On Wednesday 14 February 2001 18:10, Peter Todd wrote: > On Wed, 14 Feb 2001, you wrote: >> ... > > I re-ran my key validator on all the keys in Steve's key index yesterday. > > Only 95 out of 2340 were reachable at an htl of 30! > > Did you run the validator on the same data set as you used the first > time? From what I can see a pile of crud just got dumped onto the key > index. :( No I ran it on a snapshot taken on the morning of 13 February. Though I think that keys have only been added, not removed.
> > I originally planned to run a second pass to double check the failed keys > > but I stopped after the first pass fearing that I might damage the > > network. I have noticed that if you make too many failed requests in a > > short period of time it seems to kill routing. > > If failed requests can kill routing I think we've got a serious > problem... This assertion is based on observing the code run, not reading the code and understanding it. By the way where should I look in the source to really understand how routing is implemented? Maybe there's some other explanation. I've just observed that if I do a lot of failed requests on my local node it eventually gets into a state where almost every request for a known good URI fails. > We should only drop nodes when requests fail due to a > *connection* problem. Though let me guess... We already do that and > it's bugs that are causing the connection failures? :( > My gut feeling is that this is exactly what's happening. -- gj -- Web page inside Freenet: freenet:MSK at SSK@enI8YFo3gj8UVh-Au0HpKMftf6QQAgE/homepage// _______________________________________________ Devl mailing list Devl at freenetproject.org http://www.uprizer.com/mailman/listinfo/devl
