On Mon, Jun 30, 2003 at 04:11:50PM +0100, Toad wrote: > Why can't we just use the game theory equasion that I suggested? There > is no way we can detect when a node is maliciously returning DNF, and if > it did it would almost certainly only do it on a very few keys.
The danger is when a node returns a DNF for data which is in the network, the idea is to punish nodes for doing that without punishing them when they DNF for data that isn't in the network (since this isn't their fault). That is the point of the previous algorithm, your algorithm assumes that all DNFs are for data that *is* in the network (and are therefore unwarrented). > Expected overall time is essentially the expected time for that key for > the whole node, whether it succeeds or fails, and we add it on because > if we fail, the time for the request will be the failure time plus the > time we get from the next route. But if the data wasn't in the network in the first place it isn't fair to punish the reference for the cost of re-retrieval. The current equasion accounts for that. > This seems a rigorous and non-arbitrary > system with absolutely no voodoo/alchemy involved. As is mine - but mine is better ;-) Ian. -- Ian Clarke [EMAIL PROTECTED] Coordinator, The Freenet Project http://freenetproject.org/ Founder, Locutus http://locut.us/ Personal Homepage http://locut.us/ian/
pgp00000.pgp
Description: PGP signature
