This doesn't resolve the problem of ambiguous subnet prefixes when routing domains merge. So it doesn't go far enough IMHO.
Brian Andrew White wrote: > > Let's ask a different question. Would the following be acceptable: > > ----- > The address space FEC0::/10 is reserved for non-global use. It is intended > not to be globally routeable. All routers MUST by default blackhole any > packet destined to FEC0::/10, and MAY return a 'destination unreachable' > message. > > 'Sites' using FEC0::/10 addresses MUST implement a filter at the 'site > border' that discards source or destination addresses in the FEC0::/10 > space. Routing protocols MUST NOT exchange reachability information > concerning FEC0::/10 across the border. > > Any router or node not explicitly configured to do otherwise MAY discard > (silently or otherwise) any packet with a source or destination address in > FEC0::/10 space. > > Applications MAY choose to treat FEC0::/10 addresses differently to other > addresses, and MAY prefer or disprefer them. Applications MAY assume that > FEC0::/10 addresses will be filtered before reaching the global internet. > > ----- > > This seems to cover the minimum requirements of the relevant parties. The > only global requirement is that all routers by default black-hole > FEC0::/10. If you choose to use site local addresses, then you come under > the border router requirements given in the second paragraph. > > The only bugbear I can see is source address selection. One easy solution > is to 'prefer closest match' to whatever destination address the application > selected, which will mean SL matches SL and link-local matches link local. > Ensuring non-SL matches non-SL may be slightly trickier, depending on > prefixes, though the current policy of allocating from 0000::/12 will prefer > global-global. > > And maybe something should be added to say "If you stick these things in a > globally accessible DNS, don't be surprised when connections to your hosts > fail." And maybe a policy to NXDOMAIN reverse lookups for > [C-F].E.F.in6.arpa. > > In short, there is almost no extra effort for those who don't implement SL > addresses - all the work is pushed to those who do. > > Notice that the above text says nothing about how FEC0::/10 addresses are to > be allocated. All it does is reserve the space as 'not globally routeable' > and put policies in place to stop this information getting where it > shouldn't. > > -- > Andrew White [EMAIL PROTECTED] -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
