Re: Section 2.4 of rfc4291bis

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

On 05/08/2017 08:34, 神明達哉 wrote:
> At Sat, 5 Aug 2017 05:19:59 +0900,
> DY Kim <[e-mail]> wrote:
> 
>> OK. Let’s pin down two points:
>>
>>   o SLAAC is agnostic to the length of IID.
>>   o Explicit definition on the length of IID are with other standard track documents.
> 
> You can't ignore this:
> 
>    o RFC4291 and rfc4291bis also say the length of IID is 64 bits for
>    some set of IPv6 addresses.
> 
> Reconciling the 2nd and 3rd points would be quite tricky, and that's
> one main reason why I don't think "what's the primary definition" is
> so obvious.  So...
> 
>> Then, why not simply state as such in rfc4291bis like, e.g.:
>>
>>  “The length of the interface identifier is defined in link-layer-specific standard track documents.” PERIOD,
> 
> ...it can't be that simple anyway.  In addition, (in my understanding)
> because it's too substantial a change from RFC4291.  rfc4291bis is
> basically an attempt of elevating the original RFC to an Internet
> Standard, so, in general, we can only make clarifications or minor
> changes.

Correct. That's why I most recently suggested:

"Interface Identifiers are 64 bits long when used for Stateless
Address Autoconfiguration (SLAAC) [RFC4862]."

That's a statement of fact. IMNSHO anything more than that is basically
noise, in terms of rough consensus and running code.
 
> From a purely technical point of view, such a change can be a subject
> of discussion.  IIRC some people have actually proposed ideas like
> that.  But this won't be in the scope of rfc4291bis.

Correct.

    Brian