Re: New Version Notification for draft-chakrabarti-nordmark-6man-efficient-nd-05.txt

[Erik Nordmark <[e-mail]>]

Jinmei,

thanks for reviewing the updated draft.

On 2/27/14 7:28 PM, 神明達哉 wrote:
> At Fri, 14 Feb 2014 23:03:56 -0800,
> Erik Nordmark <[e-mail]> wrote:
>
>> Samita has already outlined some of the changes we have been making
>> based on the feedback from the list.
> Was this point addressed in 05?
>
>>> Clarify bootstrapping and DAD for link-local or not How to inform duplicate address detection during link-local address formation
Not very well it seems. We discovered shortly before the draft cutoff 
that the 6lowpan-nd approach to handling link-locals doesn't make any 
sense in this broader applicability.
Hence we added how to handle link-locals to the open issues section.

But we seemed to not have edited the rest of the draft correctly around 
DAD, including the bootstrapping section.
As the draft currently stands DAD for link-locals is (or "can be" given 
the lack of editorial clarity) replaced by NS/ARO. But address 
resolution for link-local targets needs to be done for using multicast NS.

One of the discussion on list is how to avoid the latter, with 
suggestions to add the ability to have RAs state that link-locals should 
be sent to routers. That might make sense in some deployments (when 
hosts do not depend on protocols with have a ttl=255 check).

    Erik

>>> Section 11 talks about boot strapping and the formation of a
>>> link-local address, but it doesn't talk about DAD for the LL address.
>>>
>>> Section 5 has this bullet:
>>>
>>>     o  Address Resolution and DAD uses the registered addresses instead
>>>        of multicast Neighbor Solicitation messages for non-link-local
>>>        IPv6 addresses.
>>>
>>> which seems to suggest DAD is used for link-local addresses, but it's
>>> not really clear (and this is not an addition to 05).
>>>
>>> Assuming the intent is to perform DAD for the link-local address, I
>>> suggest Section 11 explains that explicitly.  It would also be better
>>> if that point is clarified in Section 8.  And, this point of Section 2
>>> may also have to be updated:
>>>
>>>        Remove the use of multicast for DAD and Address Resolution (no
>>>        multicast NS messages), and remove periodic multicast RAs.  Some
>>>
>>> so it's clear that DAD is still needed for link-local addresses.
>>>
>>> --
>>> JINMEI, Tatuya
>>>