Re: [Idnet] IETF99 for applying AI/ML into network management: Follow-up

[Jérôme François <[e-mail]>]

Hi Sheng,

That's fine with me. So maybe, my suggestion would to have NMRG session
where people present their own use cases without giving details about
specific obtained results or ML algorithms but with a focus of what are
the different processing steps, interfaces among them, data to be
used... in an ideal case trying to see how this fits with the proposed loop.

Would that be a good plan ?

jerome

Le 27/07/2017 à 13:25, Sheng Jiang a écrit :
> Hi, Jerome,
>
> I do understand your concern. One of the historic stories is that we had two NMRG sessions/workshops back to IETF 88, 89. It served us both purposes - community building and scope refinement well. Because these two sessions, we developed two RFC 7575 & 7576. It gives us good foundation to form ANIMA WG. The problem we have now is actually TOO open and wide. If we could convergent on GENERIC problem across use case, interfaces, foundation functionalities, it would be the best. If we don't know what might be GENERIC for now, we may have to deeply study various use cases, then try to find the common among them.
>
> Regards,
>
> Sheng
>
>> -----Original Message-----
>> From: Jérôme François [mailto:[e-mail]]
>> Sent: Thursday, July 27, 2017 3:35 AM
>> To: Sheng Jiang; yanshen; [e-mail]
>> Subject: Re: [Idnet] IETF99 for applying AI/ML into network management:
>> Follow-up
>>
>> Hi,
>>
>>> Of course, we would like to target a WG if we could. However, so far, we have
>> not converged at all. It seems there are full of possibilities to apply AI/ML in
>> networks. It seems everybody is talking about their own use cases. Until we
>> converged with one or two very specific use cases that are also generic and
>> significance, or we might find some fundamental functionalities that are
>> generically support various AI use cases, we may not have anything to
>> standardized. So far, we are still in the way to find these work items. So, let's
>> keep discussing use cases and fundamental functionalities. Hopefully, with
>> more discussion, we could reduce our scope to be more and more focus. Then
>> we may have concrete work items to form a WG.
>>>
>> I agree that before formally targeting a WG, we need something specific to be
>> address. I said "something" because it could be a use case where AI is applied,
>> a generic problem across use case, interfaces....
>>
>> But my little concern is that I would avoid that planned workshop would be
>> stricly limited to particular use cases, problems or architecture.
>> In my opinion, we have to take care of the two items (community building with
>> the workshop and scope refinement) in parallel but we could, of course, expect
>> that the workshop gives us some input to refine our target in the proper way.
>>
>> jerome