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

[Owen DeLong]

On Jul 29, 2010, at 11:34 AM, William Herrin wrote:

> On Wed, Jul 28, 2010 at 5:21 AM, Joe Maimon <jmaimon at chl.com> wrote:
>> William Herrin wrote:
>>> I don't think we even give 'em point to point links. For the last /8
>>> the vendors can damn well fix their code to originate ICMP from the
>>> loop0 address instead of the RFC1918 address on the interface.
>> 
>> I completely agree. That feature would be really lovely along with other
>> control plane traffic handling improvements and wider availability of proper
>> address abstraction off of the physical interface.
> 
> How much support would there be for a policy proposal to exclude point
> to point links as a justification for any global IP addresses
> effective, say, 1/1/2012? Along with a stern recommendation from ARIN
> to the routing vendors that they update their software to prevent the
> non-availability of of addresses for point to point links from causing
> malfunctions with ICMP warnings and errors?
> 
> You'd still be able to justify an IP address for the router, of
> course, but you wouldn't be able to justify any addresses for the
> individual point to point links, regardless of technology employed. So
> you'd end up using unnumbered serial interfaces and RFC1918 addresses
> on the point to point ethernets.
> 
I would oppose such a policy. There simply aren't enough point to point
links to add up to a meaningful number of addresses. To put this in
perspective, let's assume a moderate sized ISP which has a total
address span of roughly a /14 (262,144 addresses) has as many
as 1,000 point to point links. That's still 4,000/262,144 addresses
or 1.5% of their total address utilization.

The problems caused by such a policy would exceed any possible
address conservation achieved.

Owen