The following email, sent by lorenzo to the mailing list didn't reach it because he isn't registered in Softwire. Upon his request, here is a retransmission. Regards, RD
> De : Lorenzo Colitti <[email protected]> > Date : 2012-07-06 10:59 > À : Rémi Després <[email protected]> > Cc : "[email protected] WG" <[email protected]> > Objet : Rép : TCP/IPv6 faster than TCP/IPv4/IPv6? > > On Thu, Jul 5, 2012 at 12:04 AM, Rémi Després <[email protected]> > wrote: > There has been discussions, during IETF 83 and after, to warn that, in some > circumstances, direct TCP over IPv6 can be significantly faster than > TCP/IPv4/IPv6. > A cc is sent to Lorenzo, because he privately reported some direct knowledge > of the subject (I personally have none). > > What I know is: some routers only process UDP/IPv6 and/or TCP/IPv6 on the > fast path and IPv4/IPv6 on the slow path. I know this because I have one such > router at home, and I believe that millions of users in Japan do as well. The > devices are used to connect to the NTT East/West closed IPv6 network, which > provides 100M or 1G link speeds. Some of these users have IPv6 Internet > access over these pipes, and some ISPs are thinking of tunneling IPv4 over > IPv6 to take advantage of the fast IPv6 pipe (IPv4 only provides 200M). > > My particular router will do 1Gbps of TCP/IPv6 but only about 300Mbps of > TCP/IPv4/IPv6. This particular device is fairly recent, however; I understand > that older devices have worse performance (as low as a few Mbps). > > I realize this will very much sound like a first-world problem to most people > (my router will only do 300Mbps!), but nevertheless, this is a barrier to > adoption because a service that reduces users's bandwidth by 3x to 100x is > very hard to for an operator launch. > > Cheers, > Lorenzo
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
