Le 9 oct. 2010 à 02:23, Brian E Carpenter a écrit : > The transition model in 1995 was based on the assumption that vendors > and ISPs would support dual stack globally well *before* IPv4 exhaustion. > > The fact that this did not happen is the problem. Agreed. Yet, the IETF has been IMHO, and to some extent still is, too slow to clarify the difference between DUAL-STACK SERVICE and DUAL-STACK ROUTING. In the absence of IPv6 service to hosts, generalized IPv4 address sharing will lead to port shortage, and will PROGRESSIVELY lead to intermittent and random connectivity breakages (the worse kind). With native IPv6 addresses offered to hosts (i.e. with THE IPv6 service), port shortages are completely avoided for all e2e IPv6 connections. (Besides, the danger of port shortages for the residual IPv4 traffic is consequently mitigated on paths that support the IPv6 service). Consequently, what users urgently need is DUAL-STACK HOSTS, with all the useful ways for their ISPs to assign them native IPv6 addresses (i.e. to offer IPv6 service). On the other hand, dual-stack routing isn't urgent, and may even, for some ISP, never need to be deployed. (They can first tunnel IPv6 across IPv4, and later directly move to residual IPv4 across IPv6-only). Regards, RD > > Brian > _______________________________________________ > Ietf mailing list > Ietf@xxxxxxxx > https://www.ietf.org/mailman/listinfo/ietf _______________________________________________ Ietf mailing list Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf