gotta ask since i don't grasp enough of the freenet protocol yet - is there not some way that having collected packets for a while a node can _investigate_ the local network 'weather' ? and then make more informed decisions ? punknet gives nodes ways to tracelessly verify some of the routing information they receive, i imagine this is also possible in freenet, but i simply don't know yet.
jon > > > > Brandon wrote... > > > > > > > I had a mad idea (feel free to point and laugh...) > > > > I think this is a great idea, or at least some variant thereof. > > > > I'm not so sure about fuse=0 causing Tit-for-Tat. It won't be quite so > > clear cut as that. You can't just say that if you send a message to a > > request to a node and the request fails that the node is misbehaving. The > > request might be for something not in the network or not in the specified > > HTL. It might have been dropped out of the network because there was a > > flood of information, or it was unpopular, or because the node was > > leeching, or because the node was behaving well, but the nodes it was > > connected to are leeching. You're lumping a lot of possibilities into a > > single transaction. > > > > It's important to remember that you can never deal with a node by > > itself. When you're dealing with a node, you're really dealing with the > > part of the network reachable from that node in the given HTL. > > I realise this - yes you don't want to refuse to reply on the basis of a > single message and maybe a more sophisticated metric than the simple average > would > be better - I don't know too much about this. I did think though that you > *could* hold a single node responsible for the behaviour of all nodes > downstream of it in Freenet space. It's a bit like the teacher picking one > poor kid in control of the class while the teacher is absent and threatening > him alone with detention if there is a riot when the teacher gets back > (he'll make sure they all stay quiet for his own sake :-) I'd hope that the > effect might filter along. don't quite know how you'd do this tho > > > > > But we certainly could use some way to rank the usefulness of different > > nodes/parts of the network and communicate more with those that are more > > useful. The only way we're going to get Freenet to be ubiquitous is to > > add some incentive (i.e. better service from the network) for running a > > useful node. > > > > > > _______________________________________________ > > Freenet-dev mailing list > > Freenet-dev at lists.sourceforge.net > > http://lists.sourceforge.net/mailman/listinfo/freenet-dev > > > > > _______________________________________________ > Freenet-dev mailing list > Freenet-dev at lists.sourceforge.net > http://lists.sourceforge.net/mailman/listinfo/freenet-dev > _______________________________________________ Freenet-dev mailing list Freenet-dev at lists.sourceforge.net http://lists.sourceforge.net/mailman/listinfo/freenet-dev
