Re: [6lowpan] link-local addresses

Zach Shelby a écrit :
> On Nov 12, 2009, at 16:55 , Alexandru Petrescu wrote:
>> Ulrich Herberg a écrit :
>>> Alex,
>>> On Thu, Nov 12, 2009 at 7:49 PM, Alexandru Petrescu
>>> <alexandru.petrescu@gmail.com> wrote:
>>>> [...]
>>>>
>>>> Er?
>>>>
>>>> I can see here two disagreements:
>>>>
>>>> (1) AUTOCONF document draft-ietf-autoconf-adhoc-addr-model-00.txt is 
>>>> not
>>>>   such a big consensus you make it sound like.  To me there seem to
>>>>   be disagreements on at least one issue: link-local addresses.  Which
>>>>   brings point 2 below.
>>> I have to disagree. While you personally may not like this document,
>>> the AUTOCONF WG chairs have sensed a consensus of the WG on that
>>> document. Also, some issues about the draft that have been raised
>>> (notably the text about LLs), have been solved during the AUTOCONF
>>> meeting by proposing modifications of the text. The minutes of
>>> AUTOCONF will certainly be released soon. That said, the discussion
>>> about consensus of an AUTOCONF draft is probably of no concern for
>>> 6lowpan, so this will be my only response to that topic on this
>>> mailing list.
>>
>> Ulrich, ok.
>>
>> What do you think about the 2nd point?  (the fact that 
>> draft-ietf-autoconf-adhoc-addr-model-00.txt talks differently about 
>> LLs than draft-ietf-6lowpan-nd-07.txt does)
> 
> Keep in mind that -07 uses its own definitions and model, which have 
> little chance of living on their own. So don't compare terminology from 
> -07 with the autoconf model. In practice, the autoconf model does not 
> change how our solution works. The design team is now starting to look 
> at that, so please be patient for -08.
> 
> A couple clarifications:
> 
> 1. We will initially only copy the autoconf model for our purposes. If 
> it goes forward to an RFC rather quickly, we may theb make a nomative 
> reference at some point. I at least strongly support that the autoconf 
> model work is moved forward as quickly as possible.
> 
> 2. The autoconf model very rightly points out that link-local scope (and 
> thus addresses) are of limited use.

YEs - the limit is that link.  No other limit I can imagine.

Especially here, where 802.15.4 is cited upfront.

> But it does not forbid them.

Ok.

> We have 
> exactly the same setup in 6lowpan-nd as well.

I am not sure. Here you do cite 802.15.4 whereas in AUTOCONF there's no 
link suggested.

> Let's not speculate too much on the autoconf model until we get -08 of 
> our draft out - and then you will see how the pieces fall together. 
> Based on my initial analysis it is a nice fit, but the devil is in the 
> details still. In practice we came to the same conclusions in this WG, 
> but used slightly different terminology (or it turns out we defined 
> things like subnets and links when we didn't need to).

Well - you're questioning here prior agreements.  We did seem to be in 
agreement when defining links in 6lowpan, as of April 2009, see attached 
email.

Now you do not like that anymore.

Why?

I think that links and subnets should be defined.  If AUTOCONF doesn't - 
that's an AUTOCONF issue.  That error shouldn't be done here too.

Alex

> 
> Zach
> 
>>
>> Alex
>>
>>
>> _______________________________________________
>> 6lowpan mailing list
>> 6lowpan@ietf.org
>> https://www.ietf.org/mailman/listinfo/6lowpan
> 