Datensatz

Zusammenfassung

The solution that we have developed and advocate in this paper is based on the postulation that we cited above as a motivation for the general direction of P2P search engines: \emph{give the Web back to the people}! We simply observe the queries and behavior of the thousands or millions of users that we expect to be active in a P2P network. More specifically, we monitor queries on the peers where they are initially posed and post them to the distributed directory, where they are aggregated so that interesting correlations can be inferred. Obviously, the details of this method need to be carefully worked out to ensure scalability; we will explain this in the paper. The key point to emphasize here is that a querying peer whose query keywords exhibit correlations has a good chance to efficiently find a directory entry that helps to identify the best peers with files that match multiple terms of the query. A potential caveat to our approach could be that user monitoring and query logging is a breach with user privacy. But it is exactly the key strength of the P2P approach that, unlike with a centralized Web search engine that logs queries and click-stream information, every peer is in a perfect position to define its own policies, would reveal critical data at its discretion, and has full control over the local software to enforce its specified policies. In other words, community-wide collaboration is desired and encouraged but not forced, and peers remain perfectly autonomous.