[arin-ppml] Draft Policy ARIN-2013-3: Tiny IPv6 Allocations for ISPs

[Owen DeLong]

On Apr 7, 2013, at 15:32 , Matthew Kaufman <matthew at matthew.at> wrote:

> On 4/7/2013 2:14 PM, Owen DeLong wrote:
>>>> One possible way to distinguish ISP size for IPv6 would be to use
>>>> annual gross revenue instead of total address holdings. I don't know
>>>> if that would be an improvement or not. I realize it would be more
>>>> complex to calculate and enforce. I think it would likely be more fair.
>>> More fair would be everyone pays the same. Or everyone pays based on the amount of service they actually require from ARIN.
>> I disagree completely here. An end-user with a single /48 is not in any way reasonably expected to pay the same as an ISP with multiple ASNs, a /20 of IPv6 and an aggregate /6 of IPv4 space.
>> 
>> I can't imagine in what possible universe anyone could imagine that to be fair.
> 
> Why not? As long as the price is reasonable I don't see who could possibly be opposed to a model where the only added costs are for those who actually cost ARIN more to service, not some arbitrary scaled scheme where well-run large orgs pay more than the tiny ones that need lots of handholding. Or, even better, a flat price for all. But of course there *would* be some opposed, because for every choice it isn't "fair" to someone.
> 

I noted the reasons below. It puts strong incentives in bad directions.

> This is just like the US tax system... nobody thinks that any of the models are "fair". Flat tax? Unfair to the poor. Higher taxes for the rich? Disincentive to working harder and unfair to the rich. You can't win.
> 

Actually, I think that a flat consumption tax that excludes food and clothing items and written word matter (regardless of media) under $100 would be quite fair. It isn't unfair to the rich because it's based on consumption, not earning. However, this is far afield from PPML's purpose.

> And that's why we shouldn't be discussing the fee structure here at all. We should be discussing what numbering policy results in the best addressing plan for the Internet. And I believe that handing out /36 and /40 to ISPs doesn't do that.
> 

I agree with you, but the fee structure, like it or not, is going to drive consideration of those policies, or, we're going to need a discussion somewhere that drives a better fee structure.

>>> Basing it on amount of space *or* gross revenue isn't "fair"... it is just a way to extract extra revenue from those who (in theory) have more to send to ARIN, which gives ARIN the opportunity to lower prices at the other end to keep the masses happy, or to just increase its overhead to ensure the money gets spent.
>> We can agree to disagree on what constitutes "fair" here. I believe it is "fair" for those that are extracting a greater amount of revenue from the community resources they hold to contribute a greater share to the costs of maintaining said community resources.
> 
> So we should be charging mobile operators more than we charge dialup providers because the ARPU is higher?
> 

I'm not necessarily opposed, but I'm also not convinced that's true, either.

>> 
>> I don't see it so much as extracting extra revenue so much as differentiating the allocation of the total costs.
> 
> Those are exactly the same thing. One just sounds better than the other.
> 
> The idea is that you find the people who are willing to pay more, you figure out how to charge them more. In this case, you take some of that "extra" revenue and use it to drop the prices for the people who aren't willing to pay as much. Classic pricing theory covers this, and I would expect no less from any organization. But it doesn't make it "fair" either.

Another way of considering it, however, is you try to find ways to serve those that couldn't pay as much as it would take for everyone to pay the same amount.

Over the time I've been involved in ARIN, combinations of better policy and fee reductions over time have created a better ability to serve a wider range of organizations. I think this is a good thing. Many organizations would simply be excluded from obtaining direct allocations/assignments if ARIN were to go to a flat-rate structure. I do not see that as being overall beneficial to the community.

> 
>> I don't believe for a second that ARIN is increasing its overhead just to ensure the money gets spent. I have reviewed the ARIN public financials several years in a row and I believe that ARIN does a great deal for the community at a very reasonable cost to said community.
> 
> We've had offers to do it for less.
> 

No, we've had offers to do part of what ARIN does for less. I am not convinced that those offers represented a sustainable way to do it, either.

