Re: A common problem with SLAAC in "renumbering" scenarios

[Fernando Gont <[e-mail]>]

On 20/2/19 01:54, Ole Troan wrote:
> 
> 
>> On 20 Feb 2019, at 04:35, Fernando Gont <[e-mail]> wrote:
>>
>>> On 20/2/19 00:11, Ole Troan wrote:
>>>
>>>
>>>>> On 20 Feb 2019, at 03:50, Fernando Gont <[e-mail]> wrote:
>>>>>
>>>>> On 19/2/19 10:08, Ole Troan wrote:
>>>>> Nick,
>>>>>
>>>>>> On 19 Feb 2019, at 13:57, Nick Hilliard <[e-mail]> wrote:
>>>>>>
>>>>>> Ole Troan wrote on 19/02/2019 12:22:
>>>>>>> Indeed. Wonder how these pesky mobile phone operators manage to
>>>>>>> deliver the same telephone number to a user, for years. Across
>>>>>>> different providers and contracts.
>>>>>>> I can’t think this argument is anything but a strawman.
>>>>>>
>>>>>> Ole,
>>>>>>
>>>>>> if recommending static IP addressing is an idea that 6man wants to push, you'll need to reach out to the security and ops areas to get their input on this.  I'm not sure this is an issue that 6man can resolve fully.
>>>>>
>>>>> It’s been the IPv6 addressing model for at least 20 years, so I think the other areas have had ample time to provide their input.
>>>>
>>>> For the reasons stated in draft-gont-6man-slaac-renum, I don't think
>>>> this affects the discussion we are having. But, out of curiousity,
>>>> where's the "addressing model" you are referring to documented?
>>>
>>> I can’t see slaac-renum tackling these issues.Which reasons are you referring to?
>>
>> A significant percentage of IPv6 deployments don't employ static
>> prefixes but dynamic prefixes. That's a deployment reality.
> 
> There’s a defined way to do IPv6 renumbering, which involves overlapping new and old prefixes. 

There's a defined way to do *planned renumbering*, you might argue.
That's not what we are trying to address here -- even when the
improvements we propose end up helping in that area, anyway.



> There seems to be some people arguing that renumbering without an overlap is a supported case in IPv6. It is not. 
> 
> That is although related, not the problem you are looking at. 

There's a variety of reasons for which a prefix may become stale without
explicit signaling. Jen has provided a more thorough list here:
https://tools.ietf.org/html/draft-linkova-6man-default-addr-selection-update-00
.

What triggered our I-D is one of the cases in Jen's list, where a CPE
crashes and reboots. Version -01 of our I-D elaborates on many things
that contribute to the problem.

What we have right now is a system that is not robust. We want to
improve robustness. The outcome of burying our heads into sand will most
likely be people complaining about IPv6, and IPv6 connectivity failing
and hosts falling back back to IPv4 (thanks to HE).

Modulo folks that seem to be in favor of improving things, so far I seem
to have read:

* CPE vendors are shipping crap
* ISPs are clueless in what they build
* Some of us miss some clues about IPv6
* The 37% of network suffering from these issues are just marginal.

I don't think the above is doing us any favours. Actually, it seems a
lot like "problem solving" in
https://images-na.ssl-images-amazon.com/images/I/416cjJcmM-L.jpg . I'd
rather honor the "E" in IETF.

Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492