RE: [core-dev] number of duplicate TCP queries vs. locale preferencing

It does sound like we need to make this locale stuff more probabilistic.
If we are being strict in the HostCatcher then we should add some degree of
randomness and maybe only preference the locale by 25% for example =>
particularly if that mirrors our connection usage.  i.e. 1 of 3 leaf to
ultrapeer connections.

Thanks
-greg

-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Gregorio Roper
Sent: Wednesday, September 01, 2004 3:07 AM
To: [email protected]
Subject: [core-dev] number of duplicate TCP queries vs. locale
preferencing


I did some quick and dirty tests, running LimeWire for ~ 6 hours once as an
ultrapeer with locale "de" and as an ultrapeer with locale "en". While
preferencing "de" ultrapeers, about 45% of the queries I received were
duplicates, while preferencing "en" peers, only 25% of the incoming queries
were
duplicates.

I think the solution might be to reduce the amount of locale-prefenced hosts
stored in the HostCatcher or to force some randomness regarding the locales
of
the Endpoints returned by the HostCatcher. HostCatcher.getAnEndpoint() will
almost always return an Endpoint of your own locale if the locale universe
is
large enough to allow that.

mfg
gregorio
_______________________________________________
core-dev mailing list
[email protected]
http://www.limewire.org/mailman/listinfo/core-dev



_______________________________________________
core-dev mailing list
[email protected]
http://www.limewire.org/mailman/listinfo/core-dev