Features

net.wars: Blackout in Egypt, or how to close the internet

They didn't even have to buy ten backhoes

The most fundamental mythology of the Net goes like this. The Internet was built to withstand bomb outages. Therefore, it can withstand anything. Defy authority. Whee!

This basic line of thinking underlay a lot of early Net hyperbole, most notably Grateful Dead lyricist John Perry Barlow's Declaration of the Independence of Cyberspace. Barlow's declaration was widely derided even at the time; my favorite rebuttal was John Gilmore's riposte at Computers, Freedom, and Privacy 1995, that cyberspace was just a telephone network with pretensions. (Yes, the same John Gilmore who much more famously said, "The Internet perceives censorship as damage, and routes around it.")

Like all the best myths, the idea of the Net's full-bore robustness was both true and not true. It was true in the sense that the first iteration of the Net – ARPAnet – was engineered to share information and enable communications even after a bomb outage. But it was not true in the sense that there have always been gods who could shut down their particular bit of communications heaven. There are, in networking and engineering terms, central points of failure. It is also not true in the sense that a bomb is a single threat model, and the engineering decisions you make to cope with other threat models – such as, say, a government – might be different.

The key to withstanding a bomb outage – or in fact any other kind of outage – is redundancy. There are no service-level agreements for ADSL (at least in the UK), so if your business is utterly dependent on having a continuous Internet connection you have two broadband suppliers and a failover set-up for your router. You have a landline phone and a mobile phone, an email connection and private messaging on a social network, you have a back-up router, and a spare laptop. The Internet's particular form of redundancy comes from the way data is transmitted: the packets that make up every message do not have to follow any particular route when the sender types in a destination address. They just have to get there, just as last year passengers stranded by the Icelandic volcano looked for all sorts of creative alternative routes when their original direct flights were canceled.

Even in 1995, when Barlow and Gilmore were having that argument, the Internet had some clear central points of failure – most notably the domain name system, which relies on updates that ultimately come from a single source. At the physical level, it wouldn't take cutting too many cables – those ten backhoes again – to severely damage data flows.

But back then all of today's big, corporate Net owners were tiny, and the average consumer had many more choices of Internet service provider than today. In many parts of the US consumers are lucky to have two choices; the UK's rather different regulatory regime has created an ecology of small xDSL suppliers – but behind the scenes a great deal of their supply comes from BT. A small number of national ISPs – eight? - seems to be the main reason the Egyptian government was able to shut down access. Former BT Research head Peter Cochrane writes that Egyptians-in-the-street managed to find creative ways to get information out. But if the goal was to block people's ability to use social networks to organize protests, the Egyptian government may indeed have bought itself some time. Though I liked late-night comedian Conan O'Brien's take: "If you want people to stay at home and do nothing, turn the Internet back on."

While everyone is publicly calling foul on Egypt's actions, can there be any doubt that there are plenty of other governments who will be eying the situation with a certain envy? Ironically, the US government is the only one known to be proposing a kill switch. We have to hope that the $110 million the five-day outage is thought to have cost Egypt will give them pause.

In his recent book The Master Switch, Columbia professor Tim Wu uses the examples set by the history of radio, television, and the telephone network to argue that all media started their lives as open experiments but have gone on to become closed and controlled as they mature. The Internet, he says there, and again this week in the press, is likely on the verge of closing.

What would the closed Internet look like? Well, it might look something like Apple's ecology: getting an app into the app store requires central approval, for example. Or it might look something like the walled gardens to which many mobile network operators limit their customers' access. Or perhaps something like Facebook, which seeks to mediate its users' entire online experience: one reason so many people use it for messaging is that it's free of spam. In the history of the Internet, open access has beaten out such approaches every time. CompuServe and AOL's central planning lost to the Web; general purpose computers ruled.

I don’t think it's clear which way the Internet will wind up, and it's much less clear whether it will follow the same path in all countries or whether dissidents might begin rebuilding the open Net by cracking out the old modems and NNTP servers. But if closure does happen, this week may have been the proof of concept.