage the call, we’re still only talking about 20Dec the SIP part of the call—the voice traffic will generally go directly from one phone to another once the call is set up.

Another common operation in SIP is called Registration.  In our example call, this might be how the Biloxi proxy learned where Bob was located.  The registration capability is especially useful in an environment where phones do not have static IP addresses (such as a DHCP environment or when a phone travels with its owner).  In SIP, the registration server can be co-located with the proxy server or they could be different systems.  Bob is also not limited to registering from a single location.  He could have SIP phones at home and at the office that both register with the SIP server.  Then, it is the responsibility of the proxy server to 5Bnone decide which phones to “ring” when a call comes in for Bob.  With SIP, that could mean selecting a single phone to ring, or just ringing all the phones at once.

Because SIP is used for call control, features such as voice mail and auto-attendant are not part of the SIP protocol itself.  Instead, they are provided by end points that send and receive calls themsleves.  This means that a VoIP network based on SIP has no real parallel to the “PBX” in traditional telephony.  You may hear the term “SIP Server” or “SIP PBX” used to describe the SIP proxy server, but the functionality is quite different.  However, it is possible to integrate some traditional PBX features, such conferencing into a SIP proxy server.  For example, the Asterisk SIP proxy server tested as part of our iLabs demonstration includes both voice mail and auto-attendant.  In other cases, such as a conferencing server with its heavy digital signal processing requirements, you might want a separate dedicated device.

To give you an idea of how simple SIP is, we’ve included a SIP message here: an idea of what Alice’s original INVITE to Bob might look like.   In this message, the Session Description Protocol (RFC 2327) part of the INVITE is not shown; SDP is where the voice traffic characteristics, such as choice of audio encoder, would be indicated.  SIP’s easy-to-read format has made implementation and debugging of SIP easier than other similar protocols, such as H.323.