Original URL: http://www.theregister.co.uk/2004/06/15/voip_and_skype/
VoIP suffers identity crisis
Enterprise network or PC application?
Analysis Two very different people, from vastly different backgrounds, described two virtually identical ways of working at the VON show for VoIP in the UK. Both are appealing, but very different in price, technology and approach. And the key to the future of voice over IP, is perhaps in knowing which one is right.
It may shock you to hear that one was the irrepressible techie upstart Niklas Zennstrom, author of Kazaa, and co-founder of Skype, while the other was Brian Day, a senior figure within Nortel's hosting services division. Perhaps the experience they are preaching is about the only thing they have in common.
At the heart of this divide is the fact that they both think that VoIP is something completely different from what the other believes. Zennstrom says that voice over IP is just a PC application, it's not a network, it will not replace what the community fondly calls POTS (plain old telephone system), it does not need regulating, there is no monopoly in it and it's not really telephony. It's just, as he says, a PC application.
Day already lives in a world where he is helping both Wireless and Wireline telcos make more money and reduce churn. He preaches his company's hosted systems and said: "My own personal way of working has changed beyond all recognition using these systems, and hosted services, including VoIP, is one of those things that once you've worked with it, you suddenly can't live without." Day went on to describe how once he was headquarters bound and fixed to a single desk and that the mobile phone only partially released him from this, but now he takes his office with him everywhere.
"I can work from a hotspot, from a regional office, from home, I can be Wi-Fi connected or connected through a Hotel broadband line and everything's the same. I can talk to my staff, see them on video, I can detect their presence, email or instant message them. I can explain things using co-operative browsing. It's just like being in my office," said Day.
But Day was evangelising the big corporate vision, something that later that afternoon Zennstrom tended to avoid. And yet Skype looks increasingly like the Nortel hosted tools that Day was talking about. It has pictures now, video conferencing in the near future, if offers presence through IM, and VoIP calls that beat mobile phones for quality easily, and, when they don't suddenly drop out, approach or beat fixed line voice quality. It comes with buddy lists, and will shortly have whiteboarding and text messaging and its own API for third party add-ons.
You can work on Skype in just the same way that you work on the Nortel system and on many other corporately targeted systems, but there are differences.
You don't have anyone you can call if Skype doesn't work, apart from the handful of guys that work at Skype. With an enterprise VoIP set-up you can have someone look at your routers, do an examination of your network, check for incompatibilities between your SIP (session initiation protocol) clients, and your firewall or NAT server.
Your Softswitch, a server which acts like a VoIP PBX, might need configuring and you may have interference from outside your network and perhaps the policy on your session boarder controller, a multi Gigabits per second device, might need adjustment.
The corporate version of voice over IP, and all the digital additions to telephony that can be added around VoIP, is indeed complex. It needs to be secure from attack. For instance what's the difference between an unsolicited email, and a phone call? And yet not only do you want an unsolicited phone call from a new prospect to get into the network, you want it treated with priority.
But how does your network know that a voice call on VoIP doesn't contain a virus if it bypasses the firewall, hence the border controllers that are becoming fashionable from companies like Nortel, but also Netrake, ACME, 3Com, Sonus and a host of others that are all members of the SIP way of doing things in the corporate world. And similarly how do you bridge in and out of an enterprise that still has legacy telephony systems side by side with new VoIP systems, and here more gateways are needed.
Zennstrom says he has the answer, don't use SIP at all, "Because it is a badly written protocol and it just doesn't work." Well that's a bit strong given that the entire VoIP industry has been built around it. What Zennstrom really means is that SIP doesn't work first time on every network and has trouble traversing firewalls and especially NAT servers.
In other words someone that understands networks has to make sure that your network is set up to recognise and receive VoIP packets and route them appropriately. And that, for the most part, includes more expensive, specialised hardware.
In fact this is really part of the Nortel message to operators everywhere, replace declining voice revenues, and those being lost to wireless telephony by offering managed and hosted services. Day says he has seen ARPU (average revenue per user) figures as high as $300 to $600, and on this basis the systems that he was pushing show a payback in under 18 months. Effectively he sees his role as helping to stop the operators go out of business.
The revenue pie
Zennstrom sees the situation differently. In the end he believes that pure voice will only be around a $10bn slice at the top of the revenue pie. The rest will be data, but included in that is voice data.
The interesting thing about the VON community was that Zennstrom's Skype, in the 10 short months that it has been in existence, has made it onto virtually every set of slides that were presented at the event. Skype was always in the corner, at the bottom, described as low end and not doing very much.
The next most frequent thing mentioned at the show was the Yahoo! Broadband service in Japan, where it has amassed 4 million VoIP subscribers across the broadband lines it sells there. An Alcatel spokesman described how Yahoo! employed people to give away home gateways for nothing, outside underground train stations. But no-one from Yahoo! was present to explain just what the economic model for it was.
The other thing that was mentioned "ad nauseam" was VoIP over Wi-Fi, a dream of many of the companies present, because a serious offering on this front could represent a threat to future wireless phone networks.
Wi-Fi voice would cannibalise the mobile experience and every speaker in a dedicated session on the subject listed the same reasons why it just wouldn't happen yet; quality of service needs to be embedded in Wi-Fi chips, and that won't happen until 802.11e is ratified, roaming needs to become automatic and ubiquitous, just as authentication needs to be managed on a SIM card and made foolproof yet simple. Battery life was also cited as a genuine concern here that would take several years to overcome.
