The June issue of Scientific America features a piece explaining network coding, a counterintuitive means of sending bits that could replace routers, make better use of existing systems and work better.
The thing is, it doesn't make any sense -- not to me, not even to the scientists who come up with this stuff. Who, for instance, would believe it would be more effective to send information about data than to send the data itself?
And that's exactly why it took them so long to come up with network coding: Nobody thought it would work.
The problem, it seems, is in the metaphor. The traditional metaphor for sending bits along a network is that information is like a car, and the network is the road it travels along. The routers are the traffic cops, policing the roads to make sure there are no traffic jams or accidents. Bits get sent down the "road" and they arrive at their destination. If they're going the same place, you have to work that out, just as you would if two cars were rushing to the same, one-lane bridge at the same velocity.
Except, as this piece explains, bits aren't cars. Bits don't even act like cars, really. But because we've bought into the metaphor, everyone's missed the reality: There is no road.
It reminds me of the the proverbial business puzzle with the nine dots and the dilemma about how to connect them all. Everybody thinks the lesson is "think outside the box." But really, that's not the lesson. The real lesson is: There is no box -- just nine dots and your assumptions turning the dots into an imaginary box.
Seven years ago, four researchers finally saw there was no box: Rudolf Rudolf Ahlswede of the University of Bielefeld in Germany and Ning Cai, Shuo-Yen Robert Li and Raymond W. Yeung, all then with the University of Hong Kong.
Scientific America did a much better job of explaining network coding than I could hope to duplicate here -- and it's still pretty darn confusing. You can read it for yourself if you're curious about the details. Or, you can visit the Network Coding Home Page for more information and links. It looks like it's 10 years old, but there are links from 2007, so someone's keeping it up. You can also read this primer from 2005.
For the rest of you, here's why network coding is so promising: