[arin-ppml] Draft Policy ARIN-2014-7: Section 4.4 Micro Allocation Conservation Update

[Owen DeLong]

> 
> Not really responding to you, you just happened to be the last in the thread..
> 
> Perhaps we should look at tackling some of our dwindling number
> resources issues in a different perspective.  Have we considered
> updating the policy to only issue prefix sizes which are reasonable in
> the first place?  What makes just setting up an IXP be enough to issue
> a /24?  What if this IXP is in a market in which there will never be
> more than 126 participants?  Or worse much less?  Should these IXPs be
> given /24s when a much smaller allocation may be all that's needed?
> Or should every IXP have to start small and as their participation
> increases they be issued new space to move into?
> 
> I believe the argument for global prefix visibility of IXP space has
> been largely discussed and consensus is that this space does not and
> should not be globally reachable voiding any perceived need for a /24
> I believe.

While such an optimization might be possible, I would argue that the potential
benefit is not sufficient to justify the effort.

Let’s assume a total ARIN region deployment of 25,000 IXPs. Further let’s
assume that roughly 1/4 of them fit your criteria of no potential for more than
126 participants. Let’s bound the smallest practical IXP at a /26 (62 participants).

Best case, if all of the ~6,250 participants could fit in /26s, you would save
6,250*3/4 = 4,687.5 /24s. Rounding that down (for numerical convenience
to 4096, that’s a total savings of a single /12.

In reality, the actual savings are virtually guaranteed to be much much less,
probably more like 1 or maybe 2 digits of /24s.

It’s time to realize that attempts to forestall the runout of IPv4 are futile and
stop optimizing the packing of the deck chairs.

Owen