[address-policy-wg] 2017-03 New Policy Proposal (Reducing Initial IPv4 Allocation, aiming to preserve a minimum of IPv4 space)

Hi,

> On 22 Sep 2017, at 13:56, Anna Wilson <anna.wilson at heanet.ie> wrote:
> 
> Hi Ray,
> 
>> On 22 Sep 2017, at 12:04, Jetten Raymond <raymond.jetten at elisa.fi> wrote:
>> 
>> Hi Anna,
>>  
>> I saw some calculations that with the current policy it would be 4-5 years, to run completely out, last /8 and returned space.
>>  
>> Now I don’t know if these calculations are correct, but even if they are,  or not, then I would like to know how long it should last ? 10 years, 20 , 50? 
>>  
>> I can see and understand your points, the original /8 proposal was not meant to delay v6, fully agree, but by spreading it now it will be expected to be spread out again in say 2-4 years (?) .
>>  
>> I seriously think that the more time we get,  or give the illusion that we can then rearrange it again, the more time people will ignore the fact that it will run out, regardless if we change the policy or not.
> 
> I believe you are correct that people will ignore runout, regardless of whether we change the policy or not.
> 
> My concern is that the problems of ignoring runout fall disproportionately not on existing holders who do the ignoring (who have a good chance of being able to rustle up a small amount of their existing space somewhere to run their IPv6 transition equipment) but on future new entrants (who would have no existing space to shuffle.)
> 
> That's an externalised cost, and it is the very same externalised cost that I believe the original last /8 policy was intended to address. This proposal is the best way I can think to reduce that burden on new entrants.
> 
> So to answer your first question: how long should it last? My only answer to that can be "for as long as new entrants need some IPv4 in order to use IPv6; or as long as possible if we can't get that far." We land on /24 because we think it's practical today.

But it’s not really that they "need some IPv4 in order to use IPv6”.  If they’re making content available, you need IPv4 whether you’re deploying dual-stack with IPv6 or not. If they’re deploying an IPv6-only access network, and using NAT64/DNS64/464XLAT to access legacy IPv4 content, then they’ll need less IPv4 if using IPv6, because a certain (growing) percentage of traffic will be native IPv6 and not NAT64’d (I see 30%-50% quoted in different scenarios, due to FB, Netflix, Google, etc).  The more IPv6-cabable content out there, the less need for IPv4 addresses for a NAT64 service.

So I think they really “need some IPv4 in order to access other people’s IPv4-only content". That need may drop if more NAT64 (or encapsulating equivalents) is deployed, or more content is directly IPv6-enabled or hosted on IPv6-cpable CDNs like Cloudflare, Akamai, etc.

There is some use of NAT46 (SIIT DC etc), but that seems pretty rare, at least currently.

>> Therefore I still think the current policy is sufficient.
> 
> Forgive me if I misunderstand your thinking; I believe it's this: that full runout is the remaining tool we have to get existing IPv4 holders to deploy IPv6, so we should not take further actions to delay it.
> 
> It's not an unreasonable effect to hope for. But the current /8 policy is already quite restrictive. I would be surprised if full runout would have a much greater effect on existing IPv4 holders. 

Perhaps holders might sell off some space if there’s complete RIR runout, the IPv4 price rises, and the market is the only option. So there might be a feedback loop which would generate more supply?

> And even if that effect is something above negligible, the burden of it falls disproportionately on post-runout new entrants.
> 
> So I think that's who we need to help, and why a policy change is needed.

The aim of the proposal is very well-intentioned. I’m just not convinced it will make any difference by 2021/22 when the current run-out for our region is projected.  Things will have moved on significantly by then, just like they have for IPv6 between 2012 and 2017. There was effectively no IPv6 deployment 5 years ago.

There’s an argument to track and follow policies implemented elsewhere, and keep in step with those. LACNIC has adopted /24 it seems, and ARIN have a /10 of IPv4 from which they can hand out /28 to /24. what are the current APNIC or AFRINIC policies?

Tim

> Best regards,
> Anna
> 
> -- 
> Anna Wilson
> Service Desk Manager
> HEAnet CLG, Ireland’s National Education and Research Network
> 1st Floor, 5 George’s Dock, IFSC, Dublin D01 X8N7, Ireland
> +353 (0)1 6609040   anna.wilson at heanet.ie    www.heanet.ie
> Registered in Ireland, No. 275301.        CRA No. 20036270
>