Re: [rtcweb] Proposal for a JS API for NoPlan (adding multiple sources without encoding them in SDP)

[Peter Thatcher <[e-mail]>]

On Mon, Jun 17, 2013 at 6:00 PM, Silvia Pfeiffer
<[e-mail]>wrote:

> Hi Peter, all,
>
> I'm looking at all this from the view of a JS developer and I am
> really excited that there is movement in this space. Having hit my
> head hard against the limitations of the current WebRTC API and being
> forced to learn SDP to achieve some of the less common use cases, I'm
> feeling the pain. I am therefore happy to see that there is a proposal
> for a higher-level API with similarities to the Microsoft's CU-WebRTC
> proposal and I hope that together with Microsoft's input this can
> become a really useful API.
>

I hope so as well.  Unfortunately, Microsoft's input seems to be "our way
or the highway" and the input of others seems to be "SDP ought to be enough
for anybody".  My thinking is that we can make incremental improvements
toward a cleaner API without being so extreme at either end.  And I hope I
can find others that think similarly.


What I would like to see, though, is a bit different from what you've
> proposed. In particular, the MediaFlowDescription object is the one
> object that I believe is supposed to enable JS developers to  do "SDP
> hacking" without having to understand SDP. Unfortunately, the way in
> which it is currently written, this API doesn't help a JS developer
> much. There are properties in that object like "ssrc" that mean
> nothing unless you understand SDP.
>

Actually, it helps some JS developer a lot.  It happens to be very good for
the JS I am writing :).


>
> I would therefore like to recommend making the properties on the
> MediaFlowDescription more semantic - in particular giving them better
> names - such that a JS developer really doesn't have to understand SDP
> to create/change media flow descriptions. Can we find better names for
>  id, transportId and ssrc that are more explicitly expressing what
> they stand for and when a JS developer would actually change them?
>
>
I did my best with the names.   I passionate about good names for things,
but we have certain constraints we have to work within.  ssrc is an RTP
name that cannot be avoided.  "id" of a track is the same as the
MediaStreamTrack.id, so it's already well defined in the rest of the API.
 TransportId is a little more hairy, since there is no clean API for
transports, so it has to refer to the SDP.  I could have called it
transportSdpMid to be more clear what it is, but I doubt you would have
liked that name :).  Are there any others that you don't like?


> It would be nice if the MediaFlowDescription was abstract enough such
> that in future it doesn't matter if SDP is actually underneath (with
> offer/answer),


MediaFlowDescription isn't using SDP underneath.  There would be no SDP
between the JS and the browser if MediaFlowDescription were used.  And, it
isn't offer/answer either (unless you implement offer/answer on top it,
obviously.  It's the JS's choice of what to do).



> but that's not actually my main goal. What I'm after is
> an API that can be used without having to understand what is
> underneath.
>

For signalling, what's in MediaStreamTrackDescription must be serialized,
sent to the receiving side, and deserialized.  That is a pretty fundamental
requirement, and short of deciding all RTP parameters a priori, there's no
way to do RTC without doing so.  So, you at least will need to know how to
accomplish that.  Of course, a library on to of the WebRTC API may provide
you that signaling so you don't have to.  But the WG has, correctly,
decided to leave the decision of the signalling to the JS and not back it
into the API.




