Re: default-iids: Clarifying the use of temporary addresses

[Fernando Gont <[e-mail]>]

On 05/21/2016 01:41 AM, Brian E Carpenter wrote:
> On 21/05/2016 15:29, Mark Smith wrote:
>> On 21 May 2016 at 11:47, Lorenzo Colitti <[e-mail]> wrote:
>>> On Sat, May 21, 2016 at 8:10 AM, Manfredi, Albert E
>>> <[e-mail]> wrote:
>>>>
>>>> The messages in this I-D are already clear.
>>>
>>> I disagree that they were already clear, and I think that something like
>>> Fernando's text will help a lot to clarify them.
>>>
>>> I will note, though, that this draft does the following:
>>>
>>> Deletes the text in RFC 2464 and other documents that defines how the IID is
>>> constructed, and replaces it with a pointer to this draft
>>> Makes it very clear that an implementation that does not choose to use a
>>> stable IID does not need to follow this draft
>>>
>>> The combination of #1 and #2 means that an implementation that does not
>>> choose to use a stable IID can legitimately form whatever IID it so chooses.
>>
>> Which I think is consistent with
>>
>> "Significance of IPv6 Interface Identifiers"
>> https://tools.ietf.org/html/rfc7136
>>
>> (without looking this draft being discussed because I'm about to go
>> out the door, it might be worth referencing that RFC if it usefully
>> supports some of the clarifications made).
> 
> So, reviewing the thread, I think we've been barking up the
> wrong tree to some extent (hence the new subject header above).
> 
> RFC4941 aims to be compatible with SLAAC but it includes this
> near the beginning of the protocol description:
> 
> '  2.  Create additional addresses based on a random interface
>        identifier...'
> 
> In other words it's *explicit* that there is always an existing
> address, which I think we agree is actually completely unnecessary
> for client computers. (It's also at variance with some of the wording
> in RFC4862, which frequently refers to IIDs as 'unique' per interface,
> which can't be right, otherwise RFC4941 simply wouldn't work.)
> 
> So, I think some RFC needs to state that it is legitimate to have a host
> with only RFC4941 global addresses.
> 
> Apart from that I'm quite happy with Lorenzo's comment above. Once
> the IID bits have no semantics, anything goes.

Exactly. So this is not the document to blame for "requiring" stable
addresses. If anything, such requirement comes from elsewhere (e.g.
RFC4941), and *that* "elsewhere" is the document to complain about and
to update.

Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492