On Mar 5, 2008, at 11:08 AM, jiangxingfeng 36340 wrote: > >> >> Yes, some of it can be generic, but some of it isn't, and since the >> routing needs to be done at the DHT layer anyway--and some DHTs >> certainly will want to propagate non-generic information--it's not >> clear to me that having the information be split between layers >> adds a >> lot of value. >> > One option is to let the routing decision at the peer protocol > level to override the one at the DHT level. Because peer could use > the rich information collected by the P2PSIP peer protocol and > could do the more optimal decision. This is one way in my mind how > the generic information is to be used.
I think there are some opportunities for this type of split between the generic peer protocol and the DHT (one example being the use of random replicas/random replication factor which can be an application decision vs the successor replicas that a chord-based DHT maintains based on its routing setup). However, since only the DHT knows the overlay's topology and routing algorithms, I don't see how any other layer can make such a decision. Even the use of alternate paths to avoid routing attacks by compromised peers needs to rely on the DHT layer to identify an alternate starting peer that will still reach the destination ID. Bruce _______________________________________________ P2PSIP mailing list [email protected] https://www.ietf.org/mailman/listinfo/p2psip
