Hi Les, 

> On Jul 14, 2026, at 6:16 PM, Les Ginsberg (ginsberg) <[email protected]> 
> wrote:
> 
> Acee -
> 
> Please see inline.
> 
>> -----Original Message-----
>> From: Acee Lindem <[email protected]>
>> Sent: Tuesday, July 14, 2026 1:15 PM
>> To: Les Ginsberg (ginsberg) <[email protected]>
>> Cc: lsr <[email protected]>; draft-admnr-lsr-igp-measurement-
>> [email protected]
>> Subject: Re: Questions on draft-admnr-lsr-igp-measurement-group-02
>> 
>> Hi Les,
>> 
>>> On Jul 14, 2026, at 3:41 PM, Les Ginsberg (ginsberg) <[email protected]>
>> wrote:
>>> 
>>> The latest version of the draft defines a modified form of the sub-TLV:
>>>   0                   1                   2                   3
>>>    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
>>>   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>>>   |     Type      |    Length     |   Group ID    | AMP Protocol  |
>>>   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>>>   |                                                               |
>>>   |              IPv4 or IPv6 Host Address (4 or 16 octets)       |
>>>   |                                                               |
>>>   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>>> Several questions:
>>> 1) Given the Router Capability TLV carries V4 Router ID/V6 Router ID – why 
>>> is
>> an application specific address necessary – and how would it be chosen?
>>> (I have asked this before but did not receive a substantive answer)
>> 
>> Specification of the IP endpoint allows sessions with multiple endpoints on 
>> the
>> router.
>> We can discuss this requirement but using a different EP could be used to
>> measure a different path.
>> 
> 
> [LES:] You will need more than a different node address in order to measure a 
> specific path. If that is what you are after I think you need more 
> specification of how endpoint addresses are chosen.
> Until you do, I am seeing it as redundant with the already advertised router 
> id that is carried in every Router Capability TLV.

Fair enough - we can discuss amongst the authors. I know that different EPs 
can, in fact, imply different data planes with different encapsulations and we 
may want to measure them independently. 

> 
>> 
>>> 2) You are now requiring an address per “group ID”. Why? And how would a
>> group specific address be chosen?
>> 
>> The Group ID allows AMP sessions amongst different disjoint groups of 
>> routers.
>> If this isn't a requirement, all routers in the domain can share the same 
>> Group
>> ID.
> 
> [LES:] Understood. I am just saying the group ID is underspecified. More on 
> that below.

Sure. 

> 
>> 
>> 
>> 
>>> 3) Section 4 states:
>>> “Given that an IP endpoint may be used for multiple measurement groups
>>>   and AMP protocols, a router should only establish a single session of
>>>   a given protocol type between the same two pairs of IP endpoints
>>>   (independent of how many groups the endpoint participates in).  “
>>> This would argue that the encoding of the sub-TLV should be a list of Group
>> ID/AMP Protocol pairs so that unnecessary duplication is avoided.
>> 
>> That encoding would be an option. However, this simply says that we don't
>> want to establish multiple AMP sessions for the same protocol between the
>> same two endpoints.
>> If you have two groups (1 and 2), it is entirely possible that two routers 
>> (A and
>> B) would participate in both groups, where the other routers in groups 1 and 
>> 2
>> are different. We wouldn't want multiple sessions between routers A and B.
>> 
> [LES:] Understood. I am just asking that you be efficient in your encoding - 
> something IS-IS has always been attentive to.

What if the primary use case is a single group? With a list of groups, you'd 
need explicit AF specification so really only save 1 octet per group with a 
list.  


> 
>>> 4) There is no discussion of what a group ID is nor how consistency in
>> configuration between nodes is achieved.
>> 
>> Is this a trick question? How is configuration consistency ever achieved if 
>> not
>> incumbent on those doing the configuration?
>> 
> [LES:] No - this isn’t a trick question. The group id simply isn’t specified.
> What numbers are valid? Are certain values reserved? Do they relate to 
> anything in the associated protocol?
> It might be as simple as saying "the group id is an arbitrary number from 
> 0-255" - but you haven't said anything.

Fair enough, we can add a discussion. 

Thanks,
Acee


> 
>   Les
> 
>> Thanks,
>> Acee
>> 
>> 
>> 
>>> Thanx.
>>>     Les
>> 
> 

_______________________________________________
Lsr mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to