The fashion today is to say, "It's the market, left on its own, free of government." But a number of histories of the Net - yes, histories already - show that the answer is more complex.

A recent favorite of mine is an account by John Naughton, in a book titled A Brief History of the Future (currently available only in England, or via the Web at www.briefhistory.com). In telling the story of the Internet's development, Naughton describes where innovation flourished, and where it was held back. The flourishing was sometimes inspired by government, though only when unconstrained by bureaucracy. It was sometimes inspired by business - but interestingly, and importantly, not always. Business, too, gets stuck. When, and why, is an important lesson.

Here's one example: The initial concept for the Internet was born long before it came into being. Its author, Paul Baran of Rand Corp., fixed on what would come to be called packet-switching technology - in 1964. Rand liked it and presented it to the Department of Defense. He met with resistance.

The source of the resistance soon became clear: ATT (T). In its view, packet-switching just couldn't fly. But as, of course, it eventually did fly (though not until many years after Rand tabled Baran's proposal), why was AT&T so wrong?

One clue is an exchange between Baran and AT&T's Jack Osterman. As Naughton recounts, after a long discussion, Osterman in some frustration said to Baran, "First, it can't possibly work, and if it did, damned if we are going to allow the creation of a competitor to ourselves."

"Allow."

The key to the Internet's extraordinary innovation is that it doesn't allow a term like "allow." It's architected to disallow it. The Internet is built on a principle called "end-to-end." First described by Net theorists Jerome Saltzer, David Reed and David Clark, end-to-end means the network does not choose how the network will be used. Control, or intelligence, is placed at the "end." The network is to be kept simple, incapable of discrimination. What is allowed in the Internet is what users demand. The innovations that are permitted are those that users find useful. No central or strategic actor gets to decide how the network will evolve. The network is constituted to disable that sort of control.

This end-to-end design frees innovation from the past. It's an architecture that makes it hard for a legacy business to control how the market will evolve. You could call it distributed creativity, but that would make it sound as if the network was producing the creativity. It's the other way around. End-to-end makes it possible to tap into the creativity that is already distributed everywhere. Because innovation doesn't depend on who you work for or what rank you have, innovation comes from the best ideas.

Every great story of innovation in this period has at its core a freedom from the past. It is Microsoft (MSFT)'s great strength that its chairman mercilessly shifts direction when competitive forces demand it, sacrificing legacy departments left and right. In author Jim Carlton (CCTVY)'s view, its absence is illustrated in the story of Apple (AOIXQ)'s decline (before Steve Jobs' recent return): Managed by consensus, Apple was not able to sacrifice the outdated legacies within its consensus.

And this freedom is the constitution of every great system of innovation. This is the architecture of the open-source, or free, software movement. By carrying its code with it and by leaving it open for others to modify, no one actor can maintain control over open-source evolution or make strategic decisions on how the system will evolve. The systems will evolve as innovators evolve them, not as corporate leaders "allow." Other innovators see this. They develop for this platform because they know it can't turn against them.

We should learn from this past, embrace its lessons as foundational and preserve them in the future, because there's no guarantee that end-to-end will remain a principle of the Internet. There are plenty of people who would like to return to a design in which the network controls how the network is used. End-to-end is a choice made by the Net's founders, and one which we continually must defend.

Lawrence Lessig (lessig@pobox.com) is Berkman Professor of Law at Harvard Law School.