IPv6 mistakes, was: Re: Looking for an IPv6 naysayer...

Fri Feb 18 01:39:32 UTC 2011

On Feb 17, 2011, at 4:52 PM, Mark Andrews wrote:

> 
> In message <5F90644C-5457-460F-9BC3-70802B13A270 at delong.com>, Owen DeLong write
> s:
>>> 
>>>> Cisco is just one example.  The fact is it will likely not work in
>>>> cell phones, home gateways, windows PCs, Mac's, ....  I understand
>>>> some progress has been made... but choose your scope wisely and pick
>>>> your battles and know that the weight of the world is against you
>>>> (cisco and msft)
>>>> 
>>> 
>>> I don't think I had general usage in mind, more along the lines of the
>>> "middle 4" in NAT444 that will be rolled out in many networks to
>>> conserve IP space.
>>> 
>> Infeasible. NAT444 is primarily needed to avoid doing a CPE forklift
>> for nearly every subscriber. To deploy these addresses in that space would
>> require a CPE forklift for nearly every subscriber.
> 
> Firstly it is entirely possible to do this incrementally.  Secondly
> it doesn't require a fork lift upgrade.  A minimal upgrade is all
> that is required.  For modern Linux boxes just setting a DHCP option
> would be enough.  A two line fix in a config file.
> 
Whether you do it incrementally or not, you have to upgrade every affected
device eventually. You can roll out IPv6 incrementally, too.

Most CPE is _NOT_ within the description of "modern linux boxes" so
does not apply to the discussion of the middle 4 in NAT444.

It may not require an actual forklift upgrade, but, in the real world, it will
require ISP efforts that are equivalent to a forklift upgrade, so, if you're
going to that much trouble, it's cheaper (and in many cases easier)
to go ahead and forklift your way to IPv6.

Ideally in the next round of CPE, the need for NAT444 is a non-issue.
It should support at least DS-Lite or 6rd.

Anything earlier than the next round of equipment will need to be
at least re-flashed.

Owen