> > Keith Moore wrote: > > - we need a new address space for globally-unique > > PI addresses (GUPIs) > > I might misread you, but I think that your goal is two-step: > 1. Make site-locals unique > 2. Make them globally routable (which would indeed suppress site-locals) > > 1. is a good idea but 2. is not going to work.
no, this isn't what my goals are. I don't want to make FEC0::/10s unique, nor do I want to make them globally routable. I want to discourage use of that prefix, and provide a separate set of addresses which are globally unique and routable off-site by private arrangement. If at some later date we figure out how to make them globally routable, so much the better. > > - we need an understanding that GUPIs are not > > globally routable unless/until we know how to make > > that routing scale - but NOT an architectural > > limitation against it > > Do you call recommending a default blackhole in routers an architectural > limitation? not sure. here's the question - let's say that 3 years from now we figure out how to make routing of GUPIs scale globally. having biased the last three years' worth of routers to filter them, how do we then upgrade the routers to not filter them? In general I'd like to minimize the number of hard-wired assumptions about addresses in both the addressing architecture and the code. I don't think we have enough foresight to know what our needs will be even five years from now. Keith -------------------------------------------------------------------- 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] --------------------------------------------------------------------