Hi, Jacni, Le 20 juil. 2011 à 05:07, Jacni Qin a écrit :
> hi Authors, all, > > I skimed the draft and got a question on the "Domain IPv4 Prefix" involved in > the algorithm, > > Since currently the IPv4 blocks used for accessing residential users are > probably fragmented (or not that continuous), I guess the there will be too > many domains, IMHO. During the later stage of phasing out IPv4, that might be > acceptable. Plus, it makes the algorithm complicated. So, how about removing > it, and just using the EA-bits? While, of cource the tradeoff will be shorter > IPv6 prefix is required. > Please correct me if I missed something. :-) > > Thoughts? In addition to Satoru's answer, an ISP that has many IPv4 prefixes can: - use only a few, or even only one, of its shortest IPv4 prefixes, and - use IPv4 address sharing to support as many CE's as needed. Example: an ISP has a /10, a /11, another /11, a /14, a /15, and a /16 (a real example I met). It can: - use three domains (the /10 and the two /11's), or - use only the /10, extending Port-set ID by one bit to support as many CE's In both cases, no entropy due to the fragmented IPv4 space needs to be exported to the IPv6 addressing plan. Leftover IPv4 prefixes can then be returned to the community (for free... or for a price, whatever applies). Regards, RD > > > Cheers, > Jacni > > _______________________________________________ > Softwires mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/softwires
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
