Same video stream for each user connected

[CSantosM]

Expected behavior

Unique video stream per user.

Observerd behavior

Testing the simple openvidu-ionic app, I have realised that since 6.0.8 version (included) the video stream received by cordova-plugin-iosrtc is the same that first one.

This behaviour is not happening with the 6.0.7 cordova-plugin-iosrtc version.

Steps to reproduce the problem

To reproduce this issue, we only have to join to the room with ionic ios app and connect multiple devices to the same session.
From the ios app view, the remote streams will be the same that the first remote user connected.

Platform information

• Cordova version: 9.0.0
• Plugin version: 6.0.9
• iOS version: 13.3 and 13.1.3
• Xcode version: 11.3.1

[hthetiot]

@CSantosM do you use Janus ?

[hthetiot]

Most likely related to this #467

[CSantosM]

No, i do not

[hthetiot]

@CSantosM i do not reproduce that issue using 6.0.9 on Sylaps.com using https://apps.apple.com/us/app/sylaps/id954016392

[hthetiot]

@CSantosM Can you check MediaStream.id and MediaTracks.id for remote streams.
What 6.0.8 changes does is keep the media.id from the remote instead of faking them.

https://github.com/cordova-rtc/cordova-plugin-iosrtc/pull/467/files#diff-379e16e943b1b62acdbc68ad8a80241fL20

Do you use any MCU/SFU ?

[CSantosM]

Hello @hthetiot,
Checking MediaStream.id and MediaStreamTrack.id.
I've added a print log into a PluginMediaStream.swift and PluginMediaStreamTrack.swift

The logs are the following:
When localUser(ionic - ios) is connected:

OpenViduLog: MediaStreamTrack ID =>  -6336181468174245674
OpenViduLog: MediaStreamTrack ID =>  133574601189058849
OpenViduLog: MediaStream ID =>  FFFB9244-1BD4-44B4-9B72-F53F4D2B9735

Then, When a Chrome browser join to the session. The stream looks correct:

OpenViduLog: MediaStreamTrack ID =>  6014604358066097781
OpenViduLog: MediaStreamTrack ID =>  -3513255213319178647
OpenViduLog: MediaStreamTrack ID =>  6014604358066097781
OpenViduLog: MediaStreamTrack ID =>  -3513255213319178647
OpenViduLog: MediaStream ID =>  default

Then another different user is conncted. The same stream as above will appear.

OpenViduLog: MediaStreamTrack ID =>  3716680559757491135
OpenViduLog: MediaStreamTrack ID =>  810162222955606700
OpenViduLog: MediaStreamTrack ID =>  3716680559757491135
OpenViduLog: MediaStreamTrack ID =>  810162222955606700
OpenViduLog: MediaStream ID =>  default

The mediaStreamTracks are different each others.

OpenVidu uses exclusively SFU.

It helps?

[hthetiot]

OpenVidu uses exclusively SFU.

Your SFU is not generating unique MediaStream or MediaTrack ID :)
In the past iosrtc was creating unique ID, we now reuse the ID from SDP.

What SFU are you using please, so i can make joke about them.

@CSantosM

[j1elo]

Are you talking here about the a=msid attribute in SDP?

The SFU in use is Kurento, which doesn't support Unified Plan nor it needs to, given that it only works with one single audio and one single video per WebRTC connection. With these conditions, both PlanB and Unified Plan are equivalent in practice.

Any endpoint should be able to rely on proper SDP Offer/Answer negotiation to convey that the a=msid attribute is not supported, by the fact that it is missing on the SDP Answer. So, after receiving such Answer, the local endpoint is expected to adapt to that (just like any other negotiated attribute).

Using an internally-generated UUID is the thing to do for endpoints that really require it. That's by the way what libwebrtc does (see commit), and the standard text mentions too (see PR).

[hthetiot]

@j1elo Thank you for detailed explanation.
Can we consider that generating UUID when using PlanB and not 36 char is a acceptable solution?

The thing is OpenEasyRTC and many other WebRTC library does relly on msid to match on both side, and even if i use PlanB on Chrome, Firefox or Safari since 2016, msid always matched.

[hthetiot]

I will look into WebRTC changes here https://chromium.googlesource.com/external/webrtc/+/5b1477839d8569291b88dfe950089d0ebf34bc8f and see if needed on our side then release 6.1.0

[hthetiot]

The thing is Kurento does give msid, it's just named default... may be this issue is that internaly iosRTC use those, may be the fix is not to generate UUID here but making sure iosrtc internaly make the difference using an internal uuid.

[hthetiot]

I will try to release 6.0.10 in the next 48 hours for you guys :)
With a fix.

[hthetiot]

I need more time to make it as good as https://chromium.googlesource.com/external/webrtc/+/5b1477839d8569291b88dfe950089d0ebf34bc8f keep you posted.

[micaelgallego]

Thank you!

If you need any help on our side (OpenVidu team), please ping us.

[hthetiot]

@micaelgallego I have a proper fix on a locale branch with M75 for now i did #493 dirty fix can you check.

In any case, PR are always welcome.

[hthetiot]

ping @CSantosM see #480 (comment)

[pabloFuente]

We will give it a try ASAP and give ypu feedback. Thank you very much!

[CSantosM]

Hi @hthetiot ,

We have tested this branch and it looks like the problem has been solved.
We have connected three users in the same session and we have not reproduced the bug because of the unique mediastream id.

Great, let us know when this bug fix will be released.

Regards

[hthetiot]

@CSantosM Ok will release

[hthetiot]

Done https://github.com/cordova-rtc/cordova-plugin-iosrtc/blob/master/CHANGELOG.md#version-6011

[micaelgallego]

Thanks!

[CSantosM]

Hey @hthetiot

Where can I found the 6.0.11 release? The latest version shown is 6.0.10.

Regards

[hthetiot]

Npm ?

• https://www.npmjs.com/package/cordova-plugin-iosrtc

I also just pushed tag if that what you where looking for.

[hthetiot]

cc @CSantosM

https://github.com/cordova-rtc/cordova-plugin-iosrtc/tree/6.0.11

[CSantosM]

Ok, thanks @hthetiot