On Fri, Jun 08, 2001 at 03:46:22PM +0200, Oskar Sandberg wrote: > On Thu, Jun 07, 2001 at 03:38:33PM -0400, Tavin Cole wrote: > <> > > > Once it is passed upstream the node will no longer be queried (at least > > > from that > > > direction) right? So if you delete a datum after so many requests, it is > > > trivial for a > > > specific node to force that data out of the data store. It simply > > > progresses from one > > > node to the next, until it is gone completely (i.e. the malicious node is > > > assumed to have > > > this upstream copy, when in fact, it has been performing the requests to > > > force the data > > > out of store). > > > > > > Perhaps I am misunderstanding, but this would appear to be a major > > > vulnerability. > > > > No, there is no reason why the first node wouldn't continue servicing > > requests > > for the key. It wouldn't drop out one-by-one up a chain, instead each node > > in the > > chain would periodically drop and regain it at a different frequency. > > No, he has a point. When Node A reaches your threshhold for K and allows > the next search through, then it will most probably not reset the > DataSource on the reply. So it does cause the node to get queried less for > K.
Well, depends on the semantics of DataSource reset of course, such as whether we decide to only cache on DataSource reset. But with the current model, it's true that it could reduce the frequency of requests for that key. Hard to quantify how much. But wouldn't this be good, since the requests would be directed upstream? Suppose the chance of killing the file were dependent on it's relationship to the node's keyspace focus, either as self-analyzed or based on HTL values.. -- # tavin cole # # "Technology is a way of organizing the universe so that # man doesn't have to experience it." # # - Max Frisch _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
