On 15-apr-2006, at 18:54, Christian Huitema wrote:
Clearly, the current set-up based on BGP and "default-free" tables is
not set to absorb more than a small number of PI prefixes -- maybe
thousands, maybe a few tens of thousands, certainly not a few
millions. But who says that we cannot change it?
10 years of discussions within the IETF and basic information theory
say routing won't work on the scale of the current and future
internet without aggregation with reasonable hardware price/
performace ratio assumptions.
after all, only requires a layer of indirection. We can certainly
And we have. It's called the DNS. But it seems people need more
indirection. Where do we draw the line? Portable street addresses?
Portable GPS coordinates?
It's really too bad, because we almost have what we need to pull this
whole scalability thing off in IPv6: stateless autoconfig, DHCPv6
prefix delegation, dynamic DNS updates, MIPv6, shim6. The only thing
missing is some finishing touches (ok, more than "some" for shim6,
but it is moving along) and for firewall vendors to jump on the
bandwagon and move away from hardcoded IP address based filters.
Ietf mailing list