Routing Around the Web
By Dale Dougherty
Participants of Tim O'Reilly's September Peer-to-Peer (P2P) Summit included leading developers from large companies like Microsoft, IBM, and Intel as well as from several innovative open-source projects including Freenet, Gnutella, and Jabber. One of the summit's topics was how much of the attention paid to P2P focuses on Napster, largely due to the client's notoriety. In O'Reilly's opinion, Napster is simply a part, albeit an important one, of a developing P2P story that's reshaping the Internet, yet again.
P2P has yet to be well defined. It can refer to file-sharing, distributed computation, instant messaging (IM), server-to-server communication, device-to-device connections, and more. Some P2P applications aren't necessarily purely P2P, either; that is, they may not be entirely decentralized, choosing to rely on only a few centralized services. While the P2P Summit participants didn't want to pin down a tight definition of P2P, they agreed that it isn't very new. Decentralization has been fundamental to Internet architecture since the very beginning, if you consider how packets are routed from one machine to another. What makes P2P seem new, they agreed, is that many things on the Internet are harder to do today than they were just a few years ago.
P2P technologies make sense because of the ways in which the Internet has changed. For example, many computers don't have their own Internet addresses anymore. Each computer used to be assigned its own IP number that mapped to a domain name. Today most ISPs and corporate LANs assign each computer a dynamic IP address from a pool of IP numbers.
