Matthew Toseland wrote: > We have historically created models based on Kleinberg's work - create a > big circle, then add cross-links with a 1/d distribution. > > However we have also simulated with Orkut-derived real world data. > > It would be sensible to at least start with the same data we have used > to simulate routing.
Yup, Kleinberg's model will be a useful sanity check. Could you send me a copy of the Orkut data? Otherwise I've got a copy of the LiveJournal graph that I could extract samples from (after removing unidirectional links that represent publisher/reader relationships rather than friendships). Cheers, Michael
