In message <e230de23-ad00-4f3d-b384-ba52fa7b3...@delong.com>, Owen DeLong writes: > > On Jun 6, 2011, at 4:49 PM, Mark Andrews wrote: > > >=20 > > In message <b53bef53-f327-44ed-8f23-a85042e99...@delong.com>, Owen = > DeLong write > > s: > >>=20 > >> On Jun 6, 2011, at 2:23 PM, Mark Andrews wrote: > >>=20 > >>> =3D20 > >>> In message <alpine.bsf.2.00.1106060732190.68...@goat.gigo.com>, = > Jason =3D > >> Fesler wr > >>> ites: > >>>>> But anyway, just consider it: a portion of the major websites go > >>>>> IPv6-only for 24 hours. What happens is that well, 99% of the =3D > >> populace > >>>>> can't reach them anymore, as the known ones are down, they start =3D= > > >> calling > >>>>> and thus overloading the helpdesks of their ISPs. > >>>> =3D20 > >>>> Won't happen this year or next. Too much money at stake for the = > web=3D20=3D > >>=20 > >>>> sites. Only when IPv4 is single digits or less could this be = > even=3D20 > >>>> remotely considered. Even the 0.05% hit for a day was controverial = > =3D > >> at=3D20 > >>>> $dayjob. > >>> =3D20 > >>> IPv4 will never reach those figures. IPv6 isn't preferenced enough = > =3D > >> for > >>> that to happen and IPv6-only sites have methods of reaching IPv4 = > only > >>> sites (DS-Lite, NAT64/DNS64). > >>=20 > >> I think you'll be surprised over time. Given the tendency of the =3D > >> internet > >> to nearly double in size every 2 years or so, it only takes 7 cycles = > =3D > >> (about > >> 15 years) for the existing network to become a single-digit = > percentage > >> of the future network. > >>=20 > >> Owen > >=20 > > And without there being a strong IPv6 bias in the clients they will > > continue to use IPv4/IPv6 on a 50/50 basis. I would be quite happy > > to be proven wrong and only time will tell. > >=20 > Almost every client does have a strong IPv6 bias if they have what > appears to be native connectivity. The bias degrades rapidly with > other forms of host connectivity. > > My linux and Mac systems certainly seem to strongly prefer IPv6 > from my home. YMMV.
Things like happy-eyeballs diminish it even with perfect IPv6 connectivity. 100ms rtt doesn't cover the world and to make multi-homed servers (includes dual stack) work well clients will make additional connections. > Owen -- Mark Andrews, ISC 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: ma...@isc.org
