On Mon, 24 Nov 2003 08:37:39 +0200 "EricLKlein" <[EMAIL PROTECTED]> wrote:
> Christian Huitema wrote: > > > Andrew, the draft has provision for both "registered unique local > > addresses" and "probably unique local addresses". The registered unique > > addresses are not valid on the Internet, but they definitely will not > > collide with other addresses. > > I am still have 2 concerns with these concepts: > 1. People do not want to register their secure internal network nodes (bank > central computes etc) so the "registered unique local addreses" may not meet > their needs. They do not want to have even theoritically reachable addresses > on a Cash machine or other secure node that needs to be connected as part of > the private network. > > 2. For the "approxiamtely" or "probably" unique local addresses I am > concerned that these addresses will eventually be assigned as part of the > registered addresses and can then be in conflict with "legitimate" nodes. > Probably not going to happen in the next 200 years or more, and more likely it will never happen. By the time that becomes a possibility, IPv7 or IPv8 will be ready, with, based on the IPv4 32 bit -> Ipv6 128 bit trend, 512 bit addresses ... of course, every bit added doubles the size of the address space. I'm sure somebody can provide a more accurate figure, but with all the currently planned IPv6 address space allocations, there is still in the order of 3/4 unallocated (not unassigned), for any purpose. That 3/4 would have to be used up before your concern becomes a possiblity. > So between the 2, I do not see a solution that will work better than a > RFC1918 type of address space. The more I hear about these options the more > I want to bring back site local addresses. > I suggest you have a read or review the following two IETF documents : "RFC 2993 - Architectural Implications of NAT" http://www.faqs.org/rfcs/rfc2993.html Site Locals opens the NAT can of worms in IPv6. NAT is worth avoiding. Review the SL deprecation draft. If nothing else, Section 2 - "Adverse effects of site local addresses" is worth reviewing again to understand why duplicated address spaces cause problems. http://www.ietf.org/internet-drafts/draft-ietf-ipv6-deprecate-site-local-02.txt The H/H addresses aren't perfect, but they do seem to be the best solution that meets the largest number of functionality goals, without resorting to Provider Independent addressing. PI would be perfect, except that it is likely to cause the default free routers to melt down in the longer term. As in much of life, it is a trade off. Regards, Mark. -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
