Cox et al. [18] study providing DNS service over a peer-to-peer network as an alternative to traditional DNS. They cache index entries, which are DNS mappings, along search query paths. Similarly, the TerraDir Distributed Directory caching scheme [3] has nodes along the search query path cache pointers to other nodes previously traversed by the query. In each of these examples, cached index entries have expiration times and are not refreshed or maintained until a miss or failure occurs.
Path caching of content in peer-to-peer systems has received more attention. Freenet [19], CFS [20], PAST [14], and Lv et al. [8] each perform path caching, or caching of content along the search path of a query. These studies do not focus on cache maintenance, but rather depend on expiration or cache size constraints to implicitly prevent the use of stale content.
CUP trees are similar to application-level multicast trees, particularly those built on peer-to-peer networks. These include Scribe [15] and Bayeaux [21]. Scribe is a publish-subscribe infrastructure built on top of Pastry [6] where subscribers interested in a topic join its corresponding multicast group. Scribe creates a multicast tree rooted at the rendez-vous point of each multicast group. Publishers send a message to the rendez-vous point which then transmits the message to the entire group by sending it down the multicast tree. The multicast tree is formed by joining the Pastry routes from each subscriber node to the rendez-vous point. Scribe could benefit from our CUP ideas to provide update propagation for cache maintenance in Pastry.