>
> Thanks for listening and HTH,
> Silvia.
>
>
> On Mon, Jun 17, 2013 at 10:57 PM, Peter Thatcher <[e-mail]>
> wrote:
> > Google is in full support of "Plan B" for encoding multiple media
> sources in
> > SDP, and would like to see the Plan A vs. Plan B decision resolved soon.
> > Recently, though, a third option, called "NoPlan" has been proposed, but
> it
> > lacked the details of what a JS API would look like for NoPlan.  Cullen
> > asked to see such an API proposal, and so I have worked with Emil to make
> > one.  He will be adding it to the NoPlan draft, but I will also include
> it
> > in this email.
> >
> > Again, Google is in full support of "Plan B".  But if Plan A vs. Plan B
> > cannot be decided, then we support NoPlan with the following additions to
> > the WebRTC JS API as an option that allows implementing either Plan A or
> > Plan B  in Javascript.  And even if Plan A vs. Plan B is resolved, these
> API
> > additions would still be a big improvement for those WebRTC applications
> > that don't use SDP for signalling.
> >
> > It is a bit long because I have added many comments and examples, but the
> > actually additions only include two new methods on PeerConnection and a
> few
> > new dictionaries.  So don't be overwhelmed :).
> >
> >
> >
> > Intro: This follows the model of createDataChannel, which has a JS
> method on
> > PeerConnection that makes it possible to add data channels without going
> > through SDP.  Furthermore, just like createDataChannel allows 2 ways to
> > handle neogitation (the "I know what I'm doing;  Here's what I want to
> send;
> > Let me signal everything" mode and the "please take care of it for me;
>  send
> > an OPEN message" mode), this also has 2 ways to handle negotiation (the
> "I
> > know what I'm doing; Here's what I want to send; Let me signal
> everything"
> > mode and the "please take care of it for me;  send SDP back and forth"
> > mode).
> >
> > Following the success of createDataChannel, this allows simple
> applications
> > to Just Work and more advanced applications to easily control what they
> need
> > to.  In particular, it's possible to use this API to implement either
> Plan A
> > or Plan B.
> >
> > // The following two method are added to RTCPeerConnection
> > partial interface RTCPeerConnection {
> >  // Create a stream that is used to send a source stream.
> >  // The MediaSendStream.description can be used for signalling.
> >  // No media is sent until addStream(MediaSendStream) is called.
> >  LocalMediaStream createLocalStream(MediaStream sourceStream);
> >
> >  // Create a stream that is used to receive media from the remote side,
> >  // given the parameters signalled from MedaiSendStream.description.
> >  MediaStream createRemoteStream(MediaStreamDescription description);
> > }
> >
> >
> > interface LocalMediaStream implements MediaStream {
> >   // This can be changed at any time, but especially before calling
> >   // PeerConnection.addStream
> >   attribute MediaStreamDescription description;
> > }
> >
> >
> > // Represents the parameters used to either send or receive a stream
> > // over a PeerConnection.
> > dictionary MediaStreamDescription {
> >   MediaStreamTrackDescription[] tracks;
> > }
> >
> >
> > // Represents the parameters used to either send or receive a track over
> //
> > a PeerConnection.  A track has many "flows", which can be grouped
> > // together.
> > dictionary MediaStreamTrackDescription {
> >   // Same as the MediaStreamTrack.id
> >   DOMString id;
> >
> >   // Same as the MediaStreamTrack.kind
> >   DOMString kind;
> >
> >   // A track can have many "flows", such as for Simulcast, FEC, etc.
> >   // And they can be grouped in arbitrary ways.
> >   MediaFlowDescription[] flows;
> >   MediaFlowGroup[] flowGroups;
> > }
> >
> > // Represents the parameters used to either send or receive a "flow"
> > // over a PeerConnection.  A "flow" is a media that arrives with a
> > // single, unique SSRC.  One to many flows together make up the media
> > // for a track.  For example, there may be Simulcast, FEC, and RTX
> > // flows.
> > dictionay MediaFlowDescription {
> >   // The "flow id" must be unique to the track, but need not be unique
> >   // outside of the track (two tracks could both have a flow with the
> >   // same flow ID).
> >   DOMString id;
> >
> >   // Each flow can go over its own transport.  If the JS sets this to a
> >   // transportId that doesn't have a transport setup already, the
> >   // browser will use SDP negotiation to setup a transport to back that
> >   // transportId.  If This is set to an MID in the SDP, then that MID's
> >   // transport is used.
> >   DOMString transportId;
> >
> >   // The SSRC used to send the flow.
> >   unsigned int ssrc;
> >
> >   // When used as receive parameters, this indicates the possible list
> >   // of codecs that might come in for this flow.  For exmample, a given
> >   // receive flow could be setup to receive any of OPUS, ISAC, or PCMU.
> >   // When used as send parameters, this indicates that the first codec
> >   // should be used, but the browser can use send other codecs if it
