On 12 November 2016 at 09:45, Brian E Carpenter <brian.e.carpenter@xxxxxxxxx> wrote: > On 12/11/2016 03:51, Emily Shepherd wrote: >> On Fri, Nov 11, 2016 at 01:38:34PM +0000, Khaled Omar wrote: >>> You can find the latest version of the IPv10 draft attached in this e-mail. >> >> This looks fairly similar to just using a IPv4-mapped IPv6 address >> within an IPv6 packet [RFC4038]; is there a nuance I'm missing here? > > Not a big one. Using version number 10 isn't necessary; these could > be standard IPv6 packets. But of course it doesn't solve the basic problem > that makes dual-stack or a middlebox of some kind essential: an unmodified > IPv4 host can't talk to an IPv6 host, or an IPv10 host, because it doesn't > understand the new packet format. So this solution does nothing for backwards > compatibility, unfortunately. > I've thought it can be useful to think about how hard this problem is to solve in a different context. Imagine you wanted to achieve perfect translation between two spoken languages, meaning no lost of any meaning at all after translation, including nuance, and one of the languages only has 10% or less of the words the other does. Trying to translate between IPv4 and IPv6 perfectly is the same sort of problem. Regards, Mark.