How the World Got Wired · Packets

“It Won’t Work”

The conversation between Paul Baran and AT&T is one of the great stories in the history of technology, and I want to tell it carefully, because you will see this pattern again and again in your life.

Baran went to AT&T’s engineers and explained his design. A distributed network. Messages chopped into small pieces. Each piece finding its own path. The pieces reassembled at the other end. He had the math worked out. He had the diagrams. He had simulations showing it would work.

The AT&T engineers told him it would not work.

They told him it would not work, essentially, because they were AT&T and he wasn’t. They had been building communication networks for almost a century. They were the experts. Networks were circuits. You set up a path from one end to the other, and you kept that path open. That was what a network was. What Baran was describing — pieces of a message running around independently, the network deciding paths on the fly — wasn’t a network. It was chaos. It couldn’t possibly work.

Baran asked them to at least try a small experiment. They wouldn’t. They didn’t want to. They had a perfectly good network business that worked the way they had always done it, and a young man from a think tank in California, with no experience running a phone company, was telling them they had it all wrong. They politely showed him the door.

By 1965, after years of trying, Paul Baran gave up. He couldn’t get anyone to build his network. He set the eleven volumes aside and went on to other projects. He was thirty-nine years old. He probably thought his big idea was dead.

It wasn’t. It was sitting in an office across the Atlantic Ocean, being thought up all over again, by a completely different person, for completely different reasons.

Meanwhile, in England

His name was Donald Davies. He worked at a place called the National Physical Laboratory — a British government research center — outside London.

Davies, in 1965, was not worried about nuclear war at all. He was worried about something much smaller. He was worried that interactive computing — the thing J.C.R. Licklider had been promoting, the idea that you should be able to talk to a computer in real time instead of handing it cards — was going to be too expensive to ever become normal. If every person who wanted to use a computer had to keep their own phone line open to it for hours at a time, the phone bills alone would be more than most people could afford.

There had to be a more efficient way. And Davies, sitting in his office in England, working on this purely civilian problem, invented essentially the same thing Paul Baran had invented five years earlier in California.

He chopped messages into pieces. He let the pieces find their own way through a network. He let them be reassembled at the other end.

But Davies did one thing Baran hadn’t done. He picked a better name.

The little pieces of a message, Davies decided, should be called packets. As in: a small package, a parcel, a bundle. Each piece was its own little package, with its own address label, traveling on its own.

The name was perfect. It stuck. And from that moment on, this whole technique — chopping messages up and sending the pieces independently — has been called packet switching.

Davies, unlike Baran, was actually able to build a small version of his network. The British government gave him a budget. By 1967 he had a tiny working prototype running inside his laboratory in England. It worked.

In October of that year, Davies’s team gave a public talk about their work at a conference in the mountains of Tennessee. An American researcher sat in the audience. After the talk, one of Davies’s colleagues mentioned, almost as an aside, that an engineer at RAND in California had worked out the same ideas a few years earlier — had anyone in the room heard of Paul Baran?

Almost nobody in the room had. Davies had not heard of Baran. Baran had not heard of Davies. They had invented the same thing, on opposite sides of the world, for completely different reasons, almost a decade apart.

All 24 chapters of How the World Got Wired · Library · All courses