[arin-ppml] Policy Proposal 122: Reserved Pool for FuturePolicy Development

[Owen DeLong]

On Nov 23, 2010, at 11:26 AM, George Bonser wrote:

>> You are pitting a pair of choices against each other as if they are
> the
>> only two options, when the reality is there will be a continium of
>> solutions in the middle.
> 
> I also believe to be true.
> 
Sure, there is a continuum of options available, but, I was describing
the likely consequences of the two choices proposed in order to
compare and contrast them.

>> ISP's will come back for more IPv4 space and get told no.  I would
> hope
>> a savvy ISP could set aside a /28 of their previous allocation to use
>> for transition technologies at that point. 
> 
> The savvy should *already* have some space set aside for that.  They
> know it is coming.  Having this space set aside at the RIR level simply
> enables them not setting aside of this space now at the LIR level.  
> 
What about operators that do not exist today?

What about operators that grow beyond their current ability to set aside
space?

> The savvy operators should already be handing new customers at least a
> /48 of v6 space for customers in PA space and a smaller chunk of v4 for
> use by 6to4 and such transition technologies, or maybe providing 6 to 4
> themselves if they are big enough.  The links to the customers should be
> v6 where possible and with v4 tunneled over v6.
> 
6to4 is utterly and completely useless if you have native IPv6. The purpose
of the IPv4 is addresses in 4.10 is to enable technologies at the SP level
such as NAT64 and DS-LITE. Technologies which allow native IPv6
customers to reach content and services only available over IPv4
and which have a high leverage ratio of subscribers:address.

In contrast, 6to4 is a way to allow a machine which does not have IPv6
connectivity to reach IPv6 content by using a dynamic stateless tunnel.

> The fact that many operators that consider themselves to be major
> operations are still not native v6 is going to make all of this moot, I
> am afraid.  They are going to hit the wall when runout happens and then
> it is going to be a mad scramble with a lot of operators all having
> networks in various states of misconfiguration at the same time.
> 
Hence my belief that with such a large number in this situation, prudent
stewardship of the address space requires us to use the collective
wisdom of the community to back-stop this behavior for the good of
the internet. As much as I generally oppose rewarding bad behavior,
this is a case where the collateral damage exceeds the direct
results of the bad behavior.

> For customers with v4 PI space, they should be assisting those customers
> in getting their v6 assignments NOW.
> 
Yep.

>> I pick on that size only
>> because it's a popular suggestion of how much space we should give the
>> ISP from the /10.  The more we can encourage folks to make these sorts
>> of wise decisions, the longer the /10 will prove useful.
> 
> Necessity is the mother of invention according to the old saw.  Many
> operations work on a "it isn't a problem until it is a problem" basis.
> They aren't going to change a thing until they have to.  Many (most?)
> already have the resources to start in that direction now and not wait
> for runout.  Building in more cushion simply delays things a bit.
> Having the spectre of "no space available" hanging before them might be
> enough to spur them to set aside some of their own space NOW and begin
> moving.
> 
That's a very dangerous game of chicken because it puts us on a
road where we can't flinch if the other guy doesn't. The old saw also
says "hope for the best, plan for the worst."


>> Even if the situation you describe comes to be, where folks are
>> monumentally stupid and give out every last IP of their allocations
>> keeping nothing to run a transition box your dire consequences still
>> come true.  Giving a provider who receives a /28 or two from the last
>> /10 will not change their competitiveness against another provider who
>> still has half a /8 available.
> 
> I predict many are going to give out every last bit of their space
> judging from the lack of progress so far.
> 
Yep.

>> The last /10 isn't about making a business with no IPv4 able to
> compete
>> against someone who still has half a /8 left.  That is frankly,
>> laughable.  What it can be about though is help for those who planned
>> well and ran into unexpected consequences.
> 
It was not my intention to say that it was. However, there is a difference
between the way this happens as a natural result of runout and having
ARIN actually participate in creating or exacerbating the situation.

> Or new operations just getting started who get a large v6 block and need
> a small v4 block in order to talk to the rest of the internet.  Where
> this *really* puts the damper on things is for a new operator just
> getting started and I believe there are a lot of operators out there who
> are going to want to hold onto v4 as long as possible after runout so it
> serves as a barrier of entry to competition.  Imaging trying to start up
> a new operation after v4 runout.
> 
Yes, this is much more in line with the intent of 4.10, although it is also
meant to backstop those organizations that run out and need small
quantities of IPv4 addresses for high subscriber leverage transitional
purposes.

If you want a list of the ponies that people will try to demand when we
open this Pandora's box, you need look no further than the version
of 2010-13 presented in Atlanta compared to the version that originally
went to the petition process.

>> We need to get out of the mindset that if someone has totally ignored
>> IPv6 transition we have something good waiting for them in the wings.
>> We don't.
> 
> Absolutely agree.

I couldn't agree more. If one assumes I have such a mindset, one has
not been paying attention for some time.

Owen