Re: [OSPF] Advertising S-BFD discriminators in OSPF

[Acee Lindem <[e-mail]>]

Hi Michael, Manav, Karsten, 

We did have this discussion back when we put the first piece of non-OSPF information in the OSPF RI LSA and conceded that the OSPF RI LSA could be used for all types of information.

https://datatracker.ietf.org/doc/rfc5088/

On the same subject, I have further conceded that this will be used for variable length information and am doing a bis version of the RFC4970 accommodating multiple instance of the RI LSA. I will be presenting the updates in Honolulu. 

https://datatracker.ietf.org/doc/draft-acee-ospf-rfc4970bis/

Thanks,
Acee
P.S. I’m glad I didn’t reply last night - we might have missed out on all the analogies ;^). 


On Sep 26, 2014, at 4:08 AM, Karsten Thomann <[e-mail]> wrote:

> Hi,
> 
> RFC 4970 Section 3 includes:
> 
>   The purpose of the Router Information (RI) LSA is to advertise
>   information relating to the aggregate OSPF router.  Normally, this
>   should be confined to TLVs with a single value or very few values.
>   It is not meant to be a generic container to carry any and all
>   information.
> 
> I think the first sentence allows it, as it is related to the ospf router,
> even if it is not an ospf internal information.
> Michael is right that not everything should be in the RI LSA as already
> mentioned in the RFC, increasing the count of LSAs which need to be
> flooded across the domain has also drawbacks in term of scalability.
> 
> Regards,
> 
> Karsten
> 
> 
> Am 26.09.2014 08:58, schrieb Manav Bhatia:
>> Hi Michael,
>> 
>> :-)
>> 
>> Lets not deal with analogies (my bad!) lest we get misled by those.
>> 
>> Clearly you and I look at RI LSAs differently. I dont think what youre
>> suggesting is even remotely preposterous. However, i do believe that
>> RIs can serve more than merely announcing OSPF specific router
>> capabilities.
>> 
>> And once again, we already have a precedent where RI was used for
>> announcing non OSPF specific capability -- which means that the
>> co-authors of this draft arent the only ones who think of RIs as being
>> a generic tool for router capability advertisement.
>> 
>> Lets hear what others have to say on this.
>> 
>> Cheers, Manav
>> 
>> 
>> On Fri, Sep 26, 2014 at 11:53 AM, Michael Barnes <[e-mail]> wrote:
>>> Hi Manav,
>>> 
>>> That's a pretty funny analogy. I've got another one for you.
>>> 
>>> Say the RI LSA is like a corporate learjet for a tool company. When it gets
>>> flown around it's never full so you decide to use it to ship wrenches on it
>>> along with the executives. You're proud of your wrenches and want everyone
>>> to know how great they are, but do you really want to ship them that way?
>>> Maybe it's better to put the wrenches in their own vehicle.
>>> 
>>> So to be a little more serious, if the router wants to proudly proclaim its
>>> S-BFD capability then it deserves its own special LSA rather than being
>>> packed into one used for other things.
>>> 
>>> Cheers,
>>> Michael
>>> 
>>> 
>>> On 09/25/2014 06:27 PM, Manav Bhatia wrote:
>>>> Hi Michael,
>>>> 
>>>> Very interesting.
>>>> 
>>>> I think there is a disconnect because of our interpretation of what an
>>>> RI LSA is envisioned to carry. I assume its meant to advertise
>>>> "optional router capabilities" while you believe its to be used solely
>>>> for advertising "optional OSPF router capabilities". IMO, limiting the
>>>> scope of RI to just OSPF is like using a humvee with all its bells and
>>>> whistles to distribute balloons to the children in your friendly
>>>> neighborhood park! We wanted a mechanism wherein each router could
>>>> proudly tell the world that it was an S-BFD capable node and along
>>>> with it also advertise the unique discriminator that the others would
>>>> use to reach it. RI we felt, was the perfect tool that we could use
>>>> for this purpose.
>>>> 
>>>> And btw we're not the first ones to use RI for advertising a router
>>>> capability that isnt pertinent to OSPF per se (RFC 5088).
>>>> 
>>>> Cheers, Manav
>>>> 
>>>> On Fri, Sep 26, 2014 at 12:47 AM, Michael Barnes <[e-mail]>
>>>> wrote:
>>>>> Hi Manav,
>>>>> 
>>>>> Perhaps I missed some earlier discussion, on why you decided to add the
>>>>> S-BFD Discriminator TLV to the RI LSA, but I would prefer it not be in
>>>>> that
>>>>> LSA. I would like to leave the RI LSA with only information pertinent to
>>>>> OSPF rather than pollute it with information for which OSPF has no
>>>>> interest.
>>>>> If you're concerned with a trend of creating a new Opaque type for every
>>>>> application which might want OSPF to carry information for it, then I
>>>>> would
>>>>> suggest we create a generic Application Information LSA.
>>>>> 
>>>>> Regards,
>>>>> Michael
>>>>> 
>>>>> 
>>>>> On 05/30/2014 01:47 AM, Manav Bhatia wrote:
>>>>>> 
>>>>>> Hi,
>>>>>> 
>>>>>> We had submitted the following draft a couple of weeks ago.
>>>>>> 
>>>>>> http://tools.ietf.org/id/draft-bhatia-ospf-sbfd-discriminator-00.txt
>>>>>> 
>>>>>> This draft introduces a new OSPF RI TLV that allows OSPF routers to
>>>>>> flood the S-BFD discriminator values in the routing domain.
>>>>>> 
>>>>>> S-BFD is a new charter item (will be approved very soon) in the BFD WG.
>>>>>> 
>>>>>> Would appreciate comments on this.
>>>>>> 
>>>>>> Cheers, Manav
>>>>>> 
>>>>>> 
>>>>>> _______________________________________________
>>>>>> OSPF mailing list
>>>>>> [e-mail]
>>>>>> https://www.ietf.org/mailman/listinfo/ospf
>>>>>> 
>>>> .
>>>> 
>> _______________________________________________
>> OSPF mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/ospf
> 
> _______________________________________________
> OSPF mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/ospf