>>> But we see this all over the world these days, so it isn't like I'm surprised that they would behave this way. At least they're not creating additional address registries and then telling people they need to register their addresses there too in order to protect their trademarks, like some other winners of the how-can-we-charge-for-things-that-were-free game that the Internet played some years ago.
>> This really isn't a fair characterization of either structure. First, I agree that what has happened in the DNS realm has been badly mismanaged and that most of the mismanagement has come about as a result of various profit motives.
> 
> Ok, then we agree on that front.
> 
>>  However, to claim this was strictly a matter of "how can we charge for what was free?" is not at all accurate. The original switch to "charging for what was free" was an effort to sustain the existing infrastructure from a new funding source since the previous government contracts were about to become unfunded.
> 
> I was there. It was a little of each. Lots of people volunteered to do things for free, but instead we had to play by the rules that were set up to make sure that a small number of people were able to sit at this newly-created trough. And here we are. Yes, numbering has worked out a lot better for the public interest than names did, but it still isn't great.
> 

Doing it for free sounds great, but in reality, anyone doing it for free isn't obliged to keep doing it. If you're not taking money, there's no "exchange of value" and legally that means there's no contract. Volunteers can quit any time they want.

I would argue that these functions are too important and must be sustained as continual operations. Therefore, placing it in the hands of an organization offering to do it for free isn't exactly ideal.

I actually think that the numbering situation is pretty good.

>> I think that the number registries have done a vastly better job of the transition and keeping things cost-recovery based while providing the services the community has said it is valuable for them to provide at a very reasonable cost to the community.
> 
> Agreed.
> 

Then what, exactly, are you criticizing about ARIN spending? What money do you think is being wasted or spent improperly? Where do you think this alleged reduction in revenue is feasible?

>> 
>> As much as I think that the new fee structure contains some really poor decisions that will drive poor number resource policy as a consequence, I do think that the board is trying to balance a number of tradeoffs and has made a reasonable attempt at doing so in a fair and consistent manner.
> 
> Yes. But one of those tradeoffs is that they apparently can't figure out how to run ARIN in a world where everyone with a /32 pays the kind of fees that these "extra small" ISPs claim they can afford. That's not a numbering policy problem.

True, but it is a fee structure problem. Unfortunately, if we can't solve the fee structure problem, we're probably faced with the reality of contorting number policy in order to address the reality of the situation. It's not ideal, but it's also not unusual.

For example, it's hard to prove someone was impaired while they were driving. Instead, we substitute characteristic amounts of alcohol in their blood stream as evidence that they must have been impaired.

>>>> I would like to hear about other creative ideas on ways to measure
>>>> size and/or appropriate metrics to be used for fee determination.
>>> Each email or call requires a fee payment above and beyond a tiny fee which everyone pays and which is equal for all. Or, worst case, a tiny fee per database row.
>> The problem with this is that it would again create negative incentives. It will financially incentivize the following negative behaviors:
>> 
>> 	1.	Avoid updating records to reflect changes.
>> 	2.	Downsize customer assignments to avoid having to request additional space.
>> 	3.	Avoid interacting with ARIN until it's absolutely necessary and cannot be avoided
>> 		through other contortions of ones own address space.
>> 
>> I don't thin that's a desirable outcome any more than what is proposed in the adopted fee structure combined with what is proposed in 2013-3.
> 
> Of course not. But "charge the new extra small fee for everyone with a /32, and don't allocate anything smaller than /32" (which would provide enough of a fix for enough of the people right away) we are told isn't compatible with "don't drastically reduce ARIN's expenses", which is also not something we can impact with the numbering policy process.

Right, so we agree that neither of those solutions is viable, now let's look elsewhere for one that is.

Owen