[arin-discuss] IPv6 as justification for IPv4?

[Dmitry Kohmanyuk]

Folks,

here are my few comments on a new hot topic for this spring :)

Disclaimer: I am also on NRO NC as representative from RIPE region, so I know a bit about RIPE's policies and can compare some.

On Apr 16, 2013, at 10:39 AM, Jesse D. Geddis <jesse at la-broadband.com> wrote:

> On Apr 16, 2013, at 12:07 AM, "Randy Carpenter" <rcarpen at network1.net> wrote:
> 
> Randy, I asked what would it cost if we used x baseline that I pulled out of a hat to start somewhere (as I've mentioned repeatedly). You're looking at my comment too narrowly. At the root of it was remove the ceiling. Again you're going down the ARIN cost model and based on what I've seen from folks on this list is that goes nowhere... slowly.
> 
> For me, I don't care what x org costs ARIN
> I don't care what it takes to maintain x orgs records.
> I don't believe ARIN has this data anyway. 
> 
> I care about these fees as they relate to:
> 1. Getting IPv6 widely adopted
> 2. Getting ARIN out of the way of starting a business.
> 3. Making this fee model more equitable.
> 4. Getting rid of the cut off at /14's and larger.

While we can debate here whether there should be cost dependent on total size of IP address blocks allocated ("tax the rich"),
or should tiers be narrower ("16 ranks of IP consumption"), I would also like to bring attention to new RIPE policy which just became
effective after last year General Meeting - 

RIPE NCC Members (LIRs, equivalent of ISPs) used to pay a fee depending on their tier, calculated using a formula doing a weighted
sum of allocated address blocks depending on their age ("charging scheme") used as a scope and then sorting all members by that score,
and then dividing them by categories so each class had certain percentage of total members.  Thus, fees were not exactly predictable
as they depended not just on a member paying them, but on total distribution of membership allocations and their ages ("seniority".)

Here is current schema, and historical data for reference (5 fee bands):
http://www.ripe.net/ripe/docs/ripe-566

So, RIPE members voted for flat fee structure, where each member pays exactly same amount of money, plus fixed fee per each PI object.
AS numbers are excluded from fee calculation.  So, one obtains addresses on need, and money is not a concern at all.

Right now, ARIN members vote the same but pay depending on their allocations.  New RIPE approach essentially aligns vote and payment.

On another end of spectrum ("market power") would be not just make ARIN members to pay in strict proportion to allocated addresses,
but also vote with them.  Would this option appeal to proponents of market approach ("IPv4 as commodity")?   (I am not in this camp, for sure.)

> What you suggested is going by aggregate. That's virtually identical as saying going by /x arbitrary block size. It's a linear model. That's where it's similar. I think going by aggregate is an interesting suggestion. 
> 
> […]

> It's extremely important because there is a macroeconomic impact of the current nibble policy as well as the fee structure. ARIN policy effectively created Matthew's position in the same way as Sarbanes Oxley (policy) created an entire cottage industry of document retention. Matthew is my example because he held his organisation up as one. 

So, speaking of nibble policy (I am not exactly sure what is it) - which bad effects does it have, in your opinion?

>> What you are talking about here is a serious issue. I agree wholeheartedly that ARIN needs to fix the chicken/egg problem that currently exists with new entities. However, that has absolutely nothing to do with your argument above. If a organization is being denied IPs from their upstream provider, it means that the upstream provider is dumb. They either are not properly managing their IP space, or are simply refusing because they feel like it. Do you think that changing the upstream providers fees from $32,000/year to $1,000,000/year is going to change that?
> 
> If that cost per pegged to something linear for the provider that is a fixed cost and would more likely become one for their end users. Right now for a provider like AT&T that cost is totally arbitrary. They have multiple /8's but they charge over $1,000 a year for a few static IPs that they aren't even paying a penny for. And like many providers their size they are a decade behind many of us on this list in IPv6 deployment. 

So, if we make everybody to pay same amount of $ for IP address regardless size of allocation, it would discourage address resale
and hoarding, at a cost of increasing fees of "big, bad" ISPs by factor of thousands.  Is that a good way to solve a problem? (Methinks, no.)

My suggestion would be to make it easier to obtain initial allocation (and perhaps make it smaller - say, /23 - and perhaps remove restriction on
multihoming - allowing new ISPs with no immediate need of /22 to get started on their own IP space.

Tying IPv4 requests to existing customer base of IPv6 users is interesting idea, but it may be different to quantify in policy.

-- dk@