Re: Last Call: RFC 6346 successful: moving to Proposed Standard

[Phillip Hallam-Baker <[e-mail]>]

On Wed, Dec 3, 2014 at 4:29 PM, John Curran <[e-mail]> wrote:

> On Dec 3, 2014, at 9:15 PM, Dave Crocker <[e-mail]> wrote:
> > ...
> > So, after 25 years of effort, we've achieved 5% penetration.  Wow.
>
> I'm not certain that it is appropriate to count the years of protocol
> development, testing, and deployment into operating systems and routers
> as the denominator for the "5% penetration"...  There has not been a
> strong need for IPv6 until there was actual runout of IPv4 free pool,
> and this did not occur in any of the regions until 2011 (and is yet to
> happen in the North American region)   You should either measure service
> provider enablement of IPv6 from IPv4 free pool runout dates, or need to
> consider the IPv6 protocol support that has been achieved in deployed
> devices (enabled or not) over the 25 year period.
>
> Also, characterizing IPv6 success based on one providers success is
> probably not informative... there are service providers with much
> higher enablement -
> http://www.internetsociety.org/deploy360/blog/2014/09/verizon-wireless-hits-56-ipv6-t-mobile-usa-40-att-24/
>
>
Just to clarify, what is the proposal here:

1) Address+Port become one way to manage the depleted address pool
2) Address+Port become the one and only way

I see no problem with 1, it is a sensible proposal. I would see a lot of
problems with 2.


Turning to John's point, the problem for IPv6 deployment is that IPv4
address depletion is not and will never be a reason for people deploying
IPv6. Until IPv6 delivers 100% of the functionality of IPv4, IPv4 behind
NAT will beat IPv6.

The solution for deployment then is to make IPv6 deliver 100% connectivity.
And there are several ways we might go about that. DNS64 for example. But
any such scheme is going to require a break from the pure IP end-to-end
principle because the addresses have to change from IPv4 to IPv6 somewhere
along the path.


It really isn't a major change. It does not even require a gateway to be
statefull. But some people think that it is more important to suppress such
heretical thoughts than to address the deployment problem.