On October 27, 2003 02:36 am, Tracy R Reed wrote: > I'm glad that major routing bug was squashed. Well done! But I think there > is yet one more big thinko somewhere...
> It seems that the network is not learning who specializes in what areas of > the keyspace from successful and failed requests. For a number of weeks > now I have been running my node with a specially tuned datastore. I wrote > a script to delete keys from the ds that are not in my randomly chosen > area of specialization thus forcing a strong specialization. If the > network is working properly other nodes will notice that I am unusually > successful in fulfilling requests for a certain part of the keyspace and > will route similar requests to me reinforcing this specialization. This is > not happening. Here is a histogram of my datastore: Think the routing bug could easily have caused this. Think about it. We we using the worst path. So data would follow this. Once a node got data that people we requesting it would no longer be the worst and hence we would not route to it. This would really tend to damp any specialization. Over the next two or three weeks (it will take time) we should see nodes start to specialize again. Note that just because you have your DS with lots of 'A's in it does not mean your node is specialized on 'A'. What it should do it influence it to specialize near 'A' (now that routing is 'fixed'). Specialization is not only what is in your DS but, also what keys your node can find quickly. If you have 5% of all the 'A' keys but you node knows little about routing to other nodes that have 'A' the rest of the network will not see you as specialized. Ed _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl