Daniel Roesen wrote:
On Fri, Oct 14, 2005 at 01:11:18PM -0700, Crist Clark wrote:
Actually, doing multihoming and getting PI space are orthogonal in shim6 last I knew. That is, you could get address space from your N providers and have one of the providers, say Provider X, to be the ULID for the end points. Should Provider X's link(s) go down, shim6 will ensure it all still works (which is, after all, the whole point).
But when the contract with this supplier is terminated, I have to renumber the whole network. Excellent.
Well, you can get by on shim6 until the supplier reissues the address space elsewhere. Not too reassuring? But with IPv6, renumbering is easy! Uh, yeah, right.
Getting PI space is really an administrative and economic issue. It is not a technical requirement of shim6.
That's the problem. Ignorance regarding economical problems. We don't have technical problems, we have economical ones (if at all).
Multihoming in IPv6 is a technical problem if you consider unrestricted growth of the routing tables to be a technical problem (there are those who think is it economic, "throw more memory and CPU at it"). As the section of the draft quoted, using unrouted "RFC1918-like" address space, which would presumably be PI, is also covered by shim6, but it is not the only way in which it work. There is plenty of finger pointing about the economic problems, or ignorance thereof, in IPv6, including the occasional conspiracy theory. We had similar problems with IPv4. We still feel pain from the switch from Classful networking to CIDR occasionally. That was a technical change driven by an economic reality. Then there is that evil spawn of IPv4, NAT. The way these things get worked out ain't always pretty. -- Crist J. Clark crist.clark@globalstar.com Globalstar Communications (408) 933-4387