Can Skype pull off a move to SIP?

Skype has announced it will embrace the SIP standard. Richard Chirgwin looks at what this means for telephony and asks if Skype can pull it off!

Before we dive in too deeply, there's one thing to understand: Skype for SIP is not, as it has commonly been described in the non-technical media, “a piece of software that connects PABXs to Skype”. Skype is presenting a SIP-compliant interface to the outside world, which is (as I'll discuss later on) a significant new aspect to its service.

To begin at the beginning: the world's highest-profile Internet voice provider is trialing a service which would significantly expand its business capability (presumably improving its revenue position as well).
The question is whether this will be significant for business users, who have generally not been the core of Skype's market; for telcos, which are already the target of “Skype's PSTN killer” speculation; or for Skype itself.
To tackle these questions, we first need to get under the skin of the announcement (or more accurately, what has been leaked and discussed).
What the market does not yet have is any technical detail. However, at the recent eComm conference, Mark Spencer gave some detail of the Skype offering. As creator of the popular Asterisk open-source VoIP PABX, Spencer has the profile to be a Skype insider as well.
Spencer told the conference that Skype is trialing a SIP gateway for its service – a significant development, since it means putting an open face on what is otherwise a completely proprietary environment.
SIP – the Session Initiation Protocol – is one of the key drivers behind the growing popularity of VoIP systems in the business world. SIP created a standard that described the signaling part of IP telephony. This meant software could be designed that replicated – and extended upon – the functions normally carried out by the PABX or key telephone system.
By opening up the process of setting up and tearing down calls over an IP network, SIP eliminated a pocket of proprietariness that held back IP telephony. Certainly, PABXs could use data services before the emergence of SIP – ISDN, Frame Relay and IP connections were deployed, for example, to connect systems in different offices. However, with signaling standards lacking, most such applications depended on having the same vendor's equipment at both ends of the link. The common signaling standard opens this up, since anything that can generate standard SIP messages can send those messages to another system with confidence that they'll be correctly handled.
The Closed World of Skype
Skype, however, has always used its own, proprietary signaling system. What is known about Skype signaling doesn't come from the company's own documentation, but from analysis conducted by outsiders capturing and analysing Skype traffic.
One aspect of Skype's proprietary protocol allowed the service to build up momentum and scale very quickly. Borrowing from other P2P technologies, Skype decentralised the processing tasks to the users' machines: the more people sign on to use Skype, the more processing is available to handle calls. This was a vital development. Since Skype was initially a free service with no PSTN calling capability, its user-side processing avoided the risk that popularity would kill the service either by flooding the servers or demanding such processing power as to become uneconomic.
And because Skype gained its critical mass relatively quickly, it has also been able to remain as a proprietary system. Its size made it more important for other vendors to conform to Skype than for Skype to conform to a standard – which is why so many hardware builders include a Skype client, even though it might be simpler and cheaper to follow a standard.
The proprietariness of the Skype client is also attractive to those whose business model has depended on some kind of walled garden. Mobile handset vendors are the obvious example: their key clients are carriers, who resist the idea of subsidising handsets with open VoIP clients that give their customers the chance to bypass their voice networks. However, a Skype client is a little less threatening, because it gives the carrier a way to construct plans around Skype capabilities without too far undermining their own revenues.
Skype and Asterisk
The problem is that this proprietariness has been an obstacle standing in the way of the business market.
It may have seemed that Skype was already a “force” in the business market: Skype contacts are already a common feature of business cards, and Skype already has a suite of business services.
But clearly, it was missing a point of demand. Vendors such as Spintronics offer SIP gateway boxes with SIP-to-Skype functionality, and there are also a few open source projects addressing the same market.
Skype's entry into this segment clearly offers an endorsement of its importance (albeit at the cost of attacking an existing market that had been supporting Skype as a business tool). Interestingly, while Skype seems to acknowledge that a SIP gateway is a useful tool, it's been somewhat overlooked by media commentating on the market, who haven't noticed the pre-existence of such gateways.
To understand how Skype might implement SIP support, there is an example at hand – Skype for Asterisk.
Asterisk supports SIP, but not exclusively: it also has its own signaling protocol (IAX, the Inter-Asterisk Exchange).
Skype for Asterisk is essentially a Skype client designed to accept signaling messages from the Asterisk IP PBX software. For those familiar with Asterisk, the Skype offering is presented to the Asterisk PABX as a “channel driver” (in the same way as, for example, PSTN channels and SIP channels are presented to Asterisk). In this way, Asterisk users can add Skype as a trunk, sending their Skype logins to the client.
The telephone user, in this case, would treat Skype like a dial-out code (so you might dial 0 to access a PSTN line and 9 to access a Skype trunk). Because Asterisk has some add-ons that support least-cost routing, it would even be possible to leave it to the PABX to decide which outgoing trunk to use for a particular call, based on whether the Skype trunks were available, whether a PSTN call might be free (because, for example, it's a contract inclusion), whether the call is long-distance or international, and so on.
As with the consumer world, using Skype as a trunk and adding Skype to the environment as a separate application allows the VoIP client to remain proprietary. Skype is following essentially the same approach with Asterisk as it has done with (for example) VoIP handsets, allowing its code to be embedded in a third-party environment.
Skype and SIP
In Skype-for-SIP, the company is presenting a gateway to the Internet that is SIP-compliant, so as long as the customer has a PABX that can talk SIP and (for now at least) uses the G.729 codec, Skype will be able to set up and tear down calls.
Why not offer SIP systems an installable Skype client? Because there is a host of SIP applications and environments, running on a similarly wide array of operating systems. SIP applications might run on Windows, Mac, Linux. The PABX might be a standalone system from a major telephony vendor, running an embedded operating system developed in-house.
Skype has decided to avoid such an open-ended development effort, and instead present a SIP interface to the outside world.
Will it Work?
This is the most significant aspect of the service, because it's such a fundamental departure from Skype's service history: users will be able to connect to Skype without a proprietary client. Instead, the user will attach over the Internet to a Skype-hosted interface and connect using their own software.
It's no surprise that this is a beta: it's a completely different business model from that which Skype has followed in the past.
To make its service more attractive to the business customer, Skype has decided to focus only on the service, and abandon the lock-in that has been part of its culture since the beginning.
Whether or not the service succeeds, however, is another question. The following questions are going to be critical:
  • Scalability – Since the Skype for SIP service doesn't seem to offload processing to the clients, its success as a business service will depend on Skype's ability to scale up to meet business demands.
  • Security and Privacy – Skype will have to ensure that the service offers a level of security suitable to business users. Skype for SIP users won't be vulnerable to Skype client vulnerabilities, since they won't be using the existing clients, but protecting the businesses' SIP logins will be critical.
  • Competition – If a business is already using SIP in its PABX, it's probably already a customer of a business VoIP service already. Businesses may well be prepared to add Skype as an additional outgoing trunk – but that may not make them willing to move from their existing VoIP provider to use Skype instead.
Finally, there's pricing to consider. When Skype was first introduced in the early part of the decade, it was entering a nearly-uncontested market. Its free on-network calls and (later) low-cost off-network calls were revolutionary.
In 2009, that's no longer the case. In Australia at least, traditional PSTN carriers have noticed the threat posed by VoIP, and many of their business services include various levels of free call allowances. In the meantime, the VoIP market has blossomed, and all VoIP carriers now understand the important of offering free on-network calls and cheap off-network calls.
VoIP is now a fiercely competitive market. Skype has the brand in the consumer world, but that doesn't mean Skype for SIP will be an automatic category killer.

Read more on Voice networking and VoIP