[arin-ppml] Policy Proposal 114: RFC1918 as initial ISP multihomed criteria

[Joe Maimon]

George, Wes E IV [NTK] wrote:
> I oppose this proposal.
>
> I think this is probably a legitimate issue, but given that it seems to refer primarily to IPv4 space...well...
> The last couple of proposals dealt with how to manage the last /8 of IPv4 space. At its fastest, we're talking about 6-12 months before this policy could make its way through the PDP and be adopted.
> http://www.potaroo.net/tools/ipv4/
> Have you ever heard the phrase "rearranging deck chairs on the Titanic"?
>
> There's simply not enough IPv4 space left to be seriously considering making the means to qualify for it more permissive, however noble the goals of ridding the world of NATs and cutting down on renumbering may be. The reality is that it's going to get worse before it gets better, because when we start running out, people are going to choose NATs as a way to stretch the IPv4 space they do have. I'd be willing to bet that the same networks that don't have the resources to renumber twice and are just now becoming multihomed also won't have the resources to get PI IP space on the transfer market, so unless they get lucky on the waiting list (if that gets adopted), chances are quite likely that at best, networks so situated will end up renumbering either just once (into PA space, because there won't be any PI to be had) or not at all.

I disagree.

1 - Even if things are as you say, there is about a year or so of 
difference it may make.

2 - PP110 and PP112 and potentially others would probably make space 
available for exactly these kinds of orgs for well longer than the date 
that the first justified request is denied due to lack of resources.

3 - Fragmentation and continued returns/reclaims would probably provide 
the above conditions as well.

4 - Transfers under 8.3 needs justification as well - presumably 4.2.2 
would still apply to them as applicable.

5 - The projections may be off one way or the other.

6 - If you believe it is too late to do anything, then dont oppose it, 
its a harmless no-op.


>
> If this is still a problem in terms of a network's ability to qualify for IPv6 addresses (and I'm not convinced that it is), it might make sense to modify this proposal to update NRPM section 6 (taking into account other proposals like 2010-4 that are already in progress), so that those unfortunate networks in this position are not hamstrung on the IPv6 side.
>
> Thanks,
> Wes


I would support a proposal where any address resources, routed or not, 
upstream or not, rfc1918 or not, counted towards utilization 
requirements if they were included in the subsequent renumbering 
requirements.

I believe simply removing the words "from an upstream" could cause the 
NRPM to imply this.

For example, if you have had to address your customers from rfc1918 
because you didnt have enough global ipv4, then you would qualify and 
would have to renumber those customers for additional resource requests 
to be approved, as is the case currently had you used PA space for the 
customers.

But if you burnt through /16 of rfc1918 for serial addressing and 
similar and have no intention of renumbering it, then no.

The fundamental problem is that justifying consumption based upon 
consumption does not in itself reward conservation and efficiency.

Instead it penalizes it.


Joe