Re: [v6ops] Should we use "unnumbered" in IPv6?

All,

What's one more email in such an active conversation.  Here area few 
points to take note of.

- When I suggested "administratively unnumbered", I was intending to 
find text which has less imposed (vs just "unnumbered") meaning while 
attempting to keep within the spirit of the text in section 2.1.2.

- The intention of this text was to describe the configuration choices 
for links within a network, which was to apply a ULA/GUA or to not apply 
a ULA/GUA (in which case the LL address would be the only one that remains).

- The emphasis for the draft section is about what the 
operator/administrator will (or will not) apply as configuration and not 
what potential behaviors will be exhibited (although this would have 
relevance for troubleshooting, but I don't think that is specific to 
this draft in that it's something to consider more widely).

- The action in the one case (by the administrator) would be to not 
apply addressing (additional) to a link, in which case the result is 
LL-only.

- When I re-read the text to myself, substituting "administratively 
unnumbered" , it seemed to make sense to me (I am biased), and the flow 
of the text seems to make sense as I put myself into the position of a 
operator reading this for guidance. I don't think they will be confused 
at all (given the context and full text).

- Section 2.1.2 text does present the options as (a). link-local 
addresses ("unnumbered") or (b) have global or ULA addresses assigned. 
which are in fact the choices (where I am already presuming we would 
have replaced "unnumbered" with "administratively unnumbered")

- I think it's important for us to realize the intention of the section, 
and what we are trying to accomplish by providing the general choices to 
the reader.

- We then have a reference to "draft-ietf-opsec-lla-only" which is a 
document that then goes into more detail of the approach, advantages and 
caveats of using LL-only addressing models.

I don't want to undermine the great conversation we had on the list on 
the potential impacts of using ll-only and the behaviors which we may 
see, but feel that (as I think others have noted), can be captured in a 
more appropriate document.

my two cents,

regards,

Victor K

On 2015-03-23 8:15 PM, Alexandru Petrescu wrote:
> One cant have an interface be unnumbered yet numbered by a link-local 
> address.
>
> One can't have 'an interface with only link-local address' because 
> there is the MAC address as well.
>
> One cant have a '_link_ with only link local addresses' because a link 
> is fixed into computers which have addresses configured on interfaces 
> on that link, but also on virtual interfaces, and on other real 
> interfaces.  A link by itself does not make much sense.
>
> We could as well call this a 'transit-only' link.
>
> Alex
>
> Le 23/03/2015 18:56, Hemant Singh (shemant) a écrit :
>> Sorry, but “unnumbered” has already been Nacked by EricV and others in
>> this mailer because the interface borrows the loopback interface’s ipv6
>> global address to source some ipv6 packets.
>>
>> “Link-local only” doesn’t look good to me because what if the interface
>> actually tried DHCPv6 or SLAAC for a GUA but DHCPv6 and SLAAC failed?
>> Thus “administratively” seems to be key so that one is clear the admin
>> managing the interface on the node consciously wanted only a link-local
>> address on the interface.  Why not
>>
>> “administratively link-local only”.
>>
>> Hemant
>>
>> *From:*v6ops [mailto:v6ops-bounces@ietf.org] *On Behalf Of *Mark ZZZ 
>> Smith
>> *Sent:* Monday, March 23, 2015 7:07 PM
>> *To:* Victor Kuarsingh; v6ops@ietf.org
>> *Subject:* Re: [v6ops] Should we use "unnumbered" in IPv6?
>>
>> I still think 'link-local only' is better, as it is very explicitly
>> saying what the addressing scheme is. There is no and cannot be any
>> ambiguity.
>>
>> I think 'administratively unnumbered' is close to the inverse of that
>> and less descriptive and specific.
>>
>> ------------------------------------------------------------------------
>>
>> *From:*Victor Kuarsingh <victor@jvknet.com <mailto:victor@jvknet.com>>
>> *To:* v6ops@ietf.org <mailto:v6ops@ietf.org>
>> *Sent:* Monday, 23 March 2015, 23:23
>> *Subject:* Re: [v6ops] Should we use "unnumbered" in IPv6?
>>
>>
>> Eric/All,
>>
>> What about the term "administratively unnumbered?"  This could be
>> defined as an interface where the network administrator did not
>> specifically number the interface, but does not preclude it from having
>> an address at all (i.e. LL address).
>>
>> regards,
>>
>> Victor K
>>
>>
>>
>> On 2015-03-10 5:15 PM, Eric Vyncke (evyncke) wrote:
>>> Philip
>>>
>>> In the legacy world of IPv4, unnumbered actually means using a 
>>> global IPv4
>>> address 'stolen' from another interface.
>>>
>>> In the current world of IPv6, every interface has at least a link-local
>>> address which can be used for IGP and other purposes (see RFC 7404) 
>>> which
>>> is NOT equivalent (IMHO) to 'unnumbered interface'
>>>
>>> So, I am afraid that you should use another working (albeit cumbersome)
>>>
>>> -éric
>>>
>>> On 10/03/15 15:52, "Philip Matthews" <philip_matthews@magma.ca 
>>> <mailto:philip_matthews@magma.ca>> wrote:
>>>
>>>> Just want to get some direct feedback from the WG on the use of the 
>>>> terms
>>>> "unnumbered interface" and "unnumbered link" in IPv6.
>>>>
>>>> Mark Smith, in some comments on the Design Choices draft , has said 
>>>> that
>>>> he prefers the terms "interface with only link-local address" or "link
>>>> with only link-local addresses" instead.  See the thread with title 
>>>> "I-D
>>>> Action: draft-ietf-v6ops-design-choices-04.txt".
>>>>
>>>> What does the WG think?  Should we use the "unnumbered" terms, or 
>>>> should
>>>> we use just the "... with only link-local address" terms? Is it
>>>> acceptable to mix the two sets of terms?
>>>>
>>>> Philip
>>>>
>>>>
>>>> _______________________________________________
>>>> v6ops mailing list
>>>> v6ops@ietf.org <mailto:v6ops@ietf.org>
>>>> https://www.ietf.org/mailman/listinfo/v6ops
>>
>>
>>> _______________________________________________
>>> v6ops mailing list
>>> v6ops@ietf.org <mailto:v6ops@ietf.org>
>>> https://www.ietf.org/mailman/listinfo/v6ops
>>
>> _______________________________________________
>> v6ops mailing list
>> v6ops@ietf.org <mailto:v6ops@ietf.org>
>> https://www.ietf.org/mailman/listinfo/v6ops
>>
>>
>>
>> _______________________________________________
>> v6ops mailing list
>> v6ops@ietf.org
>> https://www.ietf.org/mailman/listinfo/v6ops
>>
>
>
> _______________________________________________
> v6ops mailing list
> v6ops@ietf.org
> https://www.ietf.org/mailman/listinfo/v6ops