The Future of Federation

March 2009 represents an important anniversary in the history of our industry. No - I'm not talking about Jeff Pulver's birthday or the 5-year anniversary of the Global Crossing bankruptcy. March 2009 represents the 10-year anniversary of the publication of the first SIP standard - RFC 2543, which was issued in March of 1999.

Anniversaries, if nothing else, are an invitation for a retrospective. It has been an amazing ten years for SIP. It started out as an academic technology, born out of the early Internet. It was the new kid on the telecom block, contending with more established
technologies that evolved from the telephone network. SIP's essential vision of doing communications the Internet way - of providing enhanced functionality and integrating with what the web had to offer - ultimately won the day. SIP went on to become the technology of choice for VoIP. Of course, the story is not over yet. The future remains uncertain. Both SIP and VoIP have a long way to go before we can decommission the last circuit switch. The final frontier for VoIP and for SIP is federation - interconnecting different organizations, whether they are enterprises or service providers. Today, the vast majority of this interconnection is circuit based. This is particularly evident within enterprises. Today, IP-based voice systems account for the majority of new telephony system sales, yet the vast majority of those still use circuit interconnection to the rest of the world.

There are two possible futures on the horizon. One, called open federation, is important for realizing the vision of SIP. The other, overlay federation, is more incremental but will likely stifle much of what SIP can offer. Open federation allows for true any-to-any interconnection, without any prior arrangement. If someone in company one calls a user in company two, that call can go directly between companies one and two, with nothing in between except IP. In an overlay federation model, company A and company B connect to respective service providers which offer SIP interconnection, and those providers in turn connect to each other, possibly other service providers, passing calls amongst each other. This model is an overlay since, not only does it ultimately require IP between companies A and B, but it also requires a SIP network to be built on top of IP, in order for company A to reach company B.

In many ways, overlay federation represents the natural evolution from today's circuit based interconnection to SIP-based interconnection. The circuits that exist today between two organizations (such as an enterprise and their voice provider) are replaced with IP and SIP, but the ‘Public Switched Telephone Network' still exists, just in SIP form.

Indeed, the PSTN becomes two intertwined networks - an IP network and the SIP network that runs ontop of it. In essence, the SIP network becomes a worldwide overlay ontop of the IP network.

In open federation, by contrast, the circuits that exist today are replaced by connectivity to the Internet. The Internet acts as the universal glue, allowing all forms of traffic to flow between any two organizations connected to it. That traffic includes voice and video, but includes everything else that domains need to exchange with each other, including mail, messaging, streaming video, web traffic, and future services we have yet to see. The power of IP - and the power of the Internet - lies in its ability to carry any kind of data anywhere in the network, without any dependencies in between. That crucial characteristic - the ability to add new functionality without changing the entire network - is the engine that has driven the amazing innovation that the Internet has wrought, and that same characteristic is the essence of the open federation model.

To make this concrete, let's use an example. Consider two companies, a.com and b.com. They both deployed a new communications feature called ‘virtual handshake'. With this feature, during a phone call, either party can press the ‘virtual handshake' button on their phones, and the systems automatically exchange business cards and save them into their respective corporate address books. This feature works by leveraging SIP's ability to carry new content - an electronic business card in this case. In an open federation model, for ‘virtual handshake' to work between a.com and b.com, only a.com and b.com need to implement it. However, in an overlay federation model, not only do a.com and b.com need to implement it, but a.coms provider, b.com's provider and any other service providers in between, all need to support it. In essence, the network itself - the infrastructure that interconnects everyone together - needs to support this new feature, rather than just the two endpoints participating in the session. Can you imagine what the web would look like if each and every new web innovation - Javascript, Flash, streaming video, and so on - all needed to be supported by the browser, the website, and by the ISPs in between? The web would still be the static HTML we had ten years ago.

And so, as we stand on the cusp of the future of VoIP, the path remains uncertain. Will the future of federation follow the Internet model, allowing the same kind of innovation we've seen on the web, or will it gradually evolve into the next-generation telephone network? It will come as a surprise to no one that I strongly favor the open model; it has been a cornerstone of SIP's design. However, there are many challenges in achieving it. In my next article, I'll outline the set of technical problems that need to be solved in order to enable open federation.