[arin-ppml] Ending point to point links as a justification for a /30?

[Owen DeLong]

On Jul 29, 2010, at 7:22 PM, William Herrin wrote:

> On Thu, Jul 29, 2010 at 12:44 PM, Owen DeLong <owen at delong.com> wrote:
>> Express a 128 bit number in 32 bits such that all possible 128 bit values
>> are uniquely expressed in the 32 bit field.
> 
> Hi Owen,
> 
> It's been discussed. It was called something like "aggregation with
> increasing scope." The basic idea works like "loose source routing."
> The first address in the path is the anycast top-level waypoint for
> the destination system, followed by additional waypoints until you
> enter the final scope. The DNS for this returns a list of 32-bit
> values as the address instead of a single 32-bit or single 128-bit
> value.
> 
Right... Now, how does an un-upgraded IPv4 only host with no code
changes take advantage of this and start sending packets to IPv6-only
hosts in other scopes?

> One of the nice features is how easily it expands and contracts. If
> you need more addresses inside your system, you just add one level of
> scope and presto, you have 4 billion more addresses to play with, no
> registry fuss required.
> 
Right.. A rather pathological form of variable length addressing.  However,
not really a solution to the stated problem. Merely an obfuscation that
moves the problem around until you can pretend it is invisible. However,
without changing the code on the originating machine, it still doesn't
actually work.

Owen