[v6ops] Re: [IPv6]Re: Is the P flag even necessary? M flag already does enough (Re: Re: A detail review of draft-ietf-6man-pio-pflag-04)

[Ted Lemon <[e-mail]>]

Thanks!  I think that’s the test: if you don’t have a theory for what the
exception might be, it doesn’t make sense to have an exception. We can
always update later with a specific exception when we think of one.

There the test would be, do we imagine that a router implementing the old
spec would have some way to follow the newly specified exception other than
user configuration?  Since we don’t know what the exception would be, the
answer is no: the router implementing whatever that exception would turn
out to be would have to be implementing the new spec. So the fact that we
say MUST here isn’t a problem for that hypothetical future implementation.

Op di 6 aug 2024 om 11:34 schreef Timothy Winters <[e-mail]>

> Hi Ted,
>
> I almost made that modification and can't think of a great reason not to.
>
> OLD:
> LPD-8: The IPv6 CE Router SHOULD by default provision IA_PD IA prefixes
> with a prefix-length of 64.
>
> New:
> LPD-8: The IPv6 CE Router MUST provision IA_PD prefixes with a
> prefix-length of 64 unless configured to different prefix-length by the
> user.
>
> ~Tim
>
> On Tue, Aug 6, 2024 at 11:32 AM Ted Lemon <[e-mail]> wrote:
>
>> Yes, much better. Thanks!
>>
>> Why SHOULD and not MUST?  What is the case where they would not do this?
>>
>> Op di 6 aug 2024 om 11:26 schreef Timothy Winters <[e-mail]>
>>
>>> Hi Ted,
>>>
>>> How about this:
>>>
>>> OLD:
>>> LPD-8: The IPv6 CE Router SHOULD by default provision IA_PD IA prefixes
>>> with a prefix-length of 64.
>>>
>>> New:
>>> LPD-8: The IPv6 CE Router SHOULD provision IA_PD prefixes with a
>>> prefix-length of 64 unless configured to different prefix-length by the
>>> user.
>>>
>>> I'll make this change in the next revision.
>>>
>>> ~Tim
>>>
>>> On Tue, Aug 6, 2024 at 10:49 AM Ted Lemon <[e-mail]> wrote:
>>>
>>>> What I’m saying is that the text is ambiguous because you don’t say
>>>> what “by default” means. I am one of the people who wants to get rid of the
>>>> hierarchical model.
>>>>
>>>> Op di 6 aug 2024 om 09:05 schreef Timothy Winters <[e-mail]>
>>>>
>>>>> Hi Ted,
>>>>>
>>>>> On Mon, Aug 5, 2024 at 2:30 PM Ted Lemon <[e-mail]> wrote:
>>>>>
>>>>>> On Mon, Aug 5, 2024 at 11:16 AM Timothy Winters <[e-mail]>
>>>>>> wrote:
>>>>>>
>>>>>>> v6ops has a draft for PD on the LAN to improve this situation.
>>>>>>>
>>>>>>> https://datatracker.ietf.org/doc/draft-ietf-v6ops-cpe-lan-pd/
>>>>>>>
>>>>>>> Please feel free to send comments, we are about to do WGLC on it.
>>>>>>>
>>>>>>
>>>>>> Hey, Tim. I hadn't read the document in a while. I see this text in
>>>>>> the last requirement:
>>>>>>
>>>>>> The IPv6 CE Router SHOULD by default provision IA_PD IA prefixes with
>>>>>> a prefix-length of 64.
>>>>>>
>>>>>> I read this as "if the DHCP client doesn't specify a narrower prefix,
>>>>>> the CE router SHOULD .. 64"
>>>>>>
>>>>>> Is that what you intended? If not, I think you need to say more. If
>>>>>> that is what you intended, this won't work, because if we stack CE routers,
>>>>>> I expect every CE router to ask for a /48, rather than not specifying, and
>>>>>> that would mean that we'd always delegate the narrowest remaining subset of
>>>>>> the outer CE router's delegation to the first inner router that makes a
>>>>>> request.
>>>>>>
>>>>> That's what the working group wanted.  The original version of this
>>>>> document had more text about how to support hierarchical or flat models.
>>>>> After a round or two discussion what came out of that was routers behind a
>>>>> CE Router are no longer a CE Router as they aren't at the customer edge.
>>>>>  The draft reflects that general consensus, that leans towards deploying a
>>>>> flat model as opposed to hierarchical, which is where the /64 length
>>>>> derives from.
>>>>>
>>>>> I think it may be time for another document to specify what to do if
>>>>> you're a Internal Router (but not SNAC).  We could include all the flat
>>>>> model text for becoming a DHCP Relay and giving out IA_PD with /64 from the
>>>>> customer edge.
>>>>>
>>>>>>
>>>>>>
>>>>>