on Fair NAT: ------------- [...] # Correcting TOS for large packets with Minimize-Delay-TOS $BIN_IPT -t mangle -A $FN_CHK_TOS -p tcp -m length --length 0:512 -j RETURN $BIN_IPT -t mangle -A $FN_CHK_TOS -p udp -m length --length 0:1024 -j RETURN $BIN_IPT -t mangle -A $FN_CHK_TOS -j TOS --set-tos Maximize-Throughput $BIN_IPT -t mangle -A $FN_CHK_TOS -j RETURN [...] I think that best is: [...] # Correcting TOS for large packets with Minimize-Delay-TOS $BIN_IPT -t mangle -A $FN_CHK_TOS -p tcp -m length --length 0:512 -j RETURN $BIN_IPT -t mangle -A $FN_CHK_TOS -p udp -m length --length 0:1024 -j RETURN ## NEW LINE: $BIN_IPT -t mangle -A $FN_CHK_TOS -p icmp -m length --length 0:1024 -j RETURN $BIN_IPT -t mangle -A $FN_CHK_TOS -j TOS --set-tos Maximize-Throughput $BIN_IPT -t mangle -A $FN_CHK_TOS -j RETURN [...] because on original FARINAT, PINGS are marked as "MAXIMIZE THROUGHPUT" is correct my note ?? bests andres
On Friday 25 November 2005 21:19, LinuXKiD wrote:> $BIN_IPT -t mangle -A $FN_CHK_TOS -p icmp -m length --length 0:1024 -j > RETURN > [...] > > because on original FARINAT, PINGS are marked as "MAXIMIZE THROUGHPUT" > > is correct my note ??The code you quoted is a rule that should be protocol independent. Lower the priority of packets that claim to be interactive but are too big. Or rather, let packets keep their high priority status only if they''re small enough. It should apply to ICMP and all others as well, but obviously it doesn''t. Well spotted... I don''t like that part of the script. It''s pretty much random, I can''t even remember why I''m using 0:512 for tcp, but 0:1024 for udp as packet size criteria. The prioritization of Fair NAT is still way too static, given that everyone has different requirements. Andreas Klauer
-> On Friday 25 November 2005 21:19, LinuXKiD wrote: -> > $BIN_IPT -t mangle -A $FN_CHK_TOS -p icmp -m length --length 0:1024 -j -> > RETURN -> > [...] -> > -> > because on original FARINAT, PINGS are marked as "MAXIMIZE THROUGHPUT" -> > -> > is correct my note ?? -> -> The code you quoted is a rule that should be protocol independent. Lower -> the priority of packets that claim to be interactive but are too big. Or -> rather, let packets keep their high priority status only if -> they''re small -> enough. It should apply to ICMP and all others as well, but obviously it -> doesn''t. Well spotted... -> OK, what about: $BIN_IPT -t mangle -A $FN_CHK_TOS -m length --length 0:512 -j RETURN instead... $BIN_IPT -t mangle -A $FN_CHK_TOS -p icmp -m length --length 0:1024 -j RETURN I think that (in ICMP case) would priorize small ICMP packets. and other small packets on other protocol. BTW, I think that FairNat and JiM QoS scripts are Great! I''m studing both in order to learn about Qos. bests andres.