The meeting minutes are here: https://notes.ietf.org/notes-ietf-125-opsarea?edit
", "time": "2026-03-16T06:06:03.000Z"}, {"author": "Dan Voyer", "text": "those are great presentations - the slide 4 of this actual presentation is very much what to expect IETF to solve
", "time": "2026-03-16T07:15:59.000Z"}, {"author": "Mohamed Boucadair", "text": "For full transparency, the guidance I have provided to the presenters is to to hilghight the specifics of their deployments and how the challenges/pb to be solved are connected to their deployment.
", "time": "2026-03-16T07:26:44.000Z"}, {"author": "Beno\u00eet Claise", "text": "Thank you Med for inviting those 3 operators (and providing the presentation guidelines). When I sum up the number of their subscribers (which shows the network scale), the IETF would be really stupid not to listen carefully which problems to tackle next!
", "time": "2026-03-16T07:32:30.000Z"}, {"author": "Joe Abley", "text": "wave
", "time": "2026-03-16T07:36:10.000Z"}, {"author": "Thomas Graf", "text": "Fully agree Benoit. I am taken aback listening to those 3 presentations and clearly speak out their network operator challenges and show what is possible with AI.
", "time": "2026-03-16T07:36:11.000Z"}, {"author": "Joe Clarke", "text": "@Benoit, I agree. But when I break some of this down, I don't necessarily see \"AI\" as novel piece. In some ways data management is not new, and we've been struggling to get vendors to align. In fact, AI is interesting (in my experience) as it helps normalize some of this unstructured or unaligned data.
\nWhat spoke best to me (in my experience) is the non-determinism of AI. I'm not sure that's something the IETF should tackle. But, if I had to pull out a couple of common themes they would be digital twin, and perhaps data reduction and distribution of data processing.
", "time": "2026-03-16T07:39:48.000Z"}, {"author": "Qin Wu", "text": "Good presentations, I see Top 3 common use cases discussing here include fault management, network optimization, network change. Good to see investigate how AI can help address challenges raised here.
", "time": "2026-03-16T07:40:00.000Z"}, {"author": "Cuiling Zhang", "text": "Splitting protocol development from operational maintenance would help +1
", "time": "2026-03-16T07:42:10.000Z"}, {"author": "Mohamed Boucadair", "text": "The draft Wes is talking about is https://datatracker.ietf.org/doc/html/draft-hardaker-dns-wgs-at-ietf-02
", "time": "2026-03-16T07:43:32.000Z"}, {"author": "Beno\u00eet Claise", "text": "@Joe, I agree. This is why I also mentioned different areas, not all related to AI: digital twin, knowledge graph, opentelemetry, etc.
", "time": "2026-03-16T07:43:53.000Z"}, {"author": "Qin Wu", "text": "@Joe, If we can make data better structured with context, semantics, constraints, , these data can be better consumed by AI and Model.
", "time": "2026-03-16T07:47:36.000Z"}, {"author": "Beno\u00eet Claise", "text": "@Qin, yes. This is a prerequisite.
", "time": "2026-03-16T07:57:00.000Z"}, {"author": "Ond\u0159ej Sur\u00fd", "text": "Oh, I like the \u201cI\u2019m just a messenger\u201d role ;). I want this too\u2026
", "time": "2026-03-16T07:57:35.000Z"}, {"author": "Joe Clarke", "text": "@Qin, absolutely, but we've been trying to do this with YANG, IPFIX, structured syslog, etc. What is clear is it takes $time for people to agree and align to this. Ultimately, it might be better to spend time creating network-trained models to align the data we have now knowing that we'll likely never get to a perfect state of aligned, structured data.
", "time": "2026-03-16T07:59:47.000Z"}, {"author": "Joe Clarke", "text": "But that is likely not the work of the IETF.
", "time": "2026-03-16T08:00:08.000Z"}, {"author": "Qin Wu", "text": "have you thought about how YANG based data can be better consumed by AI and Model? I see some hybrid approach has been proposed to address non-determine challenges. Happy to talk offline.
", "time": "2026-03-16T08:02:11.000Z"}]