On Sat, 15 Apr 2000, Ian Clarke wrote: > > The qmail way, for what it's worth, is to keep a table that tracks hosts > > that have repeatedly failed to respond and not connect to such hosts for > > a period of time. Seems straightforward enough, and it's not something > > that needs to be recorded between server sessions. > > We could even be less tolerant and employ a "one strike - your out!" > strategy. One improvement on this is to, rather than just remove a node > address, overwrite it with a node address from the closest key to the > key whose node address we are replacing.
That depends if you are doing it on a reference level, then you can employ "one strike", but if we are actually blacklisting addresses we should be more lenient. I wrote a blacklisting thing which used a probability to figure out whether to actually attempt to send to a node, and this probability would be multiplied by some number (say .5) when connections failed, but incremented by another number (say .1) when they worked. I stored this with the node address in the HandshakeHandler (didn't quite have time to finish and commit). This was based on what some people were discussing here before, maybe we should use a "kill the reference" idea instead, though. > > By the way, could the Brit in the room please define the term "duff"? > > (-: > > duff (n.) "duf" - Screwed, shagged, broken, bolloxed, not entirely > working. I thought shagged meant fucked. As in literally. > Ian. > > _______________________________________________ > 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
