Working at Xerox's Palo Alto Research Center (PARC) on what they term internet ecologies, Eytan Adar and Bernardo Huberman conducted the study earlier this month to discover more about user behavior patterns on a Net music trading community.
The scientists marked off a 24-hour period spanning one August Saturday to Sunday afternoon, looking for high-traffic periods of usage on the Gnutella network. Using a modified Java-based Gnutella client, the researchers observed a final count of 31,395 hosts who shared over 3 million files and found that only a relatively small number of users share desired files. The majority of users examined in the study tended to take files without responding to queries by other people on the network looking for particular songs or artists. The scientists dubbed the music takers "free riders."
During the 24-hour period, 70 percent of Gnutella users failed to share a single file, according to Adar and Huberman. Additionally, 90 percent of the users did not respond to queries for particular songs or titles by certain artists. The pair also found that 'free riding' figures distributed equally between Net domains, showing that such behaviour spans different age groups, locations and levels of connectivity, with no single group contributing more significantly than any other.
"There was a high number of people who weren't sharing anything," Adar said. "People send out queries, and the other users don't write back at all."
The study claimed that the top 1 percent or 314 hosts contribute about 40 percent of the total number of files shared. The number of files shared skyrockets to 98 percent if the top 20 per cent of users (6,250 hosts) are considered in the data.
In some cases, music-swapping sites that use a model similar to that of Napster have centralised servers where users go in search of songs. The effectiveness of this strategy - to link people all over the world into focused Net locations - has drawn the interest and concern of music labels claiming that the peer-to-peer technology violates copyright laws on a large scale. Napster is currently battling the Recording Industry Association of America in court, with the RIAA seeking to close down the music-swapping site.
Unlike Napster, however, the Gnutella network does not maintain central servers. Instead, a user downloads or develops an application that complies with the Gnutella protocol. The application functions either as a client or a server in addition to working as a high-level network, routing information across servers and clients. To join the system, users connect to one of numerous known hosts that forward the IP (Internet Protocol) and port address information to other people on the Gnutella network.
The method of peer-to-peer file sharing conducted by Gnutella helps preserve users' anonymity and has traditionally drawn less legal attention from the music industry. Some industry observers argue one of Napster's most powerful features is the ability of the company's technology to create a huge network of users. Gnutella's less-centralised system lacks an obvious focal point and can make finding song titles slightly more difficult.
Adar said that the data in the study may provide some disturbing news for Gnutella users who rely on anonymity and the lack of centralised servers to avoid legal troubles. With only a few users providing the bulk of available files, the core file-sharing traders begin to look like a centralised repository. The system not only suffers in functionality by placing heavy traffic on a few locations but also the people making the most contributions could suffer the greatest legal consequences in the long run, the study said.
"As more people free ride, the system looks more and more centralised," Adar said. "Gnutella is starting to look more and more like Napster."