In your letter dated Tue, 19 Jul 2011 22:28:03 -0700 you wrote: >On 7/19/11 6:02 PM, Brian E Carpenter wrote: >> For example, if you're tunneling IPv6 over an IPv4 network whose PMTU (to >> the other end of the tunnel) is, to take a random example, 576, the tunnel >> end points could use IPv4 fragmentation and reassembly to provide a 1280 MTU >> for the IPv6 traffic. > >That would be true for tunneling IPv6 over IPv4. >But the topic was the wording in RFC 2460 that was added to make it >feasible to build (stateless) translators between IPv6 and IPv4.
Personally, I think it is just bad design to require extra complexity in every IPv6 stack in the world, just to accodate a transition technology that essentially hasn't been used in the last 15 years, that provides no benefit to hosts that allocate fragment IDs randomly (as is good security practice for IPv4) and provides marginal benefits for hosts that allocate IDs sequentially. I would suggest that the stateless translator protocols be updated to not use this feature and allocate a random 16-bit ID themselves. And schedule the feature to be remove from the IPv6 standard. -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
