Why organisations are right to ban Skype

Fascinating post by Bruno Giussani about the hidden underbelly of Skype. Excerpt:

Skype’s design is based on peer-to-peer, distributed networking principles. This means that the core functions of the system are decentralized, as is the database of Skype users (the tool that lets you look up other Sykpers and tells the system where to forward a call). The calls are set up and passed on among users, flowing through a chain of computers around the world without traversing any central infrastructure.

That’s good for robustness and scalability — and for Skype, which can avoid massive investments and add new users at near-zero marginal cost. For the system to work, however, some users have to take over its vital functions: routing traffic and holding portions of the database. In Skypeville, these tasks are farmed out to those users with the most powerful computers and the biggest bandwidth, such as CERN. Skype turns them into supernodes.

Only a fraction of users are elevated to this function–currently some 20,000, according to research presented at a recent conference in the Netherlands by Philippe Biondi and Fabrice Desclaux of EADS. And only a small portion of their bandwidth is supposed to be shared. Skype CEO Niklas Zennström explained it to me in an interview last year: “When you become a supernode you share some of your resources and a little bit of bandwidth, but very little; you won’t notice.”

But some do notice. San Diego-based venture capitalist and TV host Paul Kedrosky, for example, complained on his blog in January that while he was traveling his computer “was sending out enormous amounts of traffic.” The IT people at his firm discovered that the machine was routing Skype traffic as a supernode. Computerworld magazine found that “in supernode mode, Skype is reputedly able to saturate 100 Mbit/second connections.” In layman’s terms, those are fast connections. The average Skype user’s PC is much less taxed than this, obviously. The possibility of becoming a supernode is written into Skype’s end-user license agreement, but not explicitly: The word “supernode” is never used. The license speaks of “permission to utilize the processor and bandwidth of your computer for the limited purpose of facilitating the communication between Skype Software users.”

This brings up two considerations. First: Skype is using some people’s computer power and bandwidth at an amazing rate. Sure, they agreed to it when they installed the software. But since most people pay for their bandwidth, some of them may come to the idea of ask Skype to share the cost. Second: In the interview, Mr. Zennström–while acknowledging scaling issues–said Skype could basically grow indefinitely without the need for a central infrastructure. But as traffic grows, and should the current scattered grumbling by supernode users turn into more vocal complaints, Skype may have to start deploying its own supernodes. That would completely transform its business model.

Nicholas Carr adds some pithy comments of his own to this:

It’s a very nice set-up for Skype. By building its network on users’ machines and pipes, it’s able to “avoid massive investments and add new users at near-zero marginal cost,” Giussani writes. And in the past, its free-riding business model didn’t meet with much resistance. As a renegade operation – the Kazaa of telephony – Skype had an emotional connection with users that turned them into willing collaborators. But now that it’s an arm of a multi-billion-dollar profit-making company, eBay, one wonders if users will continue to happily make charitable contributions of processing power and bandwidth. Already, corporations are banning Skype from their networks.