Re: [v6ops] Implementation Status of PREF64

[David Farmer <[e-mail]>]

On Mon, Oct 11, 2021 at 1:13 AM Lorenzo Colitti <lorenzo=
[e-mail]> wrote:

> On Mon, Oct 11, 2021 at 1:36 PM Owen DeLong <owen=
> [e-mail]> wrote:
>
>> Suppose an OS developer ships a DHCPv6 implementation that requires N
>> addresses and does not enable the IPv6 stack unless it receives all of
>> them. How does that do anything to address statements such as, "my network,
>> my rules" that have been made on this thread? Any operator that wants to
>> assign only one address per device still cannot use such hosts on their
>> network, and can still criticise the implementation for not doing what they
>> want.
>>
>> At least it provides a possible way forward for networks that don’t want
>> to assign addresses outside of IA_NA or IA_PD to provide addressing to
>> android devices.
>>
>> Surely a single IA_NA + a routed IA_PD would meet your requirements,
>> right?
>>
>
> IA_PD, even by itself without IA_NA, has none of the concerns in this
> thread. RFC 7934 explicitly recommends it, actually. Not sure if enterprise
> network operators can or would want to deploy IA_PD though. Is that
> something that we can nudge the industry towards?
>

I think this is a perfectly reasonable option for the general purpose use
case, but I will note there is no clear guidance for the use case of IA_PD
on Hosts, and even the guidance for the use of IA_PD with routers is
somewhat fuzzy or at least incomplete, in my opinion. Nevertheless, this
may or may not be reasonable for special or restricted purpose use cases,
especially when compliance regimes are required, like PCI or NIST-800-171.
I support IA_PA on hosts, but whether or not that is an acceptable
replacement for IA_AN and/or IA_TA for all use cases is very much an open
question, and not entirely a technical question for that matter.

> What I think David was suggesting is: if we reach consensus on some
>> reasonable value of N, then the IETF could write a BCP to that effect. That
>> would help to some extent, because if an operator complains that the
>> implementation does not work on a network that has N=1, the OS developer
>> could point to the BCP as justification for why.
>>
>> I don’t think that’s really necessary. I think there are already adequate
>> documents about single-address is less than ideal in the existing RFCs/BCPs.
>>
>
> The existing BCPs also say that running an IA_NA-only network is not
> recommended, but that doesn't prevent folks from saying it's necessary. I
> think we'd want to make a stronger statement than that.
>

As I said, I'm open to additional BCP, and at a minimum, I think additional
Informational guidance is needed regarding DHCPv6 development.

While that is true for the general purpose use case, there is no guidance
that IA_NA-only is inappropriate for other use cases. And Android is
clearly being used in other use cases. Furthermore, that would seem to mean
a combination of IA_NA and IA_TA is acceptable even for the general purpose
use case.

Thanks.

-- 
===============================================
David Farmer               Email:[e-mail]
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota
2218 University Ave SE        Phone: [phone]
Minneapolis, MN 55414-3029   Cell: 612-812-9952
===============================================