Yes, it will lead to more hops per request, but the question is whether that is a bad thing. If those hops can be made between nodes with high capacity between one another, having more hops is actually a good thing because it leads to more caching and greater spread of addresses. And the important load to the network is mostly the bandwidth, so that isn't increasing either.
So if 15 hops with speed taken into account are faster then 5 hops without it, then I'll take 15 hops. The question is how much can this be done without breaking the algorithm, and answer I do not know (the problem with shooting blind). On Mon, 17 Apr 2000, hal at finney.org wrote: > I'm concerned that adding node latency/throughput to the routing algorithm > will make it less successful. The whole point of the algorithm is to > take me to the node that gives me the best chance of finding the data. > That's why we think that we can have a million nodes and still find > items even with HTL of 5 or so. > > Suppose I'm at a college with 100 nodes out of a million in the world, > and all of those 100 nodes have much better communication performance > with each other, compared with any node outside the college network. > That means that all 5 of my hops are likely to go to college nodes, if > we have an algorithm that gives preference in this way. If my data's > not at the college, I won't find it. > > You need an algorithm which will (A) find the data in the college > network if it's there (so as to avoid load on gateways), (B) find the > data anyway out in the world if it's not in the local network, and (C) > not require HTL to be much more than it is now. > > One possibility would be to use a strict key-closeness criterion for > small HTL (the same algorithm we have now), while weighting performance > issues more highly for larger HTL. For example, you could start with > a somewhat larger HTL, maybe 7 or so, and weight the algorithm so the > first 2 hops would be within the college network, and if they fail do > the remaining 5 hops out into the world. > > 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 #!/bin/perl -sp0777i<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<j]dsj $/=unpack('H*',$_);$_=`echo 16dio\U$k"SK$/SM$n\EsN0p[lN*1 lK[d2%Sa2/d0$^Ixp"|dc`;s/\W//g;$_=pack('H*',/((..)*)$/) _______________________________________________ Freenet-dev mailing list Freenet-dev at lists.sourceforge.net http://lists.sourceforge.net/mailman/listinfo/freenet-dev
