On 2011-08-22 5:19 PM, Tony Arcieri wrote: > This is a valid claim that any world-scale peer-to-peer filesystem must > overcome. I'd hope that such a system would require you store content you > *are* interested in much in the way BitTorrent does already, i.e. to access > content you must download it and make it available to the network for some > period of time. Perhaps it's a fair concession that people who act solely as > consumers only store and redistribute content that they themselves are > interested in. > > However, I think this same system should allow the ability to store > arbitrary content (in a similar manner to MNet/Tahoe/Freenet/GNUnet).
If it is a world wide system, will have a very large number of providers, and a very large number of files. This makes accounting potentially hard. In bittorrent, you only track reciprocity within a single torrent, so each client can track reciprocity with each peer that it deals with, bartering blocks for blocks, since pretty much all clients are interested in blocks, and one block has much the same value as another. The system I described earlier only works if the number of providers is manageable, if each client can maintain relationships with most of the other clients, for the total number of relationships is N*N, and each client tracks nearly N relationships. Somehow, have to reduce that to each client tracking log N, plus relationships with those interested in the files it hosts, and those interested in the files it seeks, or perhaps log (N*M) where N is the number of clients and M the number of files. Not sure how to do that. _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
