El Martes, 12 de Septiembre de 2006 07:34, Attila escribió: > On Montag, 11. September 2006 22:56 Thomas Bächler wrote: > >> Playing with tcp_rmen is a bad idea since the problem actualy still > >> exists. It's barely a hack which is often not enough. If Arch wants to > >> provide a workaround, I think it must be to disable TCP Window Scaling: > >> > >> echo "net.ipv4.tcp_default_win_scale = 0" >> /etc/sysctl.conf > > > > Could you provide these solutions and the explanations on > > http://wiki.archlinux.org/index.php/Configuring_network in a separate > > troubleshooting section? People will search for solutions there. > > A little hint how to test it would be fine too because me normal user see > in this thread that the download rate is bad under certain cases. > > But my problem with this very interestic informations here is that my > download rate seems to be okay (about 700 KB/s with my 6 Mbit/s download > flat rate) if i download something from a server in usa to here which is in > germany. Could this be because my chain at home looks like this > > Desktop PC (ARCH) <-> Router/Server (SuSE) <-> Cable Modem > > and so the suse pc (where i run squid as transparent proxy) and/or the > cable modem "repairs" the problems with this factor of the tcp window of my > arch pc?
Once the problem arises, it cannot be "reapaired". The problem with window scaling is that routers that can't handle it drop part of the packets; the greater the scale factor beeing used, the more packets are dropped. The only way to recover dropped packets is retranmitting them, which wastes bandwith. But this problem doesn't always happen. In fact, it usually doesn't, as most routers do support window scaling. In my case, it is my own dsl router which doesn't support scaling, so I faced the problem every time. > Sorry, if this question is too stupid.-) No question is stupid. Not asking is what may be stupid. _______________________________________________ arch mailing list [email protected] http://www.archlinux.org/mailman/listinfo/arch
