Re: [sig-policy] prop-103-v001: A Final IP Address Policy Proposal
On Jul 9, 2012, at 3:10 PM, Jay Daley wrote:
>
> On 10/07/2012, at 9:52 AM, Owen DeLong wrote:
>
>> IPv6 requires thinking quite a bit differently. In IPv4 we weren't giving a /16 to every end-site. In IPv6, we are.
>>
>> It's not about number of devices at the end. It's about number of subnets and about having enough bits to plan for automating hierarchical topologies at the end-site.
>>
>> It's about having better capabilities than we had in IPv4, not merely expanding the IPv4 paradigm and its foibles and limitations into a wider bitfield.
>>
>> If you inflict a limitation on your subscribers that prevents them from taking advantage of innovations and improvements in technology in the future, or, worse, prevents those innovations from taking place, then I'm not sure what better term to apply than victimization.
>
> We're already keeping the last 64 bits for thinking differently. Just how many bits does this thinking differently need?
>
The last 64 is for host addressing within the subnet.
You need some bits for thinking differently on the left side of the subnet boundary too.
> And if automated hierachical topologies don't happen? My point is that we should allocate according to known need and not aspirations. If we do ever get to the stage of needing automated hierarchical topologies then we can reallocate, it's not hard.
You're creating an unnecessary chicken and egg negative feedback loop. The research and development on automated hierarchical topologies is already in progress, but vendors won't productize it if ISPs won't support it. Just like vendors don't provide solutions for IPv4 that don't assume everything on a flat network with NAT in the way today. Why? Because very few households have subnetted networks or the ability to do so without NAT.
As a victim of this assumption in today's products in IPv4, I can assure you that it creates unnecessary limitations and is counterproductive to innovation.
> The argument often given by the 'allocate a /48' advocates is that we are only using a small part of IPv6 and can always use a future tranche in a different way. I would make the same argument when it comes to allocating a /56 to a household - if it's not enough then we can always change.
Meh... It's a valid point... There is a safety valve in case we're wrong. So what? The real issue is that there is a very real reason present today to allocate as designed. Most of those bits were put into the protocol for that purpose.
> History has been quite clear that underallocating any resource and then giving more is a far, far easier path than overallocating and then taking back.
History has also been quite clear that creating artificial scarcity has its own costs.
Underallocating IPv6 is an example of such a creation of artificial scarcity.
There are two ways one can waste IPv6 addresses...
One way is to delegate them to end users who don't actually use them.
The other is to ensure that there are vast amounts of IPv6 still on the shelves at IANA and/or IETF not even delegated to IANA when IPv6 is deprecated as a protocol at some time likely after I'm dead and gone.
Ideally, we have enough bits in IPv6 to take advantage of both forms of waste.
However, failing to make good use of the first form to support innovation and flexibility at the residential level is a really bad idea.
Owen