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

[Matthew Kaufman]

On 4/7/2013 4:21 AM, John Curran wrote:
> Matthew -
>   
> The current fee schedule reads as follows:
>
>     IPV6 ANNUAL FEES (NOTE: FEE WAIVERS IN EFFECT), EFFECTIVE UNTIL 30 JUNE 2013
>     Size Category   Fee (US Dollars)	Block Size
>     Small           $2,250       /40 to /32
>     Medium          $4,500       /31 to /30
>
>     There is 25% IPv6 fee waiver in effect (this started out at 100% and
>     was to be phased out over the last 4 years, although it was extended
>     at the 25% discount level for this year to carry us into the Revised
>     Fee schedule.)
>
> So, any ISP using IPv6 under _today's_ fee schedule has a minimum annual
> fee of $1687 per year [$2250 * 75%] and that covers an allocation up to
> /32 of IPv6.  Unless we continue with the Revised Fee schedule, this is
> effectively a minimum annual cost for any ISP making use of IPv6.
>
> If we want to lower that cost of using IPv6 for smaller organizations,
> we need to find some manner to distinguish these smaller ISPs from all
> ISPs, and this is typically done through the total address block holdings.

I believe that way is not appropriate for the post-IPv4 world. No ISP 
should be getting smaller than /32.

If you want to lower the cost of IPv6 for smaller organizations, you're 
free to do so, but tying to how much IPv6 space they have is likely to 
result in a very large number in the smallest category. And whether or 
not that's a problem *isn't a NRPM issue*.


>
> At this point, ARIN can sustain some number of smaller ISPs having lower
> fees, and the Revised Fee schedule supports an ISP with no more than /20
> of IPv4 space and /36 of IPv6 being categorized as "x-small" with annual
> fees of $1000/year.  This category makes IPv6 more approachable for these
> ISPs but it is indeed at the downside of a smaller IPv6 allocation.

That shouldn't be happening. No reason an ISP should be getting a /36 
instead of a /32, just encourages bad behavior.

>    As
> you are aware, /36 of IPv6 space would provide for more than 4000 /48
> and _lots_ of /56 assignments (but there would be less in practice due
> to internal hierarchy in assignment management.)

Yes. So /36 is *barely* what it takes to remind an ISP that they 
shouldn't be treating IPv6 like a scarce commodity. Maybe. But /32 would 
be a lot safer, and so that's what *numbering policy* should be doing.

>
> Draft Policy ARIN-2013-3, combined with the Revised Fee schedule as
> corrected, would continue this approach of allowing very small ISPs
> with no more than /22 of IPv4 to obtain a corresponding IPv6 allocation
> of /40 and have annual fees of $500 year in the "xx-small" category.

The draft policy is bad numbering policy, and the fee schedule will 
provide an incentive for organizations to use the provisions of that bad 
numbering policy. Don't support.

> The downside that you assert with Draft Policy ARIN-2013-3 is that
> "this type of financial pressure towards false conservation is going
> to give us things like/64-per-household instead of something sensible
> that lets the thermostat be on a different subnet than the Xbox."
>
> Given that any ISP qualifying as xx-small (even with wildly aggressive
> NAT) has no more 1024 customers each with a single IPv4 address, and the
> /40 IPv6 allocation would provide them with the ability to make 65K /56
> assignments to these same customers, it does seem somewhat strange that
> "false conservation" of those 65 thousand potential assignments would
> drive xx-small ISPs to instead make /64 IPv6 customer assignments.

Lets ignore IPv4 entirely for a moment... or pretend that in the near 
future people might single-home IPv4 with PA addresses but want their 
IPv6 to be PI space... In such a situation, organizations will choose to 
pack themselves and their customers into too little IPv6 space just to 
save a buck. I don't think that should be happening.

>
> The community can indicate that it does not support ARIN-2013-3 if the
> resulting "false conservation" is a problem, and then there will be no use
> of the xx-small/$500/yr fee category.

The community on *this* mailing list should be debating what is best 
numbering policy, and not considering the tempting fruit of this week's 
fee schedule (subject to change without community input at any time).

>   Under the Revised Fee schedule, these
> ISPs would be paying at least $1000/year based on a /36 IPv6 allocation
> (which is still better than today's fees with IPv6 use as noted above.)
>
> It would be good to hear from ISPs who would qualify for the xx-small
> $500/year category about the resulting temptation that it poses for
> making smaller IPv6 customer assignments (and how they feel safer with
> the /36 IPv6 minimum and corresponding $1000/year annual fee), as they
> are the ones who are most affected by the outcome of this draft policy
> consideration.

It would be good to hear from small ISPs that are deploying IPv6 at all, 
but that's a whole separate problem.

Matthew Kaufman