"Scientists used to need a big government grant for massive research projects, but now you'll be able do it through the Net," O'Reilly says. "Distributed computation is arriving at a time when science is realizing that it needs it."

Take, for example, Popular Power's first client, Derek Smith. He holds a Ph.D. in computer science, but his passion is virology, which simply means that he loves to study the way viruses interact with the human machine. HIV was his first target; but his present work -- primarily for the World Health Organization's National Influenza Laboratory at Erasmus University in Rotterdam, and the Influenza Branch of the Centers for Disease Control (CDC) -- focuses on how strains of the flu interact with immune systems and vaccines.

Influenza's coughs, sneezes and aches may seem like an insignificant concern. But the virus kills at least 20,000 people each year nationwide, according to the CDC, and while flu shots are the best way to keep it at bay, they don't always work. "The flu is a moving target," Smith explains. Unlike, say, the measles, it mutates into thousands of strains each year, which means that scientists must guess which strains will spread and which vaccines will be effective. So far, clinical trials show the efficacy of vaccines to be about 80 percent.

But everybody would like the vaccines to work 100 percent of the time, and Smith's research is aimed at helping that happen. For the past eight years computers have dominated his research. And, as with other applications of distributed computing -- graphics and financial modeling, for instance -- "Processing power is the key," he says.

It's essentially a matter of the more, the healthier. Before the spread of P2P, Smith would typically borrow about 15 PCs from colleagues and friends, and use the computers to model thousands of interactions between mathematical versions of viruses, vaccines and immune systems. Comparing the results of the number crunching revealed which vaccines worked best. And since it took more than three years just to code the software and write the algorithms, he felt that the graphs and charts he was sharing with fellow scientists were a sign of substantial progress.

Then, about nine months ago, he discovered Popular Power, and a whole new world opened up. "When Nelson told me what he was doing, I couldn't believe it," he says, looking back. "I told him I could use that kind of machine right away."

The excitement hasn't faded. Smith's research now runs in the background on thousands of computers, and he can't stop marveling at the power of "the giant brain." He speaks freely in terms of millions, billions -- even trillions -- of simulated interactions. He envisions using computers to build a flu vaccine that saves everyone who gets a flu shot. He even imagines a time when the research he's done on the flu will act as a model for similar studies of HIV.

"Computer modeling is a growing area of biology," he says. There's Parabon's effort to model how cancer cells react to drugs; there's the Biomedical Computing Group, formed in June by the National Institute of Health, and Process Tree, where peer-to-peer computers are being used to help a Swedish scientist build safer storage vessels for radioactive waste by simulating the radiation around an encapsulated radioactive source.

The options are endless, Smith says. "As we become smarter and smarter about how we use technology, it will become more and more important."

But demand solves only half the equation. What about supply? Will people actually find Popular Power, download the software and give back their processing power? Stavropoulos and some technologists favor skepticism. Napster has drawn 38 million users because it offered music for free. It appeared first and it caught the cultural zeitgeist, but fans mostly flocked to the service because it let them give and take something they were used to paying for.

There's no guarantee that Napster's music-inspired success will overcome the "what if it goes wrong factor," Stavropoulos says. Indeed, most peer-to-peer projects, including Freenet, still have fewer than a million users. Popular Power, for example, has fewer than 50,000 -- enough to keep Smith happy, but merely a fraction of what Napster numbered even before lawsuits made the company front-page news. And while Minar, O'Reilly and others point out that most Net-connected computers are clustered in offices, many companies won't be willing to give them up, says Dan Bricklin, CTO of Trellix, which makes Web authoring software. "It's a matter of trust," he says. Companies may not sell off their unused computer power because decentralized systems aren't as stable and secure and because they're subject to the whims of individual users.

"You have to look at the specifics of your case," he says. "The bottom line is that peer-to-peer is a technology strategy that is appropriate at some times and not in others."

Fine, says Ian Clarke of Freenet. So peer-to-peer needs to convince businesses and the world that it's safe, easy-to-use and effective. Napster did it, and so did SETI@home, lining up more than 2 million people who still let their idle computers search for aliens. Others success stories will follow.

"Maybe I'm being an optimist, but the primary problem with peer-to-peer so far is that the tools have been too difficult to use," says Clarke. Once developers make peer-to-peer a "one button, quick download, install this" application, millions will flock to it. Look at Freenet; when it started, users needed to understand IP protocols and maybe a little HTML to get it working right. But in the past few months, traffic has increased substantially because, Clarke believes, the development team has added a Windows installer and a Web component. Users can navigate through Freenet from within their browser.

"Peer-to-peer is about encouraging participation from users, so when it's very, very simple, people won't choose to be just mindless consumers," he says. "They'll become participants."

Plus, if the urge to create doesn't draw a crowd, there's always the magnetic appeal of cold hard cash. Process Tree plans to offer users $100 to $1,000 per year for their computer time; Popular Power hopes to cut deals directly with ISPs so that payment complications are minimized. Meanwhile, Mojo Nation's system includes digital payments that users dole out themselves. "Leechers" who simply try to take content -- music, movies, porn -- lose their stash, while sharing earns it back.

Ultimately, though, most peer-to-peer preachers trust that incentives won't be entirely necessary. Capitalism and Thomas Hobbes may hold that selfishness rules human nature, but Minar envisions a more complex version of both business and man. He plans to succeed and help the world; and he says it's only possible because people will follow the example. In fact, they already are.

"Eighty percent of the people who use our software give half their computer time to nonprofit causes," he says. "It's pretty amazing."

The Free Software Project
Read Andrew Leonard's book-in-progress on Linux and open source -- and post your comments.