March 21, 2021.
Meeting materials are at https://datatracker.ietf.org/meeting/materials/
Magnus to talk about Media over QUIC (a non WG BOF)
Of course, the battery dies on the microphone
Goal is to look for "solutions for today's world"
Acknowledge CDBs
and other stuff, screen moved too fast for me to type
About QUIC
- multiple streams
- unreliable datagrams
- congestion control
- flexbility in reliability between full and partial
- path migration
- user space implementation
BOF: What to Tackle (Latency)
- Interactive media
- live media
- on-demand media
Live Media - is probably a starting point
- TV production -> Broadcast Facility; Outside Broadcast; Live Streaming; Distribution platform
On top there are"Hybrid Use Cases"
For e.g Corp all employee meetings with Q&A
So, can we by thinking of these uses cases from start
- Get a more useful solution
- more long term
- not boil the ocean
Magnus shared various "useful" References
Says, dont have more but come to the BOF next Wed with ideas and questions
Sanjay Mishra asks: you mentioned CDN acknowledgement so anything you want to add?
Yes, our protocol does not acknowledge CDN but have to think of CDN support
Glenn Deen: QUIC to some other transport and back to QUIC? Some transitions?
Mo Zanaty - CDN and Relay nodes
Sanjay Mishra
Gleen Deen: One question is about transition - who will support QUIC and who will not, etc.
Magnus: come to the BOF (even thought it's early in California)
Mo Zanaty - CDN and Relay nodes
Any questions?
No questions asked.
https://dashif.org/webRTC/#documents
Marrying WebRTC and DASH for Interacive Streaming
Gives a DASH + WebRTC Use Cases. Lists three categories
- Fallback (Real-time WebRTC with fallback to DASH)
- Interleaved (RTWebRTC interleaved with DASH)
- Concurrent(RTWebRTC concurrently with DASH)
Compares DASH and WebRTC
- Shows a big chart showing FEatures for DASH and WebRTC
Hybrid Delivery - How would that work? You will have
- Publisher: Live Content
- Publisher: Pre-record Content
- Examples of Hybrid Client Architecture
Work-Flow -> Shows a slide how the work-flow will look like
Integres WebRTC Media Server and WebRTC Client with the DASH Server and DASH Player
Shows another slide "From Discover to Streaming". Shows current state of real-time Streaming and then Goal for Real-time Streaming
Work for WebRTC Extenson
- has got a lot of work to do for standardization efforts
Work For DASH
- DASH needs to determine APIs to be used between WebRTC clients and DASH
Lists various references for "Further Reading"
- DASH IF Report Summary: https://dashif.org/webRTC/
- Full Report: https://dashif.org/webRTC/report
- Interest Survery: https://forms/gle/Yy89GGeMsXYQixBZ6
Glenn Deen - Question on AB switch - how does it work when you are watching WebRT and need to pop back, which is a DASH thing, for replays, etc.
Julia - we have an app that shares the window
Glenn - for different frame rates, etc. you may run into some weird interactions about synchronization
Glenn - how do you do watermark?
Julia - we don't.
https://datatracker.ietf.org/doc/draft-ietf-mops-ar-use-case/
Will Start with AR document first
Renan will give an update - what changed?
Gives a thank list of Spencer, Rohit, Jake, Kiran and Ali
Propose replacing AR with XR (Extended Reality)
Scope of XR - Real env, AR and VR
Scode should focus on who the intended audience is so Renan is proposing a new scope in the Draft (scope is Network Operators)
Would like to solicit reviews and gives the Github repo (gives many Thanks to Kyle)
Spencer Dawkins: Comments on the scope - The proposed scope is those who are doing AR and edge computing. My question is whether there are people doing AR but NOT using edge computing. If there are, would we provide guidance for them in another document, or expand the scope of this one?
Renan says edge computing solution is an over arching umbrella where other options can be added such as multi path. We do feel multi computing solves the high intensive compute applications. He feels edge computing the key things.
Jake: Where are you looking for Edge computing? Have you looked at COINRG?
Spencer: Thing he is wondering about is that COINRG is an IRTF RG, and as such is not chartered to produce IETF Standards-track RFCs. If we have to look to an RG for guidance, are we talking about Engineering or Research?
Eric Vyncke: Agrees that need to pay attention to, i.e., check whether the contents fits the MOPS charter.
Cullen Jennings: This document is operational; Edge compute or cloud compute, the only question is just latencies; you can trade off
Suhas Nandakumar: Computing on edge and what you can offload?
Spencer: looking at the draft content, most of the content beyond the intro has not changed in a while. There is an interesting description of Latency tied to AR, but is this guidance to operators? There's a description of TCP consideration, but I'm not sure it says any more than the OPCONS draft, and if you are doing AR over TCP you will run into a lot of buildings. I have no doubt that there is Engineering work that is required here (Operations Engineering), but we should have the right scope of the document before we do a lot of word smithing.
Leslie says the document aim is clearly within scope. Have moved the document to GitHUb for comments. Focus on media delivery. Directly focused to this group.
Spencer is willing to help with the document. Let's write issues for now, instead of writing PRs.
Leslie: Working remotely has been an issue and would like to have an agreement of scope and then have the WG work on it. Have the document line up with the WG charter.
Spencer: Media delivery and what it takes to do media delivery, is that right?
Jake Holland: Agrees with Cullen. There is good Ops work to do here. How central is compute offload? What are the bandwidth and latency requirements? What kind of network is needed? Where is the scaling consideration?
Kyle: Says good to put these issues in GitHub.
Leslie: Helpful discussions.
Renan: Thanks everyone. Have just focused on Scope and Abstract and the intended audience. Have taken Spencer's comment on TCP. Want to make sure everyone agrees on scope.
Leslie: Discuss more on the list.
Leslie: Eric did his AD review. There is work to be done by draft authors.
Spncer: Noting there are two other authors and invite them to talk along with him. Says, not sure about Eric's comments on Latency. Definition of Latency category is reasonable for "streaming media", we can talk about whether latnecy for video conferencing should even be mentioned?
Spencer says Ali convinced him to distinguish between streaming video latencies and all video latencies in this document.
Spencer is comfortable on the rest of comments from Eric.
Eric Vyncke: Most comments are generally straight forward
Warren: We have tried the living document. Problem is everyone has got their own use case. In the past, what seems to have to work is the WG chairs have the document and WG Chairs drive the consensus. If folks have the stomach for engaging in discussion of a living document, they can but....
Spencer: what we had proposed a revision or two earlier was to use tinyURLs in the draft that point to a living document, and someone asked "what happens if tinyURLs go away?". We probably don't need to solve all the problems with "living documents" if we just use a URL that we control, such as an IETF GitHub wiki.
Glenn: Latency is a moving bench mark. Discussion on a living document is also questionable.
Leslie: Thanks Sanjay Mishra for shepherding the document.
Cullen: How do you rely on the SVA document? Is there just a reference of document.
Leslie: To look at the industry that is relying on our (IETF) document.
Cullen: WebRTC is a good example, where W3C references a very detailed explanation of protocol dependencies in a document, but the chairs also maintain a spreadsheet with much less detail, that gets used a lot. Both have been useful.
Spencer: We are also chartered to provide input to protocol groups about gaps operators see when they want to use our protocols. So producting a document that can provide such an input may be useful. We haven't talked about gaps much, so far, but having MOPS WG join the MOQ BOF will be good.
No other business was mentioned.
The chairs thanked everyone for attending ...