[arin-ppml] Micro-allocation policy proposal draft

[Martin Hannigan]

On Tue, Sep 30, 2014 at 7:12 PM, Bill Woodcock <woody at pch.net> wrote:
>> - increase the reserve pool to a /15
>> - increase the minimum allocation for an IXP to a /22
>
> Quadrupling the allocation while doubling the pool halves the number of IXPs served, and I think it would be unfortunate and short-sighted to let that happen.
>
> To inject some facts into the debate:
>
> http://www.keepeek.com/Digital-Asset-Management/oecd/science-and-technology/internet-traffic-exchange_5k918gpt130q-en#page78
>
> That graph is from 2011, when there were five IXPs with more than 255 participants.
>
> https://prefix.pch.net/applications/ixpdir/?new=1&show_inactive=1&sort=Participants&order=desc
>
> Today, three years later, there are six IXPs with more than 255 participants. So the portion of IXPs with more than 255 participants is holding steady at 1.5%.  In 2011, there were no IXPs with more than 512 participants, and today, there's one such, but it took sixteen years to get to that point.
>

Thanks. I spotted checked your data. Two problems. First, conclusions
from 2011 are woefully ancient. Second, your data is wrong e.g. ams-ix
you note as 617 and there current count is 680.There are quite a few
other inaccuracies contrasted to more mainstream data sources. Still,
it's irrelevant.

[ clip ]


> I support doubling the size of the reserved pool to a /15, but I don't think increasing the initial allocation size beyond a /24 is warranted yet.  I think sparse allocation is a sensible policy.  We can be reasonably certain that there will be at least 512 more IXPs before people stop caring about IPv4, but it's far from a sure bet that _any_ of those would grow beyond a /23 in that time.
>


I agree, doubling is reasonable. The allocation unit compromise is
making allocations from sparse boundaries no shorter than a /22 IMHO.

Where does this /15 come from? Existing free pool would make sense to me.

Best,

-M<