begin Neil Barsema quotation of Fri, May 04, 2001 at 11:36:33AM +0200: > To tackle this I sugested a while back to have several parrallel requests > instead of just one, and then use the quickest response for the refference, > this would introduce a bias towards nodes that are best suited. All factors > would then count (key closeness, datastore size but also physical network > topology, bandwith, usage, etc.). > Offcourse this would (dramaticly) increase the network traffic.
If you only increase traffic among very close nodes, no big deal. The "last mile" is going to be what matters to most nodeops, so it's important to address the degenerate case. For most nodes, there will be other nodes that are effectively "free" to talk to -- you wouldn't mind if every request went to all of them, because the connection to them is very large. For example, a lot of homes and offices have 100MBps switched Ethernet internally and DSL to the outside world. The nodeop would want to take advantage of every opportunity to get something from another node on the LAN, but once a request hits the DSL it doesn't matter if it's going across the street or to Uzbekistan. -- Don Marti "I've never sent or received a GIF in my life." dmarti at zgp.org -- Bruce Schneier, Secrets and Lies, p. 246. http://zgp.org/~dmarti/ (Free the Web: http://burnallgifs.org/) _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
