[arin-ppml] [Fwd: Draft Policy 2011-5: Shared Transition Space for IPv4 Address Extension]
On Feb 25, 2011, at 10:55 AM, George, Wes E [NTK] wrote:
> We also have differing opinions regarding the suitability of alternatives to prevent #5:
> a) an ISP repurposes some of their own space to use as inside NAT pool, with the understanding that
> implementing CGN will actually free up addresses that are in use today by those users moved to the
> b) squatting on unrouted but allocated space
> c) using whatever RFC1918 space meets the 80/20 rule for avoiding CPE gear defaults to ensure that
> the risk of encountering #5 is lowered, and when it happens, is more manageable to fix.
> If you are in the camp that believes that the above would solve most of the problems you listed,
> then you're wondering what problem this proposal is trying to solve. If you believe that none of the
> above are workable, or aren't workable enough for the majority of applications, then you likely
> believe that this proposal is the only correct answer.
I agree with this portrayal. Further, I've actually supported networks that were built according to each approach (a,b,c). And while they *can* be made to work, from my experiences I say that a shared transition space would be a valuable improvement.
> Finally, in addition to your questions, I'd add, "What happens if this proposal doesn't pass?" or
> perhaps "*Should* this problem be solved with policy?" Here again, we have nothing but educated
> guesses to go on, and I'm not sure "the Mega-ISPs will come in and make lots of requests for space
> for their CGN and we have to avoid that" is a good enough justification on its own, and I'm not in
> favor of policy for policy's sake.
According to my count, a /10 represents about 5% of the existing ARIN free pool (including IANA "holes"). I think it's safe to assume that at least some ISPs will use this shared transition space. So the question is whether that collection of ISPs would use more than 5% of the ARIN free pool.
I've had conversations with multiple large ISPs about their CGN deployment plans, and collectively they would support 4-5x more customers behind a CGN than a single /10 would facilitate. Even if only 1 or 2 of these large ISPs follow-through with their plans, I think there is savings to the ARIN free pool.