On 21 July 2011 12:21, Rémi Després <[email protected]> wrote: > > Le 21 juil. 2011 à 11:30, Wojciech Dec a écrit : > >> On 21 July 2011 11:06, Rémi Després <[email protected]> wrote: >>> >>> Le 20 juil. 2011 à 18:25, Wojciech Dec a écrit : >>> >>>>>> ... >>>>>>> If packets are sent to that address, they >>>>>>> get passed to NAT44. Hosts behind the CPE or apps on the CPE will not >>>>>>> use/bind to that address. >>> >>>>>> >>>>>> AFAIK, this prohibits using the full CE-assigned IPv6 on the CE LAN side. >>>>>> Such a constraint isn't needed with "4V6 encapsulation". >>>>> >>>>> No. >>> >>> A justification of this "no" is missing, because: >> >> It's not missing. Please do read once more the rest of the mail, as >> well as those of others. >> >>> - The CE 4V6 address is synonymous with a valid IPv6 address in the range >>> defined by the CE IPv6 prefix. >>> - With 4V6T, a translated IPv4 packet entering a 4V6 site cannot be >>> distinguished from an IPv6 packet destined to this valid IPv6 address. >>> >>> An example: >>> >>> Host >>> H >>> +-+ router >>> | |< 2001:db8:a:: >>> | |--------------. .-. >>> | | | | |< 2001:db8:a::/64 4V6T CE >>> +-+ LAN |--| |-----------------. +-+ >>> | | | | | |< 2001:db8:a:/56 >>> --------------' '-' LAN |---| |------------ >>> | +-+ <= IPv6 packet >>> ----------------' sent to host H >>> >>> - The IPv6 address of host H has a permitted value (at least per sec 3.2.1 >>> of RFC 3041 on privacy addresses). >> >> No. >> The above indicates some confusion between a prefix and an address >> Aside from the fact that Host H above has an address that is clearly >> not derived via SLAAC (not to mention very bogus in the context of >> rfc3041), > >> on the above link you either have TWO/THREE(?) devices with >> the same IP address (2001:db8:a::) > > Not two or three! Just one. > Having a delegated prefix at an interface (e.g. < 2001:db8:a::/64) doesn't > mean that this prefix is the interface address, does it? > Confusing prefixes, like 2001:db8:a::/64), and addresses, like 2001:db8:a::, > seems to be what YOU did, not me ;-).
Your diagram says otherwise, as it shows the routers either without an IP address, or with bogus subnetting (subnet 0 cannot be in two places/links). Good luck with making the set-up work in any case. > > As you know, we both agreed, in an of-list discussion, that the IID of 4V6 > addresses would better be a 1 than a 0. > This would avoid, as you pointed out, a conflict the IID of the Subnet-Router > anycast address of RFC 4291. > This being done, replace the host address by 2001:db8:a::1, and the example > holds better. > > If the IID of 4V6 addresses is not permitted on a link, the 4V6 CE's could be > difficult to reach from the Internet. > If it is a permitted one, a host behind the CE may have this IID. Indeed, and what you seem to be trying very hard to say is that building an IPv6 network where multiple IPv6 devices have *the same* IPv6 address is ok and reasonable, and will somehow work unambiguously with the encapsulated mode... Interesting... In general, this is known as an IPv6 addressing violation/conflict, and there are specific procedures to prevent that for good reason. Thanks, Woj. > > Regards, > RD > >> all on different links, or the 4V6 >> routers without any address. In either case the set-up is not correct >> (or at least your model set-up is the problem). >> >> Thanks, >> Woj. > > > _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
