That is the ideal situation, but I think we want the network to be as tolerant as possible to volatility and instability of nodes. Certainly effective discovery mechanisms are not a bad thing.
As I have described before, I don't believe that discovery would work at all if we pulled the current infom script. And the method that is used for discovery, the fact that the references point "downwards" rather the "upwards" along the routing for inserts is a hypothetical problem spot in that it does allow people some control of having requests for certain key value routed towards them. I would prefer if we put the references on Inserts pointing "upwards" the same way we do on Requests - simply have a randomly reset DataSource sent down with the InsertReply message. I'm not sure about Ian's greedy collection either, I think there could be good reasons for nodes to want to communicate without being discovered, and we shouldn't force it on them. Also, if the key values for the references are taken from the requests as well, then you get another situation where people might hypothetically be able to steer the routing. I think a good way to do discovery would be to be able to report yourself as a new node to another node, and it would randomly route some message it gets to you - or possibly that the other node simply resets the datasource of the one or some of the messages it passes to the new nodes address. On Sat, 03 Jun 2000, hal at finney.org wrote: > Ian writes: > > One idea might be to make nodes aquire the addresses of other nodes > > greedily - ie. if they see the address of another node (perhaps one which > > has sent a request) they add it to their datastore with a configurable > > probability. Additionally, we could support an additional field in > > messages in which a node could place the address of a randomly selected > > node from its datastore (or even more than one), and the node receiving > > that message would have the option of adding that node to its datastore. > > We could make the nodes test the other nodes with a handshake before > > adding them (perhaps with 70% probability to prevent abuse). > > I'm not sure node discovery is that big a problem in the long run. > Eventually most nodes will be up most of the time, the network will be > mature, and nodes will have the addresses of many other nodes. There will > always be a small fraction of nodes that are new, but I don't see any > reason to accelerate their acquisitio of other node addresses. > > Unless you think for some reason that in a mature Freenet, a large > fraction of the nodes at any given time will be new, node discovery > should pretty much take care of itself. > > Hal > > _______________________________________________ > Freenet-dev mailing list > Freenet-dev at lists.sourceforge.net > http://lists.sourceforge.net/mailman/listinfo/freenet-dev -- Oskar Sandberg md98-osa at nada.kth.se _______________________________________________ Freenet-dev mailing list Freenet-dev at lists.sourceforge.net http://lists.sourceforge.net/mailman/listinfo/freenet-dev
