Disclaimer: I started out reading the whole thread, but lately I've had to skim. I might have missed it if someone already said this. If people want to use private addresses to avoid having to renumber when their provider renumbers, and they use addresses their provider won't route, then they need to NAT those addresses if they want the hosts using them to be able to talk to the rest of the world. Fine. In the old days, we recommended that sites register blocks of addresses whether they were connecting them to the net or not, if they ever intended to connect them to the net in the future. That way, each address could be globally unique, since they were all going through a central registry. For several reasons, we began to have to put more stringent conditions on granting address blocks out of the pool. People stopped being able to just ask for a /24 for their own use, and get it. Partly this was because we saw the address space running out if we kept letting anyone get address space, and partly it was because people expected routing for the addresses they were granted. RFC1918 was a solution to the problems in the last paragraph. The drawbacks of RFC1918 are obvious. If an address leaks, it might make its way to somewhere where it looks locally valid, by coincidence. And when networks merge, they may have collisions. We live with it because we need it. If the IPv6 address space is big enough, we should be able to revert to the old ways. We should be able to make a policy for the RIRs that allows, once again, for people to ask for unique address space. Do we have enough address space that everyone can get a reasonable amount just by asking for it? If so, people can NAT their private nets to their heart's content, without requiring anything like RFC1918 or site-local. I've done the same thing with IPv4 with old portable address blocks. I think today's implied identity of "private address space" and "locally scoped addresses" is really muddying the discussion. If we have a world where anyone who asks can get a unique address block assigned to them to use for their private network... THEN what are the remaining reasons why people would want to use site-local addresses? -- Cos (Ofer Inbar) -- cos@aaaaa.org http://cos.polyamory.org/ "OSI is a beautiful dream, and TCP/IP is living it!" -- Einar Stefferud <Stef@nma.com>, IETF mailing list, 12 May 1992