In 2000 investigators proposed a seemingly crazy idea for limiting logjams in communications networks. Called network coding, this potentially revolutionary approach replaces routers, which simply relay messages at intersections, with network coders, which send evidence about the incoming messages instead of the messages themselves.

Network coding is faring well in experiments, most of which so far focus on sending data across multicast networks, where all receivers need to get the same information simultaneously.

It promises to make the operations of many networks more efficient (increasing capacity without having to add hardware or bandwidth) as well as faster, more reliable and better resistant to attack.