[{"author": "Hang Shi", "text": "

Hi

", "time": "2023-03-30T04:03:32Z"}, {"author": "Hang Shi", "text": "

Is the chat working?

", "time": "2023-03-30T04:03:40Z"}, {"author": "Alan Frindell", "text": "

yes

", "time": "2023-03-30T04:03:45Z"}, {"author": "Kirill Pugin", "text": "

bribes!?

", "time": "2023-03-30T04:05:36Z"}, {"author": "Kirill Pugin", "text": "

it worked at/before interim meeting...

", "time": "2023-03-30T04:07:15Z"}, {"author": "Glenn Deen", "text": "

Is J. Postel's middle name actually Bruce?

", "time": "2023-03-30T04:07:23Z"}, {"author": "\u516b\u8c37\u822a\u592a", "text": "

lol

", "time": "2023-03-30T04:07:47Z"}, {"author": "Kirill Pugin", "text": "

receivers coming and going is not specific to realtime use cases, imho

", "time": "2023-03-30T04:10:42Z"}, {"author": "Ted Hardie", "text": "

@meetecho can you increase the remote volume a bit?

", "time": "2023-03-30T04:15:48Z"}, {"author": "Lorenzo Miniero", "text": "

Ack, just a sec

", "time": "2023-03-30T04:16:28Z"}, {"author": "Kirill Pugin", "text": "

how relay would know that F5 doesn't depend on F3?

", "time": "2023-03-30T04:20:18Z"}, {"author": "Ali Begen", "text": "

Am I the only one who thinks this is really a discussion of the shade of gray when we have not agreed on some black vs. white matters?

", "time": "2023-03-30T04:24:05Z"}, {"author": "Jonathan Lennox", "text": "

No, I'm discussion

", "time": "2023-03-30T04:24:10Z"}, {"author": "Lucas Pardue", "text": "

We need priorities in the meetecho queue, so that the discussion people don't HoL block the clarifying people

", "time": "2023-03-30T04:25:01Z"}, {"author": "Ali Begen", "text": "

Just because a frame depends on a frame, it does not mean it becomes useless if the reference frame is gone.

", "time": "2023-03-30T04:25:10Z"}, {"author": "Alex Chernyakhovsky", "text": "

You mean ... like zulip topics?

", "time": "2023-03-30T04:25:16Z"}, {"author": "Lucas Pardue", "text": "

Zulip is a good example of something too complicated that doesn't get used :D

", "time": "2023-03-30T04:25:39Z"}, {"author": "Luke Curley", "text": "

don't worry I'm also confused

", "time": "2023-03-30T04:25:55Z"}, {"author": "Hang Shi", "text": "

+1 to Ali

", "time": "2023-03-30T04:26:01Z"}, {"author": "Alex Chernyakhovsky", "text": "

Said like someone who didn't get forced to use MIT Zephyr and Barnowl :)

", "time": "2023-03-30T04:26:01Z"}, {"author": "Lucas Pardue", "text": "

I am forced to use Google Chat, how does that compare?

", "time": "2023-03-30T04:26:31Z"}, {"author": "Spencer Dawkins", "text": "

Ali Begen said:

\n
\n

Am I the only one who thinks this is really a discussion of the shade of gray when we have not agreed on some black vs. white matters?

\n
\n

Ali, there are definitely prerequisite discussions for this excellent presentation. :innocent:

", "time": "2023-03-30T04:26:46Z"}, {"author": "Will Law", "text": "

@Ali - prioritization in the face of congestion is a key issue for MoQ and it informs architectural decisions about what information to expose in the objects and what rules realy must apply. I do agree that the audience is being dropped in to this with little context.

", "time": "2023-03-30T04:26:53Z"}, {"author": "Ted Hardie", "text": "

Are you folks sure you didn't intend to be in MiMi?

", "time": "2023-03-30T04:26:56Z"}, {"author": "Alex Chernyakhovsky", "text": "

@Lucas Pardue depends, do you like your pain in your left hand or your right hand?

", "time": "2023-03-30T04:27:12Z"}, {"author": "Kirill Pugin", "text": "

how would stream per group going to achieve desired prioritization?

", "time": "2023-03-30T04:28:27Z"}, {"author": "Magnus Westerlund", "text": "

So a basic question here, can one have two different streams for the same media source. I am just wondering if tree based dependencies can partially be solved by simulcast switching so that one has a chance to flatten your tree to a linear structure?

", "time": "2023-03-30T04:28:40Z"}, {"author": "Ali Begen", "text": "

@Will Law for most video scenarios, layered stuff does not apply at all. It does apply to interactive, I get that, but I am not comfortable with this topic hijacking the valuable airtime we have in this session.

", "time": "2023-03-30T04:29:25Z"}, {"author": "Ali Begen", "text": "

Further, people assume layering gives them all the capabilities one could get from rate adaptation, that is plainly wrong. There are limits in terms of what resolutions you can switch from/to. Frame rate adjustments are easy but anything else is costly and quite limited.

", "time": "2023-03-30T04:30:38Z"}, {"author": "Cullen Jennings", "text": "

Oddly this relates to how many web transport sessions we end up having between each relay.

", "time": "2023-03-30T04:31:21Z"}, {"author": "Hang Shi", "text": "

Does anyone use layer encoding beyond the simple frame rate changing?

", "time": "2023-03-30T04:33:20Z"}, {"author": "Spencer Dawkins", "text": "

Cullen Jennings said:

\n
\n

Oddly this relates to how many web transport sessions we end up having between each relay.

\n
\n

\"You are in a maze of twisty little passages, all alike\". That NEVER gets old ...

", "time": "2023-03-30T04:33:24Z"}, {"author": "Jonathan Lennox", "text": "

Chrome WebRTC uses spatial layered encoding

", "time": "2023-03-30T04:33:43Z"}, {"author": "Hang Shi", "text": "

What is the MoQ session?

", "time": "2023-03-30T04:34:17Z"}, {"author": "Alan Frindell", "text": "

i think he said it's another name for broadcast/emission/track-bundle

", "time": "2023-03-30T04:34:57Z"}, {"author": "Murray Kucherawy", "text": "

Remember to get in the queue plz

", "time": "2023-03-30T04:36:46Z"}, {"author": "David Schinazi", "text": "

Cullen was so excited to get in the queue that he did it twice

", "time": "2023-03-30T04:37:49Z"}, {"author": "Cullen Jennings", "text": "

The struggle is real

", "time": "2023-03-30T04:39:24Z"}, {"author": "Kirill Pugin", "text": "

is it required for draft to support ALL requirements before adoption?

", "time": "2023-03-30T04:42:12Z"}, {"author": "Jonathan Lennox", "text": "

No, but I think it's a good idea to at least have an informal understanding of how support for all requirements are ultimately going to fit into the solution.

", "time": "2023-03-30T04:44:16Z"}, {"author": "Ali Begen", "text": "

clarifying question: are control messages also unidirectional?

", "time": "2023-03-30T04:45:10Z"}, {"author": "Kirill Pugin", "text": "

we keep coming back to Streams vs. Datagrams...

", "time": "2023-03-30T04:45:22Z"}, {"author": "Ted Hardie", "text": "

There needs to be rough consensus in the working group that you can get from a draft to the deliverable. So it doesn't have to have everything, but everyone needs to see a path from the draft to the deliverable. The deliverables in the charter include both streaming and realtime, so they would need to be a path for both (or even a system that combines the two).

", "time": "2023-03-30T04:45:31Z"}, {"author": "Hang Shi", "text": "

if the datatream is unidiretional, does the ACK come back from control stream?

", "time": "2023-03-30T04:45:31Z"}, {"author": "Spencer Dawkins", "text": "

Ali Begen said:

\n
\n

Further, people assume layering gives them all the capabilities one could get from rate adaptation, that is plainly wrong. There are limits in terms of what resolutions you can switch from/to. Frame rate adjustments are easy but anything else is costly and quite limited.

\n
\n

I am wondering how closely this work is tied to specific TYPES of codecs and encoding. If it's closely coupled, I THINK this would require changes to the MOQ protocol to support a different kind of encoding. I THINK that's what you're pushing on.

", "time": "2023-03-30T04:45:46Z"}, {"author": "Luke Curley", "text": "

re: tagging on GitHub, it doesn't look like something I can enable for non-collaborators

", "time": "2023-03-30T04:46:58Z"}, {"author": "Ali Begen", "text": "

Spencer Dawkins said:

\n
\n

Ali Begen said:

\n
\n

Further, people assume layering gives them all the capabilities one could get from rate adaptation, that is plainly wrong. There are limits in terms of what resolutions you can switch from/to. Frame rate adjustments are easy but anything else is costly and quite limited.

\n
\n

I am wondering how closely this work is tied to specific TYPES of codecs and encoding. If it's closely coupled, I THINK this would require changes to the MOQ protocol to support a different kind of encoding. I THINK that's what you're pushing on.

\n
\n

I am simply saying that there is an elephant in the room (which is today addressed by DASH/HLS) and it is still not clear what MoQ will have to do in order to be even able to have a chance to replace them.

", "time": "2023-03-30T04:47:25Z"}, {"author": "Spencer Dawkins", "text": "

Ted Hardie said:

\n
\n

There needs to be rough consensus in the working group that you can get from a draft to the deliverable. So it doesn't have to have everything, but everyone needs to see a path from the draft to the deliverable. The deliverables in the charter include both streaming and realtime, so they would need to be a path for both (or even a system that combines the two).

\n
\n

Yes. IMO, we chartered with the goal of producing one protocol that can be used for both live and interactive applications, without defining in much detail how those two classes of applications differ. I think we're figuring that out now. :grinning_face_with_smiling_eyes:

", "time": "2023-03-30T04:49:55Z"}, {"author": "Robert Sparks", "text": "

Luke Curley said:

\n
\n

re: tagging on GitHub, it doesn't look like something I can enable for non-collaborators

\n
\n

Invite collaborators by request?

", "time": "2023-03-30T04:50:19Z"}, {"author": "Ted Hardie", "text": "

@Robert in the short term, folks can send to the chairs and we'll apply the tags.

", "time": "2023-03-30T04:50:47Z"}, {"author": "Ted Hardie", "text": "

When we get to the working group draft, we can work something more scalable out.

", "time": "2023-03-30T04:51:05Z"}, {"author": "Mike English", "text": "

good morning

", "time": "2023-03-30T04:52:55Z"}, {"author": "Spencer Dawkins", "text": "

Ali Begen said:

\n
\n

I am simply saying that there is an elephant in the room (which is today addressed by DASH/HLS) and it is still not clear what MoQ will have to do in order to be even able to have a chance to replace them.

\n
\n

That IS a bigger elephant than I thought you were pointing to.

", "time": "2023-03-30T04:53:37Z"}, {"author": "Kirill Pugin", "text": "

@Christian, is it fair to say that in realtime use case Bundle is a Conference?

", "time": "2023-03-30T05:01:28Z"}, {"author": "Alan Frindell", "text": "

my read is that bundle is hop-by-hop, so it supports the requirement suhas mentioned (splitting and merging)

", "time": "2023-03-30T05:02:16Z"}, {"author": "Jordi Cenzano", "text": "

IMHO allowing bundles to have multiple UNsyncrhonized sources is not very intuitive and creates the need for another kind of grouping (tracks from same sync domain), at least the decoder needs to know that

", "time": "2023-03-30T05:02:43Z"}, {"author": "Robert Sparks", "text": "

(surprised that the naming bikeshed didn't go down the mixing panel names for grouping things, and also surprised that there's not deep PTSD being inflicted on participants here every time someone says bundle)

", "time": "2023-03-30T05:02:59Z"}, {"author": "Kirill Pugin", "text": "

catalog can have that information

", "time": "2023-03-30T05:03:08Z"}, {"author": "Jonathan Lennox", "text": "

@jordi: Yes, so it needs to be described in the Catalog, but the Relay doesn't need to care.

", "time": "2023-03-30T05:03:12Z"}, {"author": "Christian Huitema", "text": "

Kirill, I am afraid that there is no single definition of what a conference is...

", "time": "2023-03-30T05:03:13Z"}, {"author": "Kirill Pugin", "text": "

but it sort of creates implicit grouping

", "time": "2023-03-30T05:03:16Z"}, {"author": "Christian Huitema", "text": "

No, it creats potential grouping

", "time": "2023-03-30T05:03:34Z"}, {"author": "\u516b\u8c37\u822a\u592a", "text": "

so, is Bundle basically a set of audio tracks and video tracks? then Warp can prioritize audio tracks over video tracks within a bundle as draft-04 says?

", "time": "2023-03-30T05:04:53Z"}, {"author": "Hang Shi", "text": "

Now we come back to uniqueness of track id, global/local to bundle

", "time": "2023-03-30T05:04:57Z"}, {"author": "Christian Huitema", "text": "

We have to express a property of a track, \"can be bundled with set X\", but it does not follow that the property shall be part of the name. Encoding properties n names is not generally a good idea.

", "time": "2023-03-30T05:06:06Z"}, {"author": "Mo Zanaty", "text": "

@Robert Sparks , we all still twitch at any mention of \"bundle\".

", "time": "2023-03-30T05:06:55Z"}, {"author": "Magnus Westerlund", "text": "

So a very clear terminology section will be needed for general systems. I know how much work it took to create language for all the aspect of just RTP based conferencing which I want to remind resulted in https://datatracker.ietf.org/doc/rfc7656/

", "time": "2023-03-30T05:06:58Z"}, {"author": "Jonathan Lennox", "text": "

7656 was a lot of work, but fixed enormous amounts of people talking past each other.

", "time": "2023-03-30T05:08:06Z"}, {"author": "Joerg Ott", "text": "

Let's say I am media-stream receiver and there is video offered in multiple qualities, 1080p, 720p, 480p. Would each of this be one track and I would subscribe to at a given point in time according to what my receiver figured out its path can sustain? Or would his be a single track with some magic drop priorities to realize the different resolutions and thus receive bitrates and it is a relay's job to forward to me what matches my receive path capabilities?

", "time": "2023-03-30T05:08:54Z"}, {"author": "Spencer Dawkins", "text": "

Magnus Westerlund said:

\n
\n

So a very clear terminology section will be needed for general systems. I know how much work it took to create language for all the aspect of just RTP based conferencing which I want to remind resulted in https://datatracker.ietf.org/doc/rfc7656/

\n
\n

YES. A thousand times \"YES\".

", "time": "2023-03-30T05:08:55Z"}, {"author": "Cullen Jennings", "text": "

I think we the more we talk about things like bundle, the more we realize that might not be the best name

", "time": "2023-03-30T05:09:12Z"}, {"author": "Harald Alvestrand", "text": "

is it just me, or is there a lot of heavy lifting here that is relegated to the \"catalog\", which is not yet defined?

", "time": "2023-03-30T05:10:50Z"}, {"author": "Christian Huitema", "text": "

I am very dubious that Web Transport, implemented on top of QUIC, can do pooling without support for pooling inside QUIC.

", "time": "2023-03-30T05:10:51Z"}, {"author": "Kirill Pugin", "text": "

@Joerg - it will be multiple tracks

", "time": "2023-03-30T05:14:15Z"}, {"author": "Christian Huitema", "text": "

Basically, pooling replaces the HTTP \"urgency, incremental\" by \"pool-id, urgency, incremental\", with round robin by pool id, and urgency+icremental applied inside the pool. That could be done, but it is completely unproven.

", "time": "2023-03-30T05:15:47Z"}, {"author": "Joerg Ott", "text": "

@Kirill: Ok, that's what I thought. But then a relay should not decide on a per-subscribe request basis when to have a receiver start receiving media, i.e., at which group. Otherwise, you cannot guarantee splicing the groups together without gaps.

", "time": "2023-03-30T05:16:35Z"}, {"author": "Alan Frindell", "text": "

did Luke say the name of the track depends on the name of the bundle, or that it just had to be unique within the bundle

", "time": "2023-03-30T05:20:18Z"}, {"author": "Kirill Pugin", "text": "

just unique within a bundle

", "time": "2023-03-30T05:20:34Z"}, {"author": "Kirill Pugin", "text": "

so to subscribe to track you need to know the bundle and track id

", "time": "2023-03-30T05:20:47Z"}, {"author": "Christian Huitema", "text": "

We have to decided whether there is per track metadata stating the bundling properties, or whether not. Maybe that could be left out entirely, and just trust the client. If they require a track on a connection, so be it...

", "time": "2023-03-30T05:21:55Z"}, {"author": "Luke Curley", "text": "

to clarify @Christian Huitema , a bundle != connection

", "time": "2023-03-30T05:23:21Z"}, {"author": "Ted Hardie", "text": "

For those who will managing meetech and are wondering what blocked me doing it; a search dialog came up which blocked access to the dialog.

", "time": "2023-03-30T05:23:48Z"}, {"author": "Jonathan Lennox", "text": "

[How are people @'ing other people? Do you have to do it within Zulip rather than Meetecho?]

", "time": "2023-03-30T05:24:01Z"}, {"author": "Luke Curley", "text": "

the concept of bundle is mostly there to prevent somebody setting priority=max and starving all other customers over the CDN backbone

", "time": "2023-03-30T05:24:48Z"}, {"author": "Luke Curley", "text": "

it's a way of sandboxing whatever prioritization scheme is used

", "time": "2023-03-30T05:25:22Z"}, {"author": "Christian Huitema", "text": "

Yes, but the client is making the decision to request several tracks in the same session. So the client could very well verify and only request compatible tracks, based maybe on what teh catalog says.

", "time": "2023-03-30T05:25:54Z"}, {"author": "Luke Curley", "text": "

yeah but then that has to be propagated throughout the CDN backbone, based on untrusted clients

", "time": "2023-03-30T05:26:33Z"}, {"author": "Luke Curley", "text": "

you can imagine a cache poisoning attack

", "time": "2023-03-30T05:26:44Z"}, {"author": "Ted Hardie", "text": "

@Christian compatible in what sense? Compatible for a prioritization domain?

", "time": "2023-03-30T05:26:46Z"}, {"author": "Christian Huitema", "text": "

No. The relays can get that from the track's origin.

", "time": "2023-03-30T05:27:02Z"}, {"author": "Luke Curley", "text": "

yeah exactly, that's the bundle ID written on the wire

", "time": "2023-03-30T05:27:28Z"}, {"author": "Luke Curley", "text": "

you can only trust the upstream for bundling

", "time": "2023-03-30T05:27:43Z"}, {"author": "Christian Huitema", "text": "

Yes. Suppose that the tracks have rules, such as fraction use at which priority layer, or mapping from containers, or whatever. If they all apply the same mapping and use streams makrked with priorities, then stuff works naturally.

", "time": "2023-03-30T05:28:18Z"}, {"author": "Ali Begen", "text": "

I get the idea of deadline in a conferencing application but a deadline in a streaming scenario is not that useful.

", "time": "2023-03-30T05:28:37Z"}, {"author": "Kirill Pugin", "text": "

why deadline is defined by emitter?

", "time": "2023-03-30T05:29:18Z"}, {"author": "Christian Huitema", "text": "

If you mix deadlines with dependency, your scheduling algorithm becomes an interesting implementation of the knapsack algorithm...

", "time": "2023-03-30T05:29:33Z"}, {"author": "Adam Rice", "text": "

How is time synchronised between the source and the relay?

", "time": "2023-03-30T05:29:38Z"}, {"author": "Luke Curley", "text": "

@Christian Huitema that works if every broadcast can be trusted and uses the same scheme

", "time": "2023-03-30T05:29:48Z"}, {"author": "Christian Huitema", "text": "

Yes. But that transforms a transport issue with a trust issue, for which we have solutions.

", "time": "2023-03-30T05:30:24Z"}, {"author": "Luke Curley", "text": "

but imagine a broadcast with 60 objects per second and layered encoding, versus a broadcast with a single OBJECT per group

", "time": "2023-03-30T05:30:32Z"}, {"author": "Luke Curley", "text": "

if the relay collapses the priority domain, then you get very unfair bandwidth allocation

", "time": "2023-03-30T05:30:58Z"}, {"author": "Christian Huitema", "text": "

Luke, we also get issues if each track is its own domain, because then you can only do round robin, and that's not exactly optimal

", "time": "2023-03-30T05:32:09Z"}, {"author": "Luke Curley", "text": "

yeah exactly, that's why bundles can contain multiple tracks

", "time": "2023-03-30T05:33:04Z"}, {"author": "Luke Curley", "text": "

if the producer intends tracks to be prioritized against each other

", "time": "2023-03-30T05:33:32Z"}, {"author": "Luke Curley", "text": "

you can think of it like an opt-in

", "time": "2023-03-30T05:33:47Z"}, {"author": "Christian Huitema", "text": "

You could conceive a syndicate of producers agreeing on compatible prioritizations...

", "time": "2023-03-30T05:34:10Z"}, {"author": "Jordi Cenzano", "text": "

Also I would really careful on using epoch times, most of the times we do NOT have any guarantees about the time sync of all the elements involved (client, relay, perhaps player)

", "time": "2023-03-30T05:34:25Z"}, {"author": "Luke Curley", "text": "

@Christian Huitema that's feasible too

", "time": "2023-03-30T05:34:46Z"}, {"author": "Kirill Pugin", "text": "

not most of the time we never have these guarantees

", "time": "2023-03-30T05:34:55Z"}, {"author": "Christian Huitema", "text": "

In which case we just need a verifable property, \"track X by producer Y follows priority scheme P\".

", "time": "2023-03-30T05:35:28Z"}, {"author": "Jordi Cenzano", "text": "
\n

not most of the time we never have these guarantees

\n
\n

Perhaps in a lab enviroment we do ;-)

", "time": "2023-03-30T05:35:40Z"}, {"author": "Kirill Pugin", "text": "

yeah, sure, I mean in real world :D

", "time": "2023-03-30T05:35:57Z"}]