> >   // needs to because of either bandwidth or CPU constraints.
> >   MediaCodecDescription[] codecs;
> > }
> >
> >
> > dictionary MediaFlowGroup {
> >   DOMString type;  // "SIM" for Simulcast, "FEC" for FEC, etc
> >   DOMString[] flowids;
> > }
> >
> > dictionary MediaCodecDescription {
> >   unsigned byte payloadType;
> >   DOMString name;
> >   unsigned int? clockRate;
> >   unsigned int? bitRate;
> >   // A grab bag of other fmtp that will need to be further defined.
> >   MediaCodecParam[] params;
> > }
> >
> > dictionary MediaCodecParam {
> >   DOMString key;
> >   DOMString value;
> > }
> >
> >
> > Notes:
> >
> > - When LocalMediaStreams are added using addStream, onnegotiatedneeded is
> > not called, and those streams are never reflected in future SDP
> exchanges.
> > Indeed, it would be impossible to put them in the SDP without first
> > resolving if that would be Plan A SDP or Plan B SDP.
> >
> > - Just like piles of attributes would need to be defined for Plan A and
> for
> > Plan B, similar attributes would need to be defined here (Luckily,  much
> > work has already been done figuring out what those parameters are :).
> >
> >
> > Pros:
> >
> > - Either Plan A or Plan B or could be implemented in Javascript using
> this
> > API
> > - It exposes all the same functionality to the Javascript as SDP, but in
> a
> > much nicer format that is much easier to work with.
> > - Any other signalling mechanism, such as Jingle or CLUE could be
> > implemented using this API.
> > - There is almost no risk of signalling glare.
> > - Debugging errors with misconfigured descriptions should be much easier
> > with this than with large SDP blobs.
> >
> >
> > Cons:
> >
> > - Now there are two slightly different ways to add streams: by creating a
> > LocalMediaStream first, and not.  This is, however, analogous to setting
> > "negotiated: true" in createDataChannel.  On way is "Just Work", and the
> > other is more advanced control.
> >
> > - All the options in MediaCodecDescription are a bit complicated.
>  Really,
> > this is only necessary because Plan A requires being able to specify
> codec
> > parameters per SSRC, and set each flow on different transports.  If we
> did
> > not have this requirement, we could simplify.
> >
> >
> > Example Usage:
> >
> > // Imagine I have MyApp, handles creating a PeerConnection,
> > // signalling, and rendering streams.  This is how the new API could be
> > // used.
> > var peerConnection = MyApp.createPeerConnection();
> >
> > // On sender side:
> > var stream = MyApp.getMediaStream();
> > var localStream = peerConnection.createSendStream(stream);
> > sendStream.description = MyApp.modifyStream(localStream.description)
> > MyApp.signalAddStream(localStream.description, function(response)) {
> >   if (!response.rejected) {
> >     // Media will not be sent.
> >     peerConnection.addStream(localStream);
> >   }
> > }
> >
> > // On receiver side:
> > MyApp.onAddStreamSignalled = function(streamDescription) {
> >   var stream = peerConnection.createReceiveStream(streamDescription);
> >   MyApp.renderStream(stream);
> > }
> >
> >
> > // In this exchange, the MediaStreamDescription signalled from the
> > // sender to the receiver may have looked something like this:
> >
> > {
> >   tracks: [
> >   {
> >     id: "audio1",
> >     kind: "audio",
> >     flows: [
> >     {
> >       id: "main",
> >       transportId: "transport1",
> >       ssrc: 1111,
> >       codecs: [
> >       {
> >         payloadType: 111,
> >         name: "opus",
> >         // ... more codec details
> >       },
> >       {
> >         payloadType: 112,
> >         name: "pcmu",
> >         // ... more codec details
> >       }]
> >    }]
> >  },
> >  {
> >     id: "video1",
> >     kind: "video",
> >     flows: [
> >     {
> >       id: "sim0",
> >       transportId: "transport2",
> >       ssrc: 2222,
> >       codecs: [
> >       {
> >         payloadType: 122,
> >         name: "vp8"
> >         // ... more codec details
> >       }]
> >    },
> >    {
> >      id: "sim1",
> >      transportId: "transport2",
> >      ssrc: 2223,
> >      codecs: [
> >      {
> >        payloadType: 122,
> >        name: "vp8",
> >        // ... more codec details
> >      }]
> >    },
> >    {
> >      id: "sim2",
> >      transportId: "transport2",
> >      ssrc: 2224,
> >      codecs: [
> >      {
> >        payloadType: 122,
> >        name: "vp8",
> >        // ... more codec details
> >      }]
> >    },
> >
> >    {
> >      id: "sim0fec",
> >      transportId: "transport2",
> >      ssrc: 2225,
> >      codecs: [
> >      {
> >        payloadType: 122,
> >        name: "vp8",
> >        // ...
> >      }]
> >    }],
> >    flowGroups: [
> >    {
> >      semantics: "SIM",
> >      ssrcs: [2222, 2223, 2224]
> >    },
> >    {
> >      semantics: "FEC",
> >      ssrcs: [2222, 2225]
> >    }]
> >  }]
> > }
> >
> >
> > Constructive feedback is welcome :).
> >
> > _______________________________________________
> > rtcweb mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/rtcweb
> >
>