On Friday 19 September 2003 16:43, Some Guy wrote: > > --- Gordan <[EMAIL PROTECTED]> wrote > > > > > http://entropy.stop1984.com/en/p2p.html > > > The question is in the routing. > > > > But if they use compatible FNP, then the routing > > shouldn't matter. > > If you asume the protocol and the routing are the same > thing.
No. I am specifically looking at the protocol and the routing algorithm as two completely separable things. Protocol is how the nodes talk to each other. The routing is about how the node decides which nodes would be the best ones to ask for the content it is looking for. Provided that: 1) Routing is BASED on the hash of the data 2) Hash of the data is produced based on the same algorithm in both networks 3) Nodes are communicating using the same protocol, i.e. then can understand each other's requests I see no reason why networks could not be merged. But I may well be missing something... > My point of view was that nodes are free to > route how they want and still fufill the protocol, as > long as they find data posted. That's exactly what I was saying. > > I was thinking about it by looking at Freenet and > > Entropy as different > > implementations of the same general protocol. > > Compatible FNP is in that case > > the fundamental building block, and everythins else > > can be variable, e.g. the > > data store format, routing implementation, etc. The > > advantage is that Freenet > > and Entropy nodes would be treated equally, and > > there wouldn't be two > > different networks to bridge at all, just one bigger > > network, with what is > > effectively two different node implementations. > > Hmmm, so what you are proposing is that if someone > wants to implement the exact same routing freenet has, > then they can mix thier nodes with freenet's. No, I'm saying that if the communication is in the same format (FNP), the routing cab be based on ANY algorithm that somebody chooses to implement. > 3) I'm sure CrazyNet(or whatever) has no problem > mingling with freenet nodes if we implement the exact > same routing as CrazyNet. You seem to be assuming > that freenet's routing is perfect, and everyone else > will gladly jump ship and adopt it. No, I'm assuming that provided routing is based on the same feature of the data (i.e. the key hash), it doesn't matter what routing algorithm is used. I _think_ we are agreeing here but misunderstanding each other's comments. > What I was proposing was since the DHT is such a > universal problem, maybe different competing > (friendly) systems could share the same universe of > data, and we could learn from each other. That's what I'm saying, too. But I was against using gateways between the networks and in favour of merging networks that already use the same inter-node communication protocol. This should not have anything to do with the routing. > Say someone comes up with NNG routing for freenet, > which extends FNP. Why is FNP being lumped into the same item as routing? Are they really that inseparable? The routing algorithm (e.g. NGR) shouldn't have anything to do with the way the nodes talk to each other. It is only important as a decision tool with regard to WHICH nodes a node will be talking to about a particular request. > It'd be cool if we could run a > small subnetwork of NNG nodes to see how well they > work. This would give you a better idea of NNG's > peformance, than tring out one NNG node in a sea of > old nodes. Agreed. > Can we make a hetrogeneous network? We already have one, in this exact same way. We have about 2/3 of the network using the old routing algorithm (node version <=6192), and about 1/3 of the network using the NGR algorith (node version > 6192). > Is it worth it? It seems to be working... > > This would be roughly equivalent to some people > > using Lime Wire and other > > people using Bear Share, but both programs are just > > different Gnutella > > protocol nodes. The network, effectively, doesn't > > care about how the > > implementation works, as long as it obeys the > > protocol specification. > > Right and that works because the routing is done the > same. It sucks equally, no matter what you use. I am not sure about that. I don't see why routing is inherently inseparable from the communications protocol. Once a node takes a request to find the data, I don't see why it is relevant from the operation (not efficiency!) point of view what routing algorithm it is going to use to choose nodes to forward requests to. It should be able to pick the nodes according to it's routing algorithm, regardless of whether they are Freenet or Someothernet, as the protocol for sending a request is the same. Gordan _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
