MOPS IETF 112 Minutes

November 11, 2021

Admin [5 min]

Note well
Agenda bash
Jabber scribe
Minutes taker https://notes.ietf.org/notes-ietf-112-mops#

Industry News/Experiences [20min] Video

SVA updates (Glenn Deen/Sanjay Mishra)
- slides: https://datatracker.ietf.org/meeting/112/materials/slides-112-mops-sva-update-to-ietf-112-00
- Jake: IPR rules? How much of this will come into the IETF rules at some point?
  - Glenn: There's different categories - API's and technical specs may have IP in them from SVA members and if you are a member you've signed up for the org's IP policies. The other catagories like PoC reports, White Papers, Best Practices really don't typically have IP in them and can be downloaded from the SVA web site and are intended to be read and used by the broader industry, even non-SVA members.
  - Glenn: We also have people who are actively working at keeping SVA and IETF engaged and cross-polinated

WG Docs [30 min]

Ops Cons — any follow up from WG last call which completed in early October? (Jake Holland) Video
- slides: https://datatracker.ietf.org/meeting/112/materials/slides-112-mops-status-operational-conditions-for-streaming-media-00
- draft: https://www.ietf.org/archive/id/draft-ietf-mops-streaming-opcons-07.html
- Discussion:
  - TL;DR: Need more on-list comments to demonstrate WG consensus, volunteers solicited to review and comment
  - Éric:
    - Only one review on the ML (Glenn) is a little thin to say we have WG consensus
    - Want to encourage others present to voice their opinions
    - Leslie:
      - on the agenda today to see if there are any surprises we might find in review
      - Let's take names: who is willing to review and say they've read it and it's ready
    - Sanjay: Will review
    - Glenn:
      - Already reviewed
      - Can spread to SVA folks and broader industry for more eyeballs
      - Notes that quite a bit of discussion occurred on GitHub rather than on the list, can we make others aware of this?
        
        Éric: Can mention this on the list because the list is still what "counts"
        
        Jake: Commenting on issues doesn't imply support for advancing the doc, needs separate confirmation on list, agree ML is normative venue
    - Renan: Looks good, will definitely review on the list as well
    - Leslie: need reviewers, but note that this doc is in WG LC - question is not "what doc would you write?" but "is this doc succeeding at its goals and ready to move on"
    - Ali: will ask students to review
    - Leslie:
      - Need 5 people to commit in chat to review in the next 2 weeks
        
        Mike English
        
        Renan Krishna
        
        Chris Lemmons
        
        Luis Contreras
        
        Sanjay Mistra
    - Jake: do we need author comments?
      - Magnus from chat: usually that's a request from the doc shepherd based on https://www.rfc-editor.org/rfc/rfc4858.html#section-3.1 (or rather, https://www.ietf.org/chairs/document-writeups/document-writeup-working-group-documents/)
Media Operations Use Case for an Augmented Reality Application on Edge Computing Infrastructure (Renan Krishna) Video
- slides: https://datatracker.ietf.org/meeting/112/materials/slides-112-mops-krishna-mops-ar-use-case-00
- Discussion:
  - Leslie: scope and direction questions from Spencer & Jake, can we discuss here?
    - Renan: we're looking to extract specifics of AR applications on edge computing resources/reasons to deploy edge computing, would be the direction we favor
  - Kyle: there's more applications than AR that need similar constraints
  - Jake: like what?
  - Renan: vehicle stuff, for instance. common thread is latency and battery-based power & heat constraints. Operational level there's a lot of choice of protocol defined by the application programmer, not the network operators, but the network operators need to provide certain qoe characteristics to meet the needs.
  - Leslie: what's special that requires edge computing. listed video ones, but maybe worth capturing other compute challenges that can't be captured by traditional means. synchronization of objects/shared reality and how it impacts real-time needs and locality is a key point to capture here.
  - Renan: yes, latency is key due to synchronization issues, and especially important in context of wireless networks for many of these devices, with its variability.
  - Leslie: same old sync problem as multi-cpu, it's a question of what can be pushed off elsewhere and what has to be done locally vs. what can be done non-locally.
  - Renan: yes, valuable point. Paper in the draft talks about 3 tiers, device, edge, cloud. Within these design constraints you have to think about latency and sync and how it's affected by AR requirements. Probably expand on some of that.
  - Leslie: thanks for getting doc in github: https://github.com/ietf-wg-mops/draft-ietf-mops-ar-use-case

Related IETF work [20 min]

(agenda restructure due to missing Spencer) * MOQ side meeting (Spencer Dawkins) (deferred briefly) * Follow up: Packet Difference Significance for Media Scalability (Lijun Dong) [10 min] Video * slides: https://datatracker.ietf.org/meeting/112/materials/slides-112-mops-mops-packet-significance-difference-use-case-00 * Discussion: * Jake: I wouldn't characterize my view as highly positive at this stage, but I do think this is possible in controlled networks with diffserv, and this is a much easier path than the ip options approach you mentioned, and before I'm interested in this, I want to understand from experiments how much you gain on QoE from trying to do this. * Magnus: +1 to part of jake, would want to see qoe experiments. this is kind of a simplistic view of modern codecs. * Lijun: did some research on what characteristics can be leveraged, but some of the video parts aren't our expertise. * Magnus: yes, checking what you need in the application is the homework you seem to still have to do here. * Lijun: ok. * From chat: Relevant prior work in this space in https://datatracker.ietf.org/doc/draft-even-avtcore-priority-markings/ * Leslie: maybe we can help find this work a home. * Magnus response from chat: it would be dependent on what video transport system you are targeting. Is this HTTP based, or RTP, or something else?

MOQ side meeting (Luke summary) Video
- Discussion:
  - Luke: moq in rough state at the moment, exploratory and defining use cases
  - Glenn: yes, but good stuff there
  - Luke: regarding what's out there, twitch has some webtransport stuff running, facebook doing stuff for ingest, there's also work on mapping srt over quic as well as just h3 in general for segments.
  - Jonathan: what's quic giving us?
  - Luke: congestion, hol blocking. gives some scheduling benefits. kinda too early to tell how much it matters yet, but we're looking.
  - Bechet in chat: I think main point in moq is can we use quic like RTP?

Retrospective [30 min] Video