Re: Extending a /64 (ATN/IPS worked example)

[Michael Richardson <[e-mail]>]

Brian E Carpenter <[e-mail]> wrote:
    > To be clear, the IPv6 *fixed length* address model changes the parameters
    > of allocation practice because of moving from (say) 24 to (say) 64 bits
    > of routeable prefix, but in no way changes the philosophy of allocation
    > practice. IPv6 has ~64 topology bits instead of ~24; the actual numbers
    > in the H-ratio calculation change; the potential lifetime of the address
    > space expands enormously; the economic value of address space collapse
    > enormously. But all of that breaks if you start assigning address bits
    > non-topologically. That's why IPv6 and IPv4 share CIDR as the basis
    > for both prefix allocation and wide-area routing.

ISPs that try to allocate their IPv6 residential customers based upon strict
geographical/network topology find two things:
  1) they have the wrong amount of space in the wrong place at the wrong time.
  2) they force users to renumber needlessly when they have to split
     customers in a neighbour among two routers/CMTS/BMS.

So I claim that in the 16 bits between /32 to /48 within an ISP, that there
is little in the way of network topology.  Yes, some structure, with some
layers of abstraction... when possible... but OSPFv3 deals.

I've heard that in the future that Smart ISPs will use some kind of
address allocation/management autonomic service agent over a secure autonomic
control plane.

So, instead of thinking of this as a non-topological address plan, think of
this as the internals of a really large confederated airline ISPs.

--
Michael Richardson <[e-mail]>   . o O ( IPv6 IøT consulting )
           Sandelman Software Works Inc, Ottawa and Worldwide