> > Hm. The longer I think about it, the more ways I come up with for two > > different NGrouting schemes to develop divergent views of which peers > > handle which keyspace. That would be bad. > > I am not so sure, with NGrouting there isn't really such a clearly > defined idea of "peer Y is the right place to route a request for key > X", but rather "peer Y will take time Z to get key X", where we try to > minimize Z. Provided that the algorithm's ability to estimate Z is good > enough - it should work. > > > Perhaps the convergence property should be written down in the NGrouting > > doc -- it is very important for the behavior of a large-scale network. > > I am still not sure that what you refer to as "convergence" isn't > implicit in the specification already. If nodes are making radically > different estimates to each-other about routing times, then those > estimates can't be very good. If nodes are making good estimates, then > those estimates will be similar from node to node.
Hm. Suppose that Alice and Antoine both specialize in the 0.0001 to 0.0002 sliver of keyspace, and all of Alice's immediate neighbors perceive her as the best node for that sliver and all of Antoine's immediate neighbors perceive him as the best for that sliver? There may be something about NGrouting that automatically fixes this situation, but I don't yet see it. Regards, Zooko http://zooko.com/ _______________________________________________ devl mailing list [EMAIL PROTECTED] http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
