Re: [OSPF] WG Adoption Poll for "OSPF LLS Extensions for Local Interface ID Advertisement"

[Peter Psenak <[e-mail]>]

Hi Tony,

please see inline:

On 10/05/17 17:57 , prz wrote:
> On Tue, 09 May 2017 12:54:08 +0200, Peter Psenak <[e-mail]> wrote:
>> Hi Tony,
>>
>> let me try to clarify.
>>
>> 1. This draft does not change, nor does it conflict with RFC3630 in
>> any way.
>
> Peter, my bad, I got confused forgetting that the Link ID in 3630 is a
> different beast from the 4203 link ID. Thanks for pointing that out.
>
>>
>> 2. This draft does not change anything in RFC4203 either. It provides
>> an alternative and more generic way to exchange Link Local Identifier
>> on the interface. Your are right that in our draft we need to specify
>> the behavior in case the mechanism described in RFC4203 AND the new
>> mechanism specified in our draft are both active at the same time. We
>> will add a new section in a next version that covers this part. I
>> don't believe it will be too difficult, given that the value of the
>> Link Local Identifier is the same same in both cases, the only
>> difference is the the mechanism how it is advertised.
>>
>
> Discussion ongoing ... Your question about 4203 deployment is valid and
> needs answers and yes, we need a backward compatibility section explaining
>
> a) @ which OSPF hello states the new signaling is supposed to be present

all hellos, will clarify that in the draft

> b) what happens if the value ever changes (tear down adjacency?) or whether
>     it's a violation

no need to tear down the adjacency. Just update anything that uses the 
link ID.


> c) what happens if flooded 4203 values (on LSAs) contradict signaling
>     (e.g. you're holding to the flooded LSA with "old" value) while
>     neighbor LLS'es a new value

we must set some preference, e.g. LLS value is always preferred.

> d) what happens if both signaling types show up on the link

if they are signaling the same value, there is no problem. If values are 
different, we prefer LLS.

>
> Per earlier mails, clarification WHEN the new signaling is supposed to
> be used
> would be also helpful. 4203 Link ID signaling seems to be used only in
> case of unnumbered.

any link, will clarify that in the draft.

>
> So, overall I think we agree on scope of the problem that needs to be
> addressed so we get a coherent set of standards out

so would you agree to make this a WG document?

thanks,
Peter

>
> thanks
>
> --- tony
>
>
> .
>