[arin-discuss] IPv6 as justification for IPv4?

Owen DeLong owen at delong.com
Wed Apr 17 22:50:47 EDT 2013

On Apr 17, 2013, at 10:21 AM, "Jesse D. Geddis" <jesse at la-broadband.com> wrote:

> Owen,
> Whether its framed as 'paying for IP addresses' or 'paying for registration services' is really irrelevant. The fact is is that it is tiered up to a certain point until you get different rules from everyone else at x-large. Further, those tiers are pegged to an _allocation_size_. So applying a fee scale to IPs is something _ARIN_ did by implementing this tiered structure in the first place. Not us.

So what this really boils down to is that you're upset that ARIN doesn't extend the pricing model beyond XX-L
(look at the adopted fee structure, not the current one) because a very small number of organizations at the top
don't get charged even more for their relative sizes within that category.

Frankly, the number of organizations at that tier is so small that extending the linear pricing model wouldn't make
a significant dent in the pricing at any of the lower tiers, so it's not, IMHO, worth complicating the pricing structure
to accommodate that.

> Regarding your mention of ARINs costs associated with different tiers and the assertion that x-large costs ARIN less money. No one has put forth any hard data whatsoever to support that argument. Its an unfounded assertion that's been given mythological status by you and some other folks. Indeed, even John Curran himself mentioned we can look at other models including pegging the fees to actual ARIN costs implying that they aren't right now. Based on these things I don't find this to be a valid excuse to give x-large much lower costs proportionately and to allow them to scale exponentially with no concern of fees or having to proportionately 'support' like everyone else.

They aren't pegged directly to costs because we aren't doing Fee for Service billing. I don't think you want that.

However, ARIN has shown the over all cost to ARIN per organization size based on the current fee categories and
has shown that the larger organizations do cost ARIN less per IP address to work with. There is hard data to support
that fact. Further, simple logic from what ARIN actually does makes it pretty obvious that the costs to ARIN do not
have any relationship to the size of address holdings of the various organizations ARIN deals with. It has more
to do with the number of transactions, quality of the information provided by the organization in each transaction,
and number of records held in the database. There's a little more staff time to review the greater amount of data associated with an application for a /12 than for a /22, but it certainly isn't 1024 times as much time.

X-Large pays double what Large pays. XXL pays double what XL pays. 

This is the same as XS paying double what XXS pays, Small paying double XS, Medium paying double Small,
Large paying double Medium, etc. At every tier your fee doubles and your amount of space quadruples.

So by your argument, at every tier you pay proportionately less per IP address. This reflects the fact that ARIN's costs do not scale linearly with address size, but attempts to provide a rough approximation of mapping sizes to actual costs.

Yes, the fee structure tops out at XXL. Once you reach a certain size and are paying $32,000/year, you don't have to pay more even as you get more addresses.

In reality, extending that pricing linearly beyond XXL wouldn't change pricing at the lower tiers by much. Further, it is very unlikely that those organizations are actually creating costs for ARIN that would come even close to doing so.

Let's assume, for a moment, that an ISP existed that held </4, ≥/6. By your argument, said ISP should, instead of $32,000/year, pay $256,000/year instead of $32,000/year. To the best of my knowledge, there is no such ISP and
there are probably fewer than 5 ISPs in the </6, ≥/8 category at $128,000/year, so your maximum additional yield
there is $480,000/year. Of the remaining 48 organizations in the XX-L category, I have no idea where the split would
fall between the $64,000 bracket you would establish at </8, ≥/10  vs. the existing $32,000 bracket. My best guess
would be a ~50/50 split, so let's say 24 organizations.

So, you would increase costs for top-end organizations as follows:
	5	* 96,000
	24	* 32,000

If we were to spread that evenly across the X-S, S, and M registrants (total 3818->3306 organizations), you would save each of those organizations less than $400 per year.

I simply don't buy that it's somehow more fair to inflict 64k and 128k/year pricing on to a small number of organizations at the top end to subsidize $400 discounts to 3300 other organizations.

> What I'm interested in hearing, Owen, and others. As well as, I'm guessing the poster you replied to, several other people who have also publicly questioned the hybrid scale, and the 3 dozen people on this list who emailed me privately but weren't interested in the public flogging by some you folks (Matthew, Lee, and a couple others) is a reasoned answer to the following:
> Why *should* there be a fee increasing cutoff at /14

Because beyond a certain point, continuing to double people's fees stops making sense. The size model is designed to be a convenient approximation of the cost-model that ARIN has along with some other tradeoffs. It attempts to allocate the burden along similar lines as ARIN's cost burden.

If you extend it beyond the /12 point (which is where the XX-L category actually begins), then it rapidly makes no sense and is overly burdensome as described above.

> Why should x-large (73 orgs out of thousands) get a different set of rules than everyone else.

They don't. They live by the same rules. There are a number of services with similar cost structures where you pay less for each incremental level of service until you reach a certain tier where you simply get "all you can eat" service for that price.

Some of them are even in this industry. It's not uncommon for mobile providers, for example, to have usage sensitive pricing where you pay less per GB transferred as you move up the usage stack until you finally reach a point where the next increment gets you unlimited consumption.

> And please, let's stick to the facts and avoid repeated assertions like the below that have no basis in actual data. Personally, I think trying to divine ARINs cost per tier and creating a fee structure based on that is a very bad idea. Indeed, my proposal is fundamental in getting rid of the current tier structure altogether. 

I wasn't divining ARIN's cost per tier… It was published data from ARIN.

It is documented as one of the inputs to the current fee structure.

Claiming it has no basis in facts when the ARIN CEO has contradicted you on this matter is folly.

Your proposal is fundamental in getting rid of the tier structure. It's also a really bad idea IMHO. Your proposal is grossly unfair at both the top and bottom extremes.


More information about the ARIN-discuss mailing list