On Monday 26 November 2001 12:48, you wrote: > To speed up routing, would it be possible to introduce newly discovered > nodes at a slightly lower CP value? (Like 0.9 or something.) That way, > the node will be contacted *eventually*, but after nodes that are known > to work. The routing should already be biased in favor of nodes that have been responding. > > In addition, the CP falloff algorithm doesn't seem to be that effective. > For instance, my node made several hundred connection attempts to one > node and succeeded on every one except for four -- and it had a lower CP > value than a node that it had never successfully contacted. Is this what > is supposed to happen? The absolute value of the CP doesn't mean much, because some nodes are much more favored by routing and so get tried more often. > > I understand that nodes can go offline without warning and the current CP > system takes that into account, but should working, fast, reliable (98% > +) nodes be contacted after seventeen modem users in the middle of no > where? Keep in mind that the whole CP scheme is a kind of desperate compromise to get requests somewhere that will answer without damaging routing too much. Even though another node is easy to contact, if the that node is very far away from where the node "should" be routed according to the routing keyspace sending the request there won't help you much (and will keep the network from specializing).
That being said, I have just checked in changes which I hope will improve the way connection choices are made. --gj > > > > _______________________________________________ > Devl mailing list > Devl at freenetproject.org > http://lists.freenetproject.org/mailman/listinfo/devl -- Freesites (0.3) freenet:MSK at SSK@enI8YFo3gj8UVh-Au0HpKMftf6QQAgE/homepage// (0.4) freenet:SSK at npfV5XQijFkF6sXZvuO0o~kG4wEPAgM/homepage// _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
