By James Gledhill

SIP is arguably the most important Voice over IP (VoIP) protocol on the market today. Unlike its supposed competitors — namely, H.323 and MGCP (Media Gateway Control Protocol) and its derivatives, which were specifically built to reproduce PSTN like voice conversations on an IP media — SIP was not built for voice. Rather, it was built for any type of “Session.” SIP was applied to voice after the fact. This openness of SIP, by not being built specifically for voice, gives it an ability to be adapted for many things, including, but not limited to, voice conversations. SIP is used for IVR (Interactive Voice Response) and ASR (Automatic Speech Recognition) products, voice and video conferencing, push-to-talk, instant messaging (see the SIMPLE standard), and a host of other product types. Additionally, SIP’s addressing scheme, which looks much like an e-mail address can provide a more user-based identifier, which is far more rich than the impersonal numeric addressing scheme of telephone numbers.

While H.323 is still widely used in some parts of the world, it is a forgone conclusion that it is a dying legacy VoIP protocol. There are a few non-standards based VoIP protocols being used and promoted — specifically by Skype and Cisco. But as for the standards-based VoIP protocols, SIP and MGCP are currently the predominant competitors. However, SIP and MGCP are as different from each other as POTS (Plain Old Telephone System) phones are from cell phones. While both are capable of making phone calls, they are very different in how they work “under-the-covers” — in architecture, capabilities of devices, interactions with servers, where intelligence resides, and so forth. Indeed, SIP and MGCP are very different paradigms.

MGCP simulates and tries to match the legacy PSTN network. PSTN phones connect to the phone network by constantly being connected to a single Class 5 Switch in a master-slave architecture. The Class 5 Switch controls every aspect of the phone and there is no interaction with the phone except via that switch. An MGCP device could be compared to the interaction of a dumb terminal connected with a mainframe computer. The device has very little intelligence of its own and relies on all communication going in and out of the mainframe, which controls all activities.

MGCP likes to make the case that because the phone or ATA is “dumb,” it requires less internal electronics and, therefore, should be cheaper than SIP phones. This can be somewhat true, but the capabilities of the devices are also diminished and all the processing is forced on the switch, making the switch far less scalable than in the distributed processing architecture of SIP. SIP could be compared to a PC connected to the Internet. The PC does not rely on an external mainframe computer to control its every action. With the PC, the processing is distributed among many computers, including the PC. The PC may act as a client connected to other servers, but can also act as a server for other PC interacting with it. The PC can connect with many different servers — not just a on a one-to-one relationship.

True SIP devices work the same way. They can connect with many servers simultaneously and even act as servers themselves, allowing other devices to connect to them. Although the capabilities of a SIP device (like a phone) may only allow a single “session” to exist at a given time, each SIP device is both a client and a server. Every SIP device contains a SIP “user agent,” which is made up of a user agent client and a user agent server. Two SIP devices can connect to each other without requiring any other component to be involved. (Note: Many service providers lock down the SIP device so it can only communicate with their server. While this may be a requirement imposed by the service provider, it is not a requirement imposed by the SIP standard.)

In contrast to PSTN and MGCP phones, cellular phones have much more intelligence built into them and their interaction with the “network” is not a master-slave relationship. Instead, it is a one-to-many distributed relationship, where a phone may interact with one switch now, but another switch a few seconds later. Cell phones and SIP phones have the concept of registering and re-registering a device location, which can change frequently or remain the same for extended periods. Cell phones and SIP devices work on a distributed architecture where the intelligence is distributed among a number of pieces in the network, including different servers/switches and the phones. The phone device plays a much larger role in the cellular and SIP networks than it does in PSTN and MGCP networks.

The cellular industry liked SIP so much that they have adopted it as the protocol of the future cellular networks as well as for the new cellular applications framework. So, while MGCP strives to re-model the legacy PSTN network, SIP is setting the standard for new emerging products to come. SIP is a protocol that will allow us to do things we have, until now, not been capable of doing.

Those companies that truly understand the SIP paradigm do not just think about making phone calls. They think about the possibilities above and beyond how things where done in the past. They make vastly different devices than those built for MGCP. There are a number of devices available to buy that were originally developed as an MGCP device and then later converted to use the SIP protocol. Many of these device manufacturers never made the paradigm shift from MGCP to SIP, so their devices — while using SIP as the protocol — still require all the intelligence to be on a single server. They do not move any of the intelligence to the SIP device.

At the end of the day, SIP provides the best path for future IP-enabled applications. By embracing and supporting SIP now, service providers will have a technical and operational advantage over providers that are still dependant on legacy protocols.