Rupert Goodwins' Diary

Thursday 27/07/2006It ain't half hot, mum. If we're having problems, think of the poor machines.

Thursday 27/07/2006

It ain't half hot, mum. If we're having problems, think of the poor machines. Last weekend, Yahoo and MySpace both fell off the planet for many hours because of problems exacerbated by power failures due to excessive heat-related demands. And today it's London's turn, with a power cut knocking out places like Oxford Circus and VNU. How are they going to edit those Wikipedia entries now?

It's hard to know how to make provision for power cuts these days. Not so long ago, you could get by with some sort of local backup like an uninterruptable power supply or generator. All your home and office communications would be by BT telephone -– and that had battery backup too, across the country. Even if the blackout were to last for ages, you'd have plenty of time to find out what had happened and decide what to do next.

Things are much more fragile now. You can still run your office or home equipment on UPSs, and the BT landlines still have battery backup –- not much help if you or the people you need to talk to aren't on traditional landlines. I don't know how much resilience is built into the mobile phone network in terms of power, but the system is known to run out of capacity when things get hairy. And IP services rely on each link in the chain staying up –- with no legal requirement for power outage survivability and nobody in overall charge, I doubt it's going to be robust.

As the experience of the London bombings on 7/7 showed, our modern communications systems are not to be trusted. We shouldn't be surprised: after all, we've entrusted them all to companies who cut corners and work on guesswork even when they're commercially exposed to the results: how much more likely is it that systems never tested in anger or independently, expertly audited will work when we need them?

There is an alternative -– the infrastructure-free network. A wireless mesh system is independent of central control but is both robust and self-configuring: put a load of nodes in an area, and the total bandwidth there goes up just when you need it. Such things have only been developed for specialist uses, because it's difficult to make money running something that doesn't exist. And without commercial demand, there's no pressure on the regulators to provide spectrum.

However, there's currently an interesting anomaly for rent. Between 30MHz and 80MHz are vast tracts of empty space that once belonged to old TV systems but are now fallow. There's not enough bandwidth for broadband data, the antenna aren't efficient and the radio propagation characteristics are variable: commercially, the frequencies aren't that hot. But the signals do go a long way on little power, and aren't easily put off by buildings and other obstacles.

If 10MHz' worth of this fallow spectrum were made available for licence-free low-bandwidth IP-based mesh networking -– a cross between Wi-Fi and CB, if you like -– there'd be plenty of opportunity for companies and experimenters to build inexpensive boxes that provided, say, 1Mbps links into the cloud. They'd be low power enough to run for ages from batteries, but have the range to get across cities (and further: if the idea was a success, point-to-point high bandwidth links bringing together areas of high density would be a natural sequel). The result would be a robust, free usage communications system that would be highly difficult to disrupt, built out with minimal investment and little risk, and one that overlaid our existing systems with a complementary set of functions that we badly need. In time, voice-over-IP handsets, PC cards and the like would follow.

It's hard to see how to make billions at it. That doesn't mean it shouldn't be done.