Ian Clarke wrote: >> Yes. (The red number is route success rate, and blue the mean number >> of steps.) > > > Why doesn't this contradict Kleinberg's result?
a) It is an asymptotic result. It doesn't say that every distribution that looks correct will always lead to good results. b) Dependencies. Having a harmonic distribution of link lengths is not the same as having drawn the links independently from a harmonic distribution. To see this clearly, look over at the right hand side of the picture in my last mail. See all the nodes that form a chains - it is extremely unlikely that you would have such chains of nodes if the links had been drawn independently. // oskar
