[arin-ppml] Opposed to 2010-9 and 2010-12

Tony Hain alh-ietf at tndh.net
Tue Oct 26 14:29:13 EDT 2010

Milton L Mueller wrote:
>worth pursuing further. 
>>Putting in hard dates to try to sunset an address range will 
>>just lead to endless debate. 
>Agreed, but the escalating fees also require a somewhat arbitrary 
>setting of the escalation rate. Double every year, or every two 
>years? Doubling starts after [three][five][N] years?

Yes the values are arbitrary, but not completely. Just about everyone can
find a way to move infrastructure in 3-5 years, particularly when there is
an obvious financial motivation. Also, no matter what values are picked,
people will game the system to their local advantage. Picking too short a
window will lead smaller players to wait to start any IPv6 deployment until
they can see that they will be able to complete an infrastructure swap
before the price increases significantly. Pick too large a window and the
motivator to move off the transition doesn't kick in for a longer period. At
the same time the price increment can be just about anything. I picked
doubling because that is easy to understand and just about everyone has seen
the math about where that quickly leads. 

The trick here is finding the balance that provides motivation to move off
6rd in the 5-7 year timeframe, without being so threatening that the large
players object up front because there is a real possibility they will end up
at a financial disadvantage after the price increase kicks in because they
have more infrastructure to move. If the initial fee is low enough, they
will not even notice the first couple of doublings, so I think double every
year after 3 is a reasonable middle ground to give the large players a good
5 year window before the price really starts to pick up. 

>>Putting in an explicit cost that is only borne by those who 
>>need the crutch will result in a natural weaning.  
>>As much as I don't want to see 6rd in a different prefix range, 
>>maybe that is the out here. Create a specific fee schedule for a 
>>dedicated use prefix range, which would enable a path forward in 
>>the short term at the same time it motivates providers to get off 
>>of the technology asap. 
>That might work. Herrin was proposing segregation of prefix ranges 
>for other purposes, got some flak but I can't remember exactly what 

I am not overly happy with the idea of 6rd being in a separate prefix from
what the isp would normally use, and I am not overly happy with the idea of
different charges for different address ranges. That said, I don't see any
clean way to deal with sun-setting a transition allocation other than
financial feedback. As you noted, even that path is arbitrary so it is not
all that clean. The advantage it has is the ability for people to build an
internal business case about moving off the transition allocation. Any
ad-hoc attempts at artificial reclamation and/or filtering will be met with
legal action because they will not be universally applied to or by all. As
much as there is an arbitrary nature to an escalating fee structure, if it
is agreed to up front and universally applied there is little ground for
future legal action to leverage.

>Another possibility might simply be to impose a 6rd surcharge that 
>increases. If I am correct in assuming that 6rd users have to 
>declare what they are doing in order to qualify for the larger blocks 
>of addresses, the fee could be triggered by the declaration rather 
>than the use of a specific address range. 

This would address my concern about using an identifiable range, and if
policy evolves to get us past the /32 strangulation mindset, might even
result in most providers keeping their initial allocation rather than having
to migrate between a transitional allocation and a long term one. A /32 is
intended for a startup with 0 customers, so every isp with an existing
customer base should have a good sized allocation to begin with, and there
is no reason to force them to renumber except to reclaim the 6rd oversized
allocation. If they all got a /24 now to support 6rd, then in 3 years had to
justify keeping that based on existing customers with native IPv6 or pay the
increasing 6rd fee, worst case should result in the smallest players
shrinking from a /24 to a /28. As long as they know up front that will be
the case, they can build their long term address plan to fit within the /28
so there is no need to renumber once 6rd is decommissioned. 


More information about the ARIN-PPML mailing list