Despite an avalanche of eager adopters, agreement on what P2P means is sorely lacking. Some P2P applications predate the Internet itself, such as the Domain Name System. This creates a quandary: If the backbone of the Internet, DNS, is a P2P network, does the whole Internet also qualify?

Defining P2P
P2P arouses revolutionary fervor among its proponents, although the most prominent victories pale by the standards of past revolutions: women's suffrage versus… free music? Nevertheless, some politically motivated P2P projects have merits beyond technical wizardry. The Publius and Freenet projects, which provide anonymous and uncensored publishing via P2P networks, restore a measure of freedom to an increasingly commercial Internet. Other P2P ventures operate on the less- revolutionary principle that even spare processing power is a commodity; but these examples only prove that a single buzzword can accommodate polar philosophies. Let's examine the technical definitions instead:

"The P in P2P is people."
--From Dave Winer's Sep. 13, 2000 column

Perhaps something less abstract? The Peer-to-Peer Working Group's definition of P2P will serve:

"Peer-to-peer computing is sharing of computer resources and services by direct exchange."
--From http://www.peer-to-peerwg.org/index.html

That's better, but still vague. To complicate matters, advocates often ignore malicious implementations when discussing P2P. For instance, Distributed Denial of Service (DDOS) applications use multiple computers to pepper a target with spurious network traffic. Similarly, email viruses consume resources on many machines as they replicate. Both skirt the edges of this broad category, yet are rarely mentioned in the same breath as P2P.

Dissecting the Definition
Three terms in the Working Group's definition bear explanation: sharing of resources, peers and direct exchange.

P2P's unique contribution to computing is not the sharing, but the distribution of resources. No longer do servers house all the resources and dole them out to clients; a P2P network makes use of dispersed computing resources. Though most P2P applications share storage or processing power, any computing resource can theoretically be shared-monitors, printers, and so on. P2P levels the client-server relationship, creating peers. Legally, peers are on equal footing, and the same holds true in computer terms. In a "pure" P2P network, each node is client and server.

The last term is more difficult to grasp. Where does "direct exchange" occur in a P2P application like Napster, in which a central server handles all the searching? Isn't this simply a glorified client-server relationship?

A few case studies will clarify the Working Group's definition.

Distributed Computation: SETI@Home -- The Search for Extraterrestrial Intelligence
The SETI@Home program, distributed as a screen saver, sifts minute quantities of astronomical data for evidence of extraterrestrial life. Results are sent back to the SETI project at the University of California at Berkeley. With approximately three million charitable users, the SETI@Home network is the most powerful supercomputer on the planet, but boasts no interaction between peers at all. The only beneficiary of this orgy of screen saving is the central SETI@Home database.

Centralized P2P: Napster
The music sharing service Napster is the flagship P2P application, with an installed base of over forty million. Napster's index of music files is centralized. The content, however, is decentralized, and resides not on company-owned servers but on millions of peer computers. Like a phone book, Napster provides addresses, but won't deliver. Napster's own servers take no part in file transfers, which occur entirely between peers. Unfortunately, this distinction provides no legal insulation in today's climate, and the company was recently served with an injunction forbidding the trading of copyrighted material via its service. In general, a centralized P2P model aids searching, but replicates content excessively.

Decentralized P2P: Freenet
Freenet is a completely decentralized P2P file sharing service with an emphasis on anonymity. There is no centralized server. Each computer in the Freenet runs a piece of software that turns it into a node equivalent to all the others, and uploaded content is distributed in encrypted form between all the nodes. Currently, searching the Freenet is impossible. Instead, in order to retrieve a given document, the user must obtain the appropriate key.

A Jury of Peers
The P2P acronym gets applied very generously. Some of the foregoing examples boast no peer-to-peer interaction at all, but all reside on O'Reilly's P2P Directory.

P2P networks have certain technical characteristics that most definitions fail to capture. All P2P networks utilize widely distributed computing resources. SETI@Home shows that certain problems (those that can be broken down into small, distributable chunks) are amenable to P2P solutions. Freenet shows that certain tasks (searching and indexing) are difficult in proportion to the degree of decentralization. Ironically, Napster became the flagship P2P application precisely because of its impure pedigree: a centralized, efficient search engine. These criteria hardly seem like the founding principles of a computing revolution, though. What else is going on here?

The P2P media frenzy began when Napster became synonymous with larceny in the eyes of intellectual property lawyers. This is an old story, oft repeated. Each time a recording medium gathers significant users, content producers attempt to quell the market and fail. Battles over the legality of videocassette and CD recorders have long been ceded in the public's favor, and the skirmish over the MP3 format seems bound to follow, as MP3 walkmans and MP3-enabled CD players are increasingly commonplace. Content producers are redirecting their efforts towards:

• Developing viable content encryption schemes to prevent piracy.
• Attacking ISPs that permit piracy.

Encryption schemes encoded into hardware are almost predestined for circumvention; witness the rapid dissolution of DVD encryption. The latter tactic, however, is something new. Traditionally, the stance of American regulatory organizations has been that "carriers," such as phone companies, cannot be held responsible for material transmitted over their infrastructure. Although Internet Service Providers (ISPs) receive similar protection against liability under the Communications Decency Act (CDA), the Digital Millennium Copyright Act (DCMA) compels an ISP to remove infringing content once notified of the copyright violation. Other nations are not even this permissive, as indicated by a recent German court case in which AOL was held liable for copyright infringement conducted by its users. Additionally, bandwidth is asymmetric: That is, users download a great deal more than they upload. P2P networks often reverse this relationship by turning previously taciturn clients into chatty peers, thereby placing additional load on ISPs.

Thus, the Achilles' heel of any P2P network is the network itself. Unlike the bootleg VHS market, the once-booming Asian Video CD market or the traffic in gold Playstation discs, Internet piracy relies on the Internet itself. Back in Internet prehistory, billboard service (BBS) operators used upload/download ratios and other mechanisms to reign in rampant users. Similar measures can be concocted for the Internet, the most obvious being bandwidth metering (which may actually be impossible with cable modems… perhaps their only advantage). Sufficient pressure on ISPs can also halt the most egregious instances of Internet piracy. The DMCA gives content providers a surefire legal strategy for curbing such activity when it's discovered within America, and other nations are similarly inclined even without an equivalent to the DMCA.

Choose Your Peers Wisely
I do not condone or defend piracy. As the author of this article, I'm keenly aware how easily digital media can be appropriated. Nevertheless, Napster's case is an unfortunate legal precedent for benevolent P2P applications like Freenet, Publius, and the many distributed computation projects. Since ignorance of technical detail often results in knee-jerk ISP usage policies and blanket legislation like the DMCA, the aggregate "P2P" banner may inadvertently stifle the development and application of honest peer-to-peer technology.

Meanwhile, vendors are jostling to benefit from the same P2P technologies that ISPs are encouraged to discourage in the public. In some cases, the public's own computers are the profit center. Companies like Porivo allow users to donate "wasted" processing power in return for a free entry in a sweepstakes or a cash pittance. The spiel is generally couched in conservationist terms. Nobody likes waste, especially when it's unprofitable.

Luckily, resilient P2P applications use technical wizardry to circumvent restrictions. A flawed first-generation alternative to Napster named Gnutella communicates via HTTP, the lingua franca of the web itself. Unless the Internet itself is shut down, the stronger second generation of P2P applications may prove insurmountable. Legal acrobatics will no doubt ensue, but I'm optimistic about the prospects for public peer-to-peer networks, given the unruly nature of the Internet.

And frankly, I already give enough of my processor time to flashing browser ads. The rest is going to charity .

What's your take on P2P? Post your comments below.