>> Now if the routing algorithm can take into account previously requested >> keys, why can't it take into account performance/reliability metrics? This >> information is determined by the results of previous requests, so it is >> still a function of the previously requested keys. >Because unlike the key and the references to other data, the >performance/reliability metrics aren't a consistent network-wide set of >measurements. Each node can and will often have a different view of the >networks performance, so this yields to anarchistic and ineffective >routing.
Don't let Travis hear you saying that anarchy is ineffective. ;) I understand that all nodes must use the same measure of key closeness in order for document clustering to be useful. (The sets of documents that a node caches for its neighbours should overlap.) However, each node has a different view of the network's topology, and routing still works, so I'm still not convinced that each node having a different view of the network's performance will break routing. I'm not sure how to reconcile those two positions, but I'm working on it. Michael _______________________________________________ Freenet-dev mailing list Freenet-dev at lists.sourceforge.net http://lists.sourceforge.net/mailman/listinfo/freenet-dev