The VON audience of mostly operator and equipment maker personnel were asked when they thought all wireless networks would interoperate and hand off to one another and they mostly voted for sometime in 2006/7. The were also asked if WiMAX would be important in this process, and only one hand went up to say no, with an overwhelming number aware that new wireless networks were on there way, built around IEEE 802.16 technology. British Telecom's BluePhone experiment was naturally discussed, and the fact that this contradicts every speaker's view in terms of obstacles and timeframe.
Bluephone is planned to bring the best of both worlds for both voice and data calls through the same devices, at broadband speeds. Whenever customers are within reach of a BT wireless access point in their home or office, they will be able to connect at the best available speed and quality, through the BT network. If they move out of coverage range, they will seamlessly link to a Vodafone cellular GSM or 3G network for voice and data, giving them the best available connection wherever they are.
Project Bluephone has undergone trials with 50 users over the past two months and the technology is now proven says BT. Now it is to be brought to market by BT in close collaboration with Alcatel, Ericsson and Motorola. A 'soft launch' involving more than 1,000 users is planned for this summer, with a full launch later this year. BT will include its own hotspots that it is rolling out in place of its fixed public phone boxes. Vodafone is pretty happy to do all this because it has weak network coverage in many residential areas and this will be taken up with home Wi-Fi networks and routed to the wires, and so this is really a way of getting BT to shoulder the work-load around residences.
BT in turn is facing as many as 20 per cent of UK homes not having any kind of fixed line and this is its way to stop the rot. At home, where congestion on a broadband attached Wi-Fi network is under the customer's control this is fine and it will work. At the office where a wireless LAN can be set up with an expensive proprietary quality of service protocol on a central switch, this will work. But out in the world of hotspots, the Bluephone will probably just make Vodafone money, for all the reasons that the VON discussion group highlighted.
As a result of so many references to Skype and in turn to the key protocol of SIP throughout the days at VON, there was naturally a lot of attention on Zennstrom once he began to talk.
If Skype is a revolution, Zennstrom is the revolutionary. It uses the Joltid peer to peer network which he also wrote, and this allocates certain logged-on machines in the Joltid network, as supernodes. These supernodes can retain elements of key encrypted information, such as the directory of 13 million Skype users, in a distributed data-base, complete with redundant copies.
Switch off a supernode and more are re-established. As a result Skype can give away free telephony because it has no central server overheads, does no switching, and needs no central billing. It will launch a prepaid Skype-out service soon, whereby a discount calling service will offer Skype customers a cheap way of letting their phone calls emerge into POTS.
"Not all calls can be to another computer and someone you know," said Zennstrom. "You have to call a taxi or order pizza sometime." Zennstrom also said that he would be giving out phone numbers to people in most countries which were local numbers to use with Skype, so that people on POTS could call Skype numbers. Would he make Skype open-source? No - that would make its strong 1024 bit encryption and security vulnerable: "We could do it but only if we re-engineered the way it works and we don't have the time right now." But he would open an API.
Zennstrom pointed out that for making money Yahoo! had an ARPU of about $9.5 per annum and that his approach would be similar. Then the key question was asked, when would he make Skype inter-operate with SIP? "If you interconnect to SIP based systems, the chain is only as strong as its weakest link and I don't think we will do that until SIP systems are better."
Quality of Service
Cisco's Rohan May didn't quite jostle Zennstrom at the end of his speech, but you could tell he wanted to and demanded to know when Zennstrom would build a connection to SIP. He said Skype had no Quality of Service and effectively challenged Zennstrom to some kind of demonstration, Skype against SIP.
Privately Zennstrom said: "It isn't when Skype is going to be compatible with SIP, but the other way around. I wrote Skype partly because SIP was such a bad protocol. It can't tunnel through a NAT server, or firewalls."
He wasn't the only one to say this, with most people content to say it in whispers over coffee, that most versions of SIP were incompatible and that it required a re-engineering of enterprise networks to get VoIP calls through reliably.
But if that gives a potted view of the two worlds of VoIP, the personal and the enterprise, there is a third world that is the preserve of telecommunications operators only.
British Telecom coincidentally made the announcement of its future IP architecture on the last day of the conference. BT's network will go IP end-to-end, creating massive savings, and it will also begin trialing fibre to the premises, echoing recent moves made by Telcos in the US.
There will be large scale migration of voice services from 2006, the first stage in the migration pilot will involve the bypass with multi-service access nodes of the core PSTN network link between two major network nodes at Cambridge and Woolwich. An extension is planned later in London.
And if calls are packet switched instead of circuit switched throughout the network, then this will give BT and just about everyone great advantages when it comes to moving VoIP data around the country. And just what will the regulators say about a national telephone network that is really a set of servers and softswitches with IP traffic. It will certainly make for interesting times and enterprise telephony will plug more easily into the future BT network, complete as it will be by 2009, as long as that enterprise traffic is IP ready when it leaves the building.
Remember BT's telephony won't be on the Internet, it will be on a private Internet called British Telecom's network. So is VoIP a PC application as Zennstrom says, or is it a set of enterprise servers and services, as Day believes, or is it simply the telephone network? We're inclined to think it's all three.
© Copyright 2004 Faultline
Faultline is published by Rethink Research, a London-based publishing and consulting firm. This weekly newsletter is an assessment of the impact of events that have happened each week in the world of digital media. Faultline is where media meets technology. Subscription details here.