Why not just send a ping to every node and count the number of hops to them?
I guess this wont happen untill 0.9 or something, but if it were to happen I think you would have to use two different systems. One for CP and one for Performance. So each node has a CP and a Performance number. So if the node has reliable routing you would favor the nodes with good Performance, if the node experiences many failed connects it would favor nodes with good CP.. The Performance number should be based on Latency, average throughput and maximum throughput. The thoughput and latency numbers should be scaled in % against the best throughput / latency found. Just my 200 Lire. On Mon, 26 Nov 2001, Will Glynn wrote: > One other thing that could speed up routing is a CP increase routine that > takes the destination machine's IP into account. For instance, a node > talking to another node at my ISP will be faster than my node talking to a > node halfway around the world. A bunch of Freenet nodes at a college would > talk to each other a lot faster than talking to my node. Et cetera. > > This increase in efficiency could be gained by changing the CP reward system > to take into account distance in the IP address space. Basically, compare > the destination IP with the local IP (the one that's used to talk to this > node), count how many of the first bits match, and adjust your multiplier > accordingly. > > This would make a useful feature if the CP scheme is going to be re-worked > anyway. > > > _______________________________________________ > Devl mailing list > Devl at freenetproject.org > http://lists.freenetproject.org/mailman/listinfo/devl > -- | Are ?rseth | elrond at idi.ntnu.no | http://ctemp.dhs.org/ | | tlf: 73886876 | mob: 410 00548 _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
