Re: [v6ops] I-D Action: draft-jjmb-v6ops-unique-ipv6-prefix-per-host-00.txt

[Brian E Carpenter <[e-mail]>]

Ole, Fred,

On 30/10/2015 11:03, Templin, Fred L wrote:
> Hi Brian,
> 
>> -----Original Message-----
>> From: [e-mail] [mailto:[e-mail]]
>> Sent: Thursday, October 29, 2015 2:59 PM
>> To: Brian E Carpenter
>> Cc: Templin, Fred L; Tore Anderson; [e-mail]; 神明達哉
>> Subject: Re: [v6ops] I-D Action: draft-jjmb-v6ops-unique-ipv6-prefix-per-host-00.txt
>>
>> Brian,
>>
>>>> Hi, I won't disagree with Brian and Tore but on some IPv6 links waiting to
>>>> receive an RA may be redundant and unnecessary before directly
>>>> attempting DHCPv6 PD. On those links, DHCPv6 messaging alone may
>>>> be all that is necessary, and no RS/RA is required.
>>>
>>> How is an IPv6 stack fresh from the factory supposed to guess that?
>>> Surely we don't want to add configuration knobs?
>>>
>>> I think RA is a required feature. RFC 6434 says:
>>>
>>> 7.2.2.  Use of Router Advertisements in Managed Environments
>>>
>>>   Nodes using the Dynamic Host Configuration Protocol for IPv6 (DHCPv6)
>>>   are expected to determine their default router information and on-
>>>   link prefix information from received Router Advertisements.
>>> ...
>>> 12.2.  Neighbor Discovery for IPv6 - RFC 4861
>>>
>>>   Sending Router Advertisements and processing Router Solicitations
>>>   MUST be supported.
>>
>> in the general case:
>> it isn't required for router to router links.

That may be so in practice, but is it actually documented that way?
(I agree that it makes sense.)

>> nor is it required for DHCPv6 PD.

I don't see any wiggle room in the section 12.2 quote above. But anyway,
the only practical interpretation of that is "if you think you might
like to get an IA_PD, then try DHCPv6 regardless of RA."

>>
>> in the 7084 case:
>> we've tied the CE router's default routing to ND Router Discovery.

Right... which is a safe choice, because it's consistent with RFC 6434.

> 
> The same as what Ole said. Also, an IPv6-over-(foo) document can specify that
> DHCPv6-only autoconfiguration is used on that link. 

Really? RFC 4861 does allow exceptions, but mainly in terms of NBMA
links. Also, DHCPv6 is only a SHOULD in node requirements.

> In that way, IPv6 nodes
> that connect to a (foo) link already known in advance what sorts of link
> ND/DHCPv6 services they can expect.

Well, the IPv6-over-foo modules can know in advance. But host stacks as a whole
need to figure this out dynamically, I think. Again, I am assuming that we want
to get avoid configuration knobs.

Regards
   Brian