Just because it is not easily quantizable DOES NOT mean there is no error in it.
On Mon, Dec 01, 2003 at 07:40:58AM -0500, Zlatin Balevsky wrote: > > >We need to think about why NGR isn't benefiting from a similar > >self-reenforcing effect. I don't think it is as simple as "its > >estimates aren't good enough", any more than it is valid to say that > >classic routing's routing decisions are perfect. > > If we view classical routing as ngr with a single estimator, which is > the distance between the requested key and the reference to the node, > then this estimator is always 100% correct because we know both the > requested key and the reference to the node with 100% accuracy. > > Ngr is working with probabilities; in case where the estimates for the > probability are severely off-track, which is possible when two or three > estimated variables are somewhat off their actual values, the result of > following the formula is as good as random. (and if my suspicion that > the estimators give us highly erroneus values of <50% accuracy is true, > then the whole thing is as random as random gets) > > In both cases we route to the node which is the best candidate according > to certain characteristics, but in the case of classic routing this > decision is based on 100% accurate estimate. Even if the different > nodes have different references for each other, the one that is the > proper one will eventually spread to other nodes through reference > passing, etc. In ngr, unless the error in each estimate is miniscule > then the final result of many nodes using a formula with high error > margin result in nearly random routing and will prevent the > self-fulfilling proficy from happening. > _______________________________________________ > Devl mailing list > [EMAIL PROTECTED] > http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so.
signature.asc
Description: Digital signature
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl