Internet Engineering Task Force 92, 25 Mar 2015

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF92-PAYLOAD-20150325-1520

Description

PAYLOAD meeting session at IETF92
2015/03/25 1520

A

B

Are girls often after.

C

Your instance, this cool this would be crystal- will have time for that. That may exist in the first room shortly: okay,.

C

Okay, these are specific reason why you want to be first. Oh you.

A

Are gonna call the time here, No, okay,.

C

Okay, can you push the door Oh sober.

A

C

C

Well, good afternoon, everybody this is the payload working group trying to see that in making typos having to do all working groups and our lives here and me chairs the note. Well, you should be aware of what they say about IPR known I PR so you've seen it already for a couple of ten times. It's Wednesday today agenda the payload status update. Then we have the payload format for the map codec, which I will present, because the authors are not here, there may be in the Java room, Jonathan.

C

Okay and we have the SPT 291 ancillary data, and we have the thick payload and the vp9, which will be presented by Justin, not by Jonathan, Wright least I got the presentations also from him. We have one load taker whose bow anyone I would. I would like to get a second mode taker anyone volunteers to take note just action items.

C

Ok, Stefan was good enough to do that. So, hmm okay, no problem, so the document status we didn't have any RFC published. We have the RTP how to in the ethical q, I think it's mixing for some reference we have in is G processing the g7 1100 520 55, there's a 05 version that was uploaded today and I hope it will open close the open issues after the discussion that they had aged 0 to 60 507.

C

The some discussion with the ad there was discussion with the ad reach out on the mailing list. The previous ad. We have the new ID, but there were some open issues ever think that they're going to be a revision. What's the short Stefan.

A

So Stefan Vega one of the co-authors of this thing. Yes, there will be a revision that a provision is of editorial nature. Steak head Tori changes, but they are editorial. Oh okay,.

C

A

On top of that, through some discussion in the IMT see which trickled back to us and so on, but more mostly, we came up with one other change: that's potentially not editorial, and that has to do with this decoding order. Number presence being signaled through the trend through the mode of it through the transmission mode, Mrs, T, srst and mr empty.

A

It may be worth to be couple this coupling and we'll have a proposal doing that in the very near future. Okay,.

C

A

That's not an editorial change, so I guess until.

C

You have to come back to the word. You look for.

A

My first one to the working group with us: okay,.

C

A

I present, our estimation is that it will, you know if I wouldn't tell you most of you would probably not even seen it, because the changes are really really really minor and eligible, but it's still a non editorial change. The world come back, but it's not going to be. There's not going to be a need for foreign. You know to review all the whatever 90 pages of what is now in over it by working angles. Okay,.

C

So I have to ping Kuechly to update the documents. Yes, okay,.

A

I'll, do that? Okay, your browser! Thank you. Okay,.

A

So quick question on that: that's right now has been public and it's just in the upright state. I think.

C

A

Back there does it need to be back in the you weren't all the way back into the into that into.

C

The working group back into the working group doesn't matter because what we'll do he will publish a new will take, it will make a last call and we'll say: will free it back to you to publication. Whatever.

A

The stadium we're going to we're going to working good man.

C

So leave it as an ad it's under ad review and that's that's what what will keep it for the moment. Okay,.

A

There will be a new version anyway. There.

C

Will be a new version here or I need to do anything yeah yeah? We will keep you up to date. Okay, the opposite is in is G last call and what's the status of the vp8 forgot, this.

A

C

Idf last Korea, sorry about that the and the vp8 14.

C

As far as the authors know, the vp8 document is ready for anything they work. Your chest want to do to it. Yeah, because.

A

The meeting one new version comment at this one, so.

C

Color with rottier they were covered from about the the beverage and all these things I think there was some comments on the list which I didn't feel like they're all fixed okay. So we will try to see how to finish them. Okay, the milestones I, don't think I don't want to go to that. It's just there there. We still have these documents that there's some some codecs that are not being progressing.

C

We'll have to go back to those people and see if they intend to do what they want to do like the Bluetooth, the preamble, one, that's that's early and I need to discuss, and because we are, the will was our work and I think that's the MVC. What the MVC yeah I don't know what's happening. Is it you Cree Stefan? You look every see.

A

D

Manga regarding the NBC.

A

Codec embassy, for those who are not aware, is a dialect of h.264 I.

A

Don't know for sure what I do know with some reasonable certainty is that NBC is not a particularly popular forward over the wire right. It's it's one.

C

A

Things which you find on DVDs occasionally but not anywhere so while I can't really speak for you agree on this one. Okay,.

C

A

Be surprised that some of us have that he or other people would have lost interesting.

C

Okay, so I'll ask I'll: ask you about it? No problem, okay. So this is the chair status, not the next presentation.

C

It's the we have the mmm Emil p payload format.

C

So this is an individual draft. It's a payload for the map, codec, it's a M LP and there's an Emmy LPE voice, codec! It's it said, letter standard as see here. What is the where it's being used.

C

It was submitted. The draft was submitted in previous versions of this.

C

Now it has, it has 3 mods this 612 hundred and twenty four hundred room. This is not that important. Now as I read it, but has anyone else read this document? Are you ready to okay? So they have some I sentiment, a response on the list on some of these like they have they. The thing is that the right thing is that, first of all that they expect to be symmetrical on both sides. So if you offer 12 to 2 2400, 600 and 1200, it means that you can dynamically switch between them.

C

The preferred one is the first one and the response, the the the offer and the response can select what he wants and he's dictate the order, and it needs to be symmetric. It means that if I didn't ask, why I mean that's the way it's used? Okay, they want it to be symmetric, so I didn't care, it's something. You really can ask why it's need to be symmetric and inside an RTP.

C

It cited it in an RTP packet that can be multiple friends, but they need all of them to be at this from the same rate, you cannot switch dynamically inside an RTP packets. That's the way it's got its build.

C

The other things there too, to note is that they are requesting to have the payload subtype name will be. They have em ELP and ELP e, and they have also ones that are with the beach right inside, like the 1200, the 600 and I said we don't usually like it, because we prefer to advocate as a parameter which they're having the definition, but they said it's already being used currently, so they want just to capture the weight is being used. Instead of of changing look at it, send the comment whatever to do.

C

Is this: since it's a it's a it's a regular audio codec payload specification will ask for a milestone for the adoption and try to to work would be so that's the plan for that. Any comment of that.

C

E

Jonathan I Simon have a little scared. This unity. We asked yesterday over stamping they've already done, rather than applying our best practices to how we think it should be done. So hopefully I mean if they are they'll, be unhappy. If we change how it works, then I don't think that'll be a bit outcome for anyone. So I mean I. Think that I mean what way you were saying is like you know, we are a foundation or does this way, so we can't change that I. Think.

C

They only the only point, it's about think. The only point is about the payload subtype limbs and we can discuss this. I mean I'm. Just saying what I got. I sent a comment about it. I it's on the mailing list. We sent a comment. We got a response, we can still discuss this, but I mean at least it's a lot. It's an RTP payload for an audio codec search. We save this scope of this working group and that's what we are doing it. So that's a problem. The issue is whether we can progress it Aziza.

C

We wanted to make changes and that they'll.

E

Have to make a decision.

E

Okay: okay, okay,.

C

So so it will be good sent also in this on the mailing list and we'll take into consideration before adopt adopting it as a as a work item in the working good to see that they are okay with essential. We have some agreement about how to proceed on that part. Okay, okay, so think this is the. This is the information about this one okay, so the next presentation we have is the.

C

The SF this one, the theory, search.

C

Note I'll present from the powerpoint, like you asked, so I could maybe be.

C

Well, there's a flexible, fax, sorry.

E

C

Conclusion for the Prince feran station is that will first let some ppl response on the list on the way it's constructed and they have any concerns about it and then, after we finish that we can go to the next step of having a milestone, just to be sure that they are flexible with accepting that, if you bring it here chain, it means that they may need to do some changes and how flexibilities yes, yeah just wants to remark.

A

Sure I keep trying to just me remind you, you know basically highly charged on specification, requires I mean they could go to another body, get specified and.

F

A

F

Effectively be asked to rubber stamping so.

C

Of course, what.

A

I'm, saying is: don't apply your criteria to rush if it does harm and obviously confine yeah.

C

But but at least I wanted to have some I mean some little discussion before we go to the next step and if some people are concerned, I already said look at what the current exchange it was on the list and if you're, if you have more concerns than we can try and see, and then we can progress that okay, so.

F

Discuss Cynthia t91 ancillary data in RTP.

F

Okay, so simply what is this society of motion picture and television engineers? It's the standards, development organization for professional video applications and today, professional video inside the broadcast plant is carrying something called a serial digital interface or SDI. This is either a megabit per second for SD or 1.48 gigabit per second for HD serial digital signal. That's carried over 75 ohm, coaxial cables.

F

When these technologies were developed in 1990s, they were very exciting and they are pushing the edges of technology at the time today, when we have 10 gigabit 40 given hundred Gigabit Ethernet, it seems like maybe it's time to go, send it send video over over IP.

F

So in order to do this, we need to start taking a look at how we've been sending video of our SDI, breaking it down into its individual components and mapping these into an RTP infrastructure, so that we do these standards compliant / RP. So some of these elements include many rfcs which the IETF is already developed. One is RFC 4175 I, believe in mr. Collins developed this for the carriage of uncompressed video over IP 10 years ago works just fine. It seems to meet a lot of our needs.

F

Another one is RFC 3190 the character, 24-bit uncompressed audio PCM, / RTP. This has just been adopted by the audio engineering society amias as their 80s 67 standard and that's gaining ground in the audio world and we're bring them to carry audio or video world as well, potentially 1588 from the I Triple E. The precision Time Protocol is a great way to do the timing that we need in order to time our systems make sure the audio/video synchronization and that we can do frame accurate switching.

F

This leaves a little bit of something called ancillary data and thus the internet draft bored. So the serial digital interface mocks at several different areas. Boop you've got the active video. This is what you actually see on your television at night. This is where the actual video is carried within the SDI box.

F

There is a vertical ancillary area and boom a horizontal ancillary area. These are areas outside of the active video which is used for carrying the ancillary data. They are historical inheritors.

A

F

Vertical blanking and the horizontal.

A

F

Period is back when we all had CRTs, which of course not. Unless do anything.

C

F

So what is ancillary data? Its non video data, which is embedded in the STI outside of the active video area? It's typically 10-bit words and most of our applications today, unfortunately not matching LT octet boundaries, it's identified by a pair of the data ID and the secondary data ID and as d id MSD ID cynthy terminology it's registered through cynthy, including the active format descriptor, which helps us with. We go from 16 by 9 representations in beyond, convert before by three representations to make sure important things don't get cut off.

F

The screen can scan data for movies study. One of four messages for use in ad insertion triggers DVD vbi photos of your lyric electra, PVR ancillary timecode, and also the closed captions of both and Celia 708 and cea 608 closed caption, normally dia, subtly registry. So closed captions are very important. You know, frankly, broadcasting only one thing: we need to have all the air, the u.s. that's closed. Captioning, that's the one thing we can get fined for not having. We can have no video, no audio. Nobody cares.

F

You get up close captioning in your trouble, so we also have a submission at the bottom.

B

There's a URL the.

F

Simply registry of ancillary data types, so the goals for this payload are to be able to carry ancillary data, whether it's from standard definition or high definition, video sources being in the vertical ancillary area or horizontal minutes Larry area, either in the Y or luminance channel or the sea or chroma channel, and also have sufficient information provided such that we can recreate in the decoder we're in the STI raster. The original ancillary data came from this is for it is important for some specific legacy uses in the future. We probably won't need this.

F

As we move to all-ip infrastructures, we don't care where in SD I box, it came from we're all going into and luckily simply re has a standard which has a data model for kit for handling all these requirements.

F

Cynthy 2038, which is the carriage of ancillary data in MPEG transport stream, and this is the data model again topics down roots in the chromatids primitives channel, the line number and the SDI its downfall, horizontal offset is found on the d ID and the SD ID the street, the actual user data wards of the ancillary data and check some word of the ancillary data packet.

F

So in the proposed RTP payload for the arts of the timestamp, we intend he's a sampling instant of the frailer field, to which the ancillary data in the arts of you packet belongs, and this is cribbing from RFC 4175, the RTP record it with either last ancillary data found in a frame or field again.

F

This is driven from RC 3497 RC 4175 we'd like to use it 16-bit, extended sequence number, that's also found in RFC 4175 we like to have the look of the payload and octet we'd like to have the accounts of the ancillary number of ancillary data packets carried in the RTP payload and finally like to have some octet aligning bits at the end of every ancillary data packet information so that the beginning of the next one is octet aligned to make a processing just a little bit easier.

F

Okay, we must be back to pc conversion problem. You get the point on top and black is typical. Rtp payload orange we've got the ancillary data hearts of the table. Header at the extended sequence number the length knock text, the counts of the ancillary number of ancillary data packets. That which are in the payload and then blue is the ancillary data packet itself with weather growing or luminance line.

F

Number horizontal, offset eid best eid, the data count and the user data words of the ancillary packet to check some word and then some bits in green if needed, or october lineman before the beginning of the next ancillary data packet.

F

So for a media type, we're suggesting it be under video because it is highly associated with the use of video and that it be cynthia smpte 291, which is the name of the standard for ancillary data packets and serve. You we'd like to include an optional sdp format, parameter for identifying it flow is made up exclusively of a particular ancillary data type as a FN PT mmq down here here, indicating the div 0 x, 61, st id0 XO, to for example. This is 608 closed captions in this particular.

F

This allows for the easy identification, the ancillary data type of flow. If you user, should want to yeah I think this phone yeah, you know- I'm, not I, was just thinking about that. Big and I need to take a look at the draft to make sure that we can have multiples.

F

Ok and that's it so we'd like to you know, put this to the working group happy. It's ok to have it be taken up by the WG. Ok,.

C

Anyone read this comments on that.

C

Ok, so hmm I said it did anyone read it or have any comments on that.

C

Ok, you ready Colin, also yeah, okay, so.

E

Typically, what city were created to carry it yesterday, usefulness is it sit for GSP else and.

F

We're not spelling that out at this internet.

E

No no I mean about giving in these days. People are using it in I.

F

You know the broadcast industry is getting its head kind of wrapped around how that's going to work right now. I think we, in the eve of the AES radius of 70 kind of innovating they'd like to sip, although they've found that more challenging than I think they first expected so I was suspicion. It's going to end up being like web services.

E

I'm a most wondering whether there is declarative or offer ass or SUV, so whether it's you can have my directional flows and if so, can they have different values for them give these different outlets negotiated or is it simply a claritin where you say this is what it is: ticularly yeah.

F

I think in I think, in the draft we open, we left open the opportunity for offer answering only able to handle a particular ancillary data type. I don't know whether that's going to be a real use case or not. I figured we keep it over the dose all.

A

Think about beautiful one.

E

Better yeah sure that makes it so the other question I have to assume this would typically be tightly associated with it. Actually, no video overall relatives will be together and though he'll welcome to be constant right. Yes, so there's the usual complexity. Where I'm, like this guy sitting behind me, for the fact that you're not allowed to use the same archetype steps, because the have traditionally randomized you will have to have various complexity exception.

E

We should go where the where you'll have the computers have the of the direction signaling mtv timestamps and the library so which, given? How can the time step locked? Everything? Is it's probably not that the end of the world for you, but should look at the make sure you have all of that birthday and cozy.

F

And I don't know just answer that we are taking a look at how you use rtcp sender reports and also how sdp files can contain. A linkage between a particular are to be timestamp. And we have half of a particular piece of be.

E

Growing- and you probably also want to have some sort of stp grouping that says hey, this- could only wanted this probably never go.

C

Together, is it tiger to the clock, source work.

E

I think you say good things go together. Yeah.

F

Yeah we get we particular cat box, aureus, RFC yeah, and you know we're not necessarily ruling out the use of the RTP header extensions. I think RFC 6051, that has a 64-bit time staying up an endurance tension, but I was suspicion like abs, 67, we're going to try to stick with the hawk stores in the stp as well as this.

D

Clue I concocted sigh, deny all responsibility of anything to do with this setup. Yes, sir I absolutely think we should do do this.

D

Looking the draft I notice, you've copied the extended sequence number thing from 4175, please for 4175 that's needed, because the secrets numbers were out too quickly. It wasn't clear to me if you had enough packets in this format that you actually needed the extended series number equally I can see. You may want it just to make things easier, because it's the same as the rest of the vid. So that's something to perhaps look at the other thing which I think you should take.

D

A careful look at when you think about bringing this into the warning is nothing to bring things in along with the unrest. Video format, especially at school experience, is that the rtcp packet rate is some surprisingly high for these types of applications and the video that the counters wrap very quickly. So it may well be that a lot of ER tcp extensions, don't quite work when you're sending rtcp at the fair sentencing thank its personnel.

F

When you see our new CRTC extensions in, you are hot sauce or Olympic. So.

D

By anyone in particular I know, we have a bunch of reporting extent of your course enhancer, most of which are I suspect, assuming a much lower packet rate. You may well find that there are hidden assumptions and packing crates at some of those and.

F

With regard to the extent the header, that's our the extended time stamp, I went back and forth in my head about that. I figured it's hard to go along with it, because we don't know exactly how much ancillary data people really wanna push push broom, it's just all the ones for unpacking per frame, maybe doesn't matter for all fish out of your packets per frame. I might be more than issue.

E

Relay across fitness on general because they were a bit about the use of media tech, video above early video because officially associated with video, we have other examples of metadata article formats and what media types do they use. Speaking for myself, I think we have had some, which typically have any milk really like fat, will use typically use video for protecting video I'm, not sure we have meditative proper as opposed to a pair of clothes and there's, like you know, cursor position uses video things like that. So I think.

A

E

So precedent for this it could also be application, but I mean I, don't know if we've had application as an Argentine one yeah I have the high score article /, arka p, so and.

F

I had a problem with this too I was looking at application. Kicking you know. The definition about location does not smell.

D

F

A

D

We used video for the 3gpp time text, which is kind of reverse closed captioning, I. Think in press my.

C

D

Against HQ for.

C

It so politically, because for being just.

D

Make sure I sleep a.

C

Lot easier, if this week for a camera crew provide any great I, don't know how to fund its about.

F

A table someone smile about Romanian, detox, ancillary, so.

A

I am planning how to all this town. Now you plan to put this in the same are to be session, where the other did.

F

Right now, audio and video, again sorry Gator and separate RTP sessions.

F

If that sounds like at that well and then what reason you want to be able to mix and match, for instance, if we have an English version of a German version or french version of the audio, and we have to make a couple of different distributions of this device. Cable will pull the RTP German over here with this video and then pull the French for that videos. I think the flexibility makes it reasonable that we have these and separate our community sessions. This isn't really to be bundled together for end users.

F

This is for production, so.

C

A

C

Basically, thank you, then,.

A

Application is not the problem, it's the old full story about the a.m line, having the top level hi.

E

John flags are further relate for us. He said EJ's approaches the corporate farming box and religious about to have a video data which might still wanna have this type of vanilla burger. If I had 10. You basically just said about the you know some way of indicating which closed captions are written, which in mind there's the outside available list. Now.

F

Be outside the scope again, the goal here is to carry cynthy rich tree ancillary data in the city, registration, all those upper level, closed, captions, notnot languages, and you know I think over time we may as an indicator broadcast industry. We may want to take a look at breaking the closed captions out the separate time text streams. Unfortunately, we're not at a point where we can really do that. There's still too much legacy built around is 608 the 708 closed-captioned.

A

Ben campbell jonathan made me think of something that probably not something need to worry about right now, but if you've been mentioning the use case of identifying springs by different language or closed captioning by different language, there is a kind of pre buff effort going on right now and ryone, which includes doing some language tagging for our TV screens. It's really more designed for interactive telephony type, in fact applications, but there may be some applicability here, at least when it comes to how you identify these things, just something to keep in mind. Okay,.

F

Yeah, that's that's great information. What one can ways you have a closed caption. We are very strict. The regulations are very certain, of course, by the FCC. Now, I'm sure shortly will be enforced by telecommunication regulators around the world. So we do have to be careful about exactly what you know how power carrying labeling these things, but I people continue to watch that sometimes.

E

I, just hooked up to a price question HQ floor, which is the prime account controller, does reciprocation, so we have precedent, graphic application, mimetypes for so.

F

C

It's a phone camera control.

A

While 224 has something to do with video in the sense that it controls the camera, 224 is something that goes from the receiving end to the sending end. Only so it's a totally legitimate use to control the camera promises. It took you to use to 24 to send camera control information from a system that has never seen a camera in its life right, so I think that's a difference, because this stuff is really clearly basically, as I was here. It.

C

Was just too it was just to say that there's there's a there's, a tailored format, RTP payload, which is the application? That's all! Ok, that's! Oh there's one already.

E

I think because, as you actually getting it's not just metadata, is that I made an actually has some notion of where it is inside, of the cover video frame. So.

C

It's my penis. It's.

E

Certainly an odd video format, but you know it is in some sense it goes inside. You know the you know the whatever the bsci.

F

D

E

Sense, video in essence on.

D

Competence I've moves that we agree that this interest in this work, okay,.

C

Okay, so I can see there's there's interest in this work and it's in line with what we are doing so will I can ask if people thinks that it's for the working group to work on that who agrees that we should work on this say in this working group should raise their hand.

C

Okay, anyone objecting to that should raise their end. Ok, so we will go to the list of tool to get the approval on that, and then we will continue with this with the process. Okay, okay,.

B

C

We will need to have a milestone on that after that, okay, so no you got the action item. Ok, so the next one and.

C

We'll ask you to: are you going to leave some time also for.

C

Because otherwise, Justin would be will not like it so go to this one.

B

On this presentation of mine- perhaps because maybe next time this so.

C

Just I uploaded the new revision about half an hour ago, so we love raffle now. So, if you had earlier, then you should try to look again at the side. Just ok! You sorry! Yes,.

B

Sir, the exception is looking document and we just submitted 00 Asian. We got some comments from Magnus. Thank you we'll get to those comments soon.

B

So there are a couple of poppin issues that have been carried over from Honolulu. The first one was how to be associated the repair package, streams to the source, packaging screams, and this I think in Honolulu with the music. There was no final decision, so since it's called flex back, we decided that we would go with the dynamic Association would be armed and keep the stb very simple. Thus we won't have a good music again. Perhaps just htp direct rate can vary over so.

A

So there's no association in.

B

Stp we just declare them in, as the example shows at the bottom. I think one question to ask right now: is the people are happy with this step, if not doing any another payload type or SSRC declaration.

E

B

Okay, so next slide place, then so so the current problem is perhaps as simple as that. You have effect RTP stream and you have source streams from different media sources, so you want me to even we do have to cram when the receiver received it. There is no actual actually currently way of associating them. So there are several.

A

Ways we can do this.

B

They're exemplified in the next couple of.

E

Slides also thanks.

B

To Peter Thatcher, who make these slides than the next four in Senegal next slide, please so left for bias, SRC, which is basically putting and it's SRC in the effect stream in same there is a reference to certain source, RTP stream them that's as simple as protection in having severe on stupid media source. The next time the next one was hard to fit using em, ideally powered. How would we do this? This is kind of difficult, especially if there is.

B

There are multiple sources of our TV screens for the media source, so so their work, a 1, a 2 then made a would not work.

A

But if there was only a.

B

Single week via source, like just read a 1, then this will be easy and it would be sufficient next slide. So there's also proposal of doing an RTP source ID, where it would be as simple as adding another identifier to to both the source cream and the repair stream, where both of them have the identifier to, and you would be able to receive when you receive that you will be able to map them next slide. So.

A

Something like this, you.

B

Could do it without the msid, just RS I mean by itself. So in this case you say to me to and before- and this is another way of signaling this so I guess the question next.

A

B

The pic so this would look like in the article hair extensions as unexciting one for the sauce streams, which would be on every packet and the facts. Teamwork. Whenever the practical drive would have the same, so you will be able to associate this game if you are multiple streams, the same thing you are correct, so so I guess the open question for this first issue is which of the three mechanisms or four mechanisms are ideal of just using SSRC using made + RS, ID or just RS ID.

E

I think that doing s Circe's explicitly as a developer, a terrible idea, because I mean all sorts of things that expected all the recording stuff and everything else is out of scope, whispered and, additionally, I think we also need to solve this problem for our TX, which isn't already defined, feel a little workout. So whatever we do should be better extensions or other better data. So I think that we should say we're not going to us our seeds and we should further rescue worker with scope, which is either injector.

B

Didn't lose his job to EDX, know.

E

B

E

Easy, like the flexpay itself, let's go for this, but.

C

But I you associate.

E

Both affect & RPX with source dreams, so the other side equals ours.

B

E

Or something like that yeah so either I'm using gravity exactly where goes to bed little social creatures.

D

Colin Perkins I. I think I got up stick something very similar. I.

A

Don't think the.

C

D

Works of a sudden, you need a separate suppose identifier to do this. I would suggest that probably the right way to do that is in which could be would be signaled an sdp or was in other extensions like this.

D

So basically you.

E

Want to generic.

D

Mechanisms firm saying these streams get together, which is not, if you see specific.

D

The other side.

B

Eve would be, we don't have to worry about. Maybe.

D

Mid already has a different purpose. You use it for its existing focus. If you need another end into Earth, to signal best that you can.

D

Okay, they say so.

E

Filter point I'd make: is that finishing that doesn't mean to block this draft? Is this product would work fine, they're, not wiggle our business in the modern world, with just sending them on separate unbundled? My ins abusing.

E

Whatever connect, everything is cool, okay, so I think there is a working case. Those blocked with do things where this is usable without any dividing any.

C

Okay, so I think to conclude that what I understand is I think what Jonathan and I agree with John Kahn is that you have a solution when it's in a separate. This one is, if you want to to have them in the same m line, or also you need, maybe in the case where you have one having to support even one fax stream for two separate em lines, or things like that. You need some identifiers for that, then, in this case, you need to the other side.

C

I think people agree or did the RSI do some. Some generic token is needed, but I think it needs to be presented in their music right. Does, okay, that it's not to pay it sets at a payload. Okay, then you just you will have it here is an example that you have to define the generic this generic one in a different working groups 0 through that's a conclusion.

B

C

Just to find this viable, but you have a solution. That's currently working using the the separate headline, yeah.

D

Contact as a procedural thing to make this go quickly, I would write this payload format, just on the assumption that everything is sense of separate Airlines and you use the grouping that way. This can be done very quickly. I would then in another working group, do this other stuff and then that draft then refer to this and r-tx and anything else where it useful I'm waiting UT couple one, and so that would you you're.

E

D

E

D

It's just a separate thing which says you: this could be used, for example, but it would normally update.

B

C

Don't need to go with you just see, you say when you use it with that, that's how you use it in the other draft. Okay. Yes, that's.

D

How you do this because they get a draft, could be defining in generic mechanism, make it say, go for example, you could use this.

C

With ya transmission.

D

You could use it with a flexible fat, etc, etc, etc. Grouper.

E

This graph says you could use as a proving framework or any other mechanism for associating sorcery with American citizens drop defined in the future. Okay.

E

C

So we have such a drop.

B

A

B

Want to make sure that begins is right, because perhaps you don't have been two more cycles: okay,.

C

I can help it was this one who to do that. That's a problem. Okay,.

B

C

B

Other open issues which were I think simple to solve the first one is we have mm, which are in the rdp headers or in the fact, our DB others and.

G

B

N equal to 0 and in equal to one book mean that's a row back.

B

We wanted to just overload this the slightly by saying that if you want to just send row fact, then you said it as any 20 and if you think that you're going to send both raw and contracting set this as n equal to 1, so that the receiver has an idea that, with n equal to 0 when it receives the packet, that's the end of the row and it does happen, wait for more packets and maybe send a retransmission packet of more than one package missing with n equal to 1.

B

It knows that that it would anticipate more back. Packets will be coming, which would be callin fact and therefore be able to repair like worst loss of packet. So I guess.

A

The question is.

B

Is it ok because I think last time when we would discuss this, we had second bullet, and now we want to just sufficiently slightly optimized for a case where we good, we good guess. If the, if there's a loss, then is it recoverable or not, with the anticipation of more effect packets to come.

A

Mostly I just understand spawn, but they well. The other changes. I think you got to point out later is that there's a used to be SDP level description of whether or not this was 1d or 2d one day in early 2d or whatnot interleaved, and this is basically allowing dynamic in RTP description of that.

A

So again, we're trying to divorce ourselves of anything in stp will try to make everything flexible and dynamic at the sender's discretion in RTP, and so by doing this we can eliminate the stp part and, and the RTP sender can dynamically signal whether or not to D is coming. Here's one D but to D is coming.

E

Actually, like most motivation, because if your instance in one of the cases that definitely care about is, if you're applying the fact to video and video bandwidth and swing around so widely, you want to be dynamically. Changing your structure. You know like, depending on whether it's decided to 76 megabits or 50 kilobits,.

C

Okay, so so I understand people agree with this proposal. Anyone objecting to this proposal, so so, okay, so I. I think that's! That's! Okay!.

B

So we have couple of operations based basically with this in SD, be the first ones the type of protections. If you go with the flexible one that we could do away with this top pitch, which was 0 1 and 2, to indicate that it's interleaved on, interviewed and and or both if you do it dynamically in RTP, that there is less use of signal exist in stp. However, I believe there might be some cases where it may not. They might be static, so if they want to do it is nice to be so.

B

What you would like to do is remove it, but you open to suggestion if you should keep it or not, so any feedback on that. If, if not, we will remove it.

E

That means there anything picture that you would do.

C

If you were to negotiate that that's a question.

E

Amazing, not gonna, basically taking out as to be movable in negotiating whatever the sender wants to do is what senator wants to do, but.

A

Leaving knobs like this and stp encourage people to do modes and then get being flexible and stuff. I would personally just try to bake it in you can't get rid of the stp cold so.

B

Just remove that as well, so the last open issue is nds TV. This is again the length of the depth of the the fixed 2d protection. The main point was that conveyed the M&N. We can only go to 56 cross 256 packets, which is quite many, so we could remove it and be happy and just look at the constraint that that one comes out a predominately 56 well.

A

Actually, you say that it's too is 250 cc is too large, but actually for high bitrate applications.

C

It's not really that much I mean.

A

If you consider uncompressed video, for example, 256 package is not.

E

A

E

So I mean I think that if we need fact for identity, we're going to have a modified, primitive as easily spent a sequence on the family, that I think that you know if you're, not when you forgive this attentive use case of this is for things of use or dressing team can see those numbers. I think that you know being able to protect no more than entire signals. Number roll over is entirely fun. Okay, I.

A

Think also it it's not really about the hackett rated, the original videos that length of the protection period. You need speed box go by the way, so, if you're sending even with uncompressed video, you might not need more than this.

B

So so the proposal is to remove it. I'm here no objection to its ok: let's remove it, monks, okay, so next steps it just gotta accept it as a working group document that some work still to be done. We would appreciate feedback, especially interviews, so that we can fix this quickly. Going forward, perhaps be done with it. Okay,.

C

Okay, so you need to revise pressed of all the decision that was made here and then we try to see how to continue and finish is that if there are no comment, okay and we'll talk about the about the thing about the sdp, I think like what colin suggested that you should just use the to.

B

Make it simpler, yeah, yeah, yeah.

C

So we can be done with that. Okay, thank you.

C

But the last one we have a couple of more minutes for Justin.

G

All right craft with the poultry video Jonathan gaining a video on Big John Taylor format. So if you cannot click overview, much better compression than vp8 or the previous generation with codecs forty percent, our team quality of forty percent pure bits, the Codex, much more flexible and ask for eight reference.

G

Pictures in each reference picture can be a different resolution than the frame that's actually being encoded, does off sampling or down seven or whatever, to basically go from the previous frame reference frame to their for picture, and this allows or basically scalability that you can have a current picture at 720p, be predicted from a base layer at 360p, and you know that frame that should be cut it either from the base layer and a spatial sense that also from previous temporal frame and be pretty good for multiple reference. Pictures and.

E

G

E

G

D

With this, we can.

G

Do full scalability and dad one of things that comes along with this is that, if we're basically supporting star sending a stream a middle box, that's aware of scalability can basically drop out higher layers in these higher layers can be removed without affecting the integrity of the lower layer next slide. So we introduce some terminology into in this document, basically defining two types of things: a superframe and a layer frame and the superframe our mayor fraim is this is saying for a given. You know special size.

G

This is base one street one frame captured once in a time and the superframe is the combo this across all spatial layers. So this basically just makes it easier to refer to much other concepts. It's referred to frequently in the document next time, so the packetization covers a bunch of things. One things we need to talk about is: if is a frame actually something you know what layer is it in, but especially in temporarily, what's this prediction structure or we're coming from?

G

Is this thing discardable much the bits that are talked about in one of the presentations yesterday about basically having frame markings in RTP header extension are also sir presence here. You know it indicates winners of packets, the start of the frame packets at the end of a frame, whether it's a keyframe, whether the frame is discarded and in addition, it also has other sort of things about. You know take a temporal and spatial scalability um there is. There are two modes of how the scalability can be indicated.

G

You can either say here are all the pictures that this particular frame depends on or you can sort of say well, there's a sort of pre de there there's gonna be a special pattern that these frames are going to follow for temporal structure. You might say I'm going to send temporal layer, 0 first and then two and then one and two and always moving the structure and so based on the actual sort of frame number.

G

You can figure out kind of basically implicitly determine what temporal reference pictures are going to be used, and that makes a there's much redundancy in each specific packet, and that makes things a bit more efficient.

G

One of the other things that receivers always need to know is that you know when a picture comes in, they might get the basin layer first and then I could be enhancement layer, but if the code needs no, do I omit the picture as soon as the bass players complete or do I wait for all the layers to be present, and so we have this thing called the scalability structure that sent early in the stream.

G

We basically know how many layers are currently being transmitted, and if you know that there is a base in an enhanced, you know that if you're only decoding the base so far, you need to wait for the enhancement to also come in before you basically admit the picture from the decoder.

C

Can it be changed? Yes,.

A

G

Excited so this is the flexible versus not flexible mode, the non flexible mode. Basically, you send the structure. We thought group of frames and in this thing basically sending here are the dependency orders.

G

This first up is specified the scalability structure, and then you can refer to this with a more efficient coding of the packetization and allows you to determine which are there actual temporal references, you're flexible mode. You can actually basically say here are three excesses of reference pictures that this thing depends upon and it's more flexible as the name implies, but it's also slightly more verbose in terms of how many bits requires the packetization header. So here's, the overall big picture of the techno sation, looks just like any other.

G

Codec I've got this chalo descriptor here next slide, and so here's the actual Caleb scripture for the flexible mode, there's much things that we present, but.

D

There was listen.

G

Lee 80 and ass layer indication. That's president at this elbit is said: we've also got the other bits. I talked about earlier, the beat bit, which is the beginning bit of a frame or being a packet of frame II for the in packet, in a frame as well as I, for you know key and some other things as well, but you can see that this could actually be several bites for the actual packet hair next slide and in the non flexible mode.

G

You know it's Trump down a little bit more, and so the actual later thing is determined from the gof index. The special mayor is still the sky explicitly next slide, and so you don't need to send all these bits all the time if you're just doing hi poppy stream, you can do a just a single bite, the technical bite and you don't want to use them to be in the.

A

G

You're doing just temporal and not spatial, then you have a special skill, bony structure, the SS structure, but you only fill in later 04 special scalability, but the actual temporal narrowly indicated through the goth index and if you're doing everything, then you'll fill in the s and the D bits as well. I'd indicate that this is you know, layer, 0, mayor, one spatial risky bet on overall I'd. You know this can be good fairly efficiently. You know in you, don't pay for stopping Nash buddies, oh yeah!

G

This is this is the way that you know early on the stream. You would send this. You know with your initial keyframe, basically indicating here stuff, that's going to be set right now and if the metal box then decides hey, I'm gonna actually started pouring the base layer only or basic plus enhancement. They can update the scale. That is structured and say here are the layers that he was a endpoint Lashley receiving, and then you can decide from your decoder. Will you act when you should actually emit a frame?

G

Do you wait for basement enhancement or you just stop as soon as your people base.

A

A

G

Would need to go check. I think my Jonathan Guillen yesterday photos for the frame girl, entire.

E

Think its lair and.

A

E

Our assumption is that spatial 0 begin a player and.

A

G

Is expedient spatial I? Did you go back.

A

G

I mean we went up to figure out kind of how we have this preset structure the Gotham neck stuff with like out, because that hasn't explicit E&P feel as opposed to my heavily implicit from like the cough structure, but yet a spit. What the s number is always independent and, as you know, occupied by three kids, okay,.

C

Well, we're little bit over time so.

G

Exactly what we actually accounted for, I think that I was thinking about this disposable layer.

C

Okay, so what's the mixed up for this to connect.

G

See you know, p p9 and we have actually be non-running side. Chrome, it's a behind a flag. This is not turn on by default, but getting the packetization so standardized as one of the things that's keeping us from acting front of the switch we like to see this adopted as a working group I, don't so that we can then sort of give us on the standards track and turn into an RC. Ok,.

C

So again, I think in terms of in general terms, I mean it's in disappeared for Matamata pyramid for codec, so I don't see any anyone thinks that it's not part of the payload working group took another pair after deep enough of everything that I don't think. So anyone objecting that it said item for the working group to work on anyway, who supports that is working for the working toward you bring work going ok, so we will ask for a milestone for that and we'll start to work into the working group document.

C

But of course we need to review them to get a better and to get to the people interested. Do some work on the brain.

G

Exactly MO and Jonathan I need to talk about.

G

C

G

Doctrine probably should speak to that.

C

Ok, ok, ok! So this was the last presentation and thank you and we'll see you tomorrow with other meetings.

C

Now we first have to us, we have to also a milestone and and then confirm on the list and then and we'll tell you that well, but.

C

But you should the tissue and.

C

The only thing you should make us to do this step, something that's like to remind us that we need to ask for.

A

C

I'll, be there a little Shepherd for this document as well. Sir I'll take care.

C

We need well to ask for agreement of all the doc way.

E

B

C

The true character yet.

A

C

What's the last mountain, then we can start putting the loan because it is what them series MLP. They want autonomous.

A

Never with the.

C

Candle stuff, Crandall.

C

This is the cutest one. Is the.

A

Meeting go: the company pays my last I.

C

Don't work just a physically, they are.

E

C

To put us here, no, no we need to do is go to a desk from these monks. Now this is just so. You know.

C

There are planning our.

C

Operational management.

A