[Galene] Re: Experimental outbound activity detection

[Galene] Re: Experimental outbound activity detection

[Juliusz Chroboczek]

> I was trying to understand what's missing for outbound activity detection
> (aka "Is this thing on?"), and learned that there is a "totalAudioEnergy"
> property for outgoing WebRTC streams as well, but it's not yet implemented
> in Firefox:
> https://developer.mozilla.org/en-US/docs/Web/API/RTCAudioSourceStats/totalAudioEnergy

Yes.  Firefox's support for WebRTC is unfortunately incomplete, there are
a number of workarounds in the code for Firefox's bugs.  I really need to
check if they are still needed in recent versions (and display a warning
if the user is running an old version of Firefox).

A workaround would be to push the output through a WebAudio component,
I hear that WebAudio is well supported under Firefox.  I don't know if it
would increase CPU usage by much.

(WebAudio is a lot of fun, by the way, it's what's used on the login page
for the Fourier transform.  It could be used for implementing audio
filters, it's extremely flexible if you don't care about CPU and battery
usage.)

> https://github.com/jech/galene/compare/master...alxndr42:galene:outbound-activity

I don't think you need maxAudioEnergy, there should be only one audio track.

> It does seem to work in Chromium, using largely the existing code. Here's
> an experimental branch, please let me know what you think:
> https://github.com/jech/galene/compare/master...alxndr42:galene:outbound-activity

I haven't looked yet, but one thing comes to mind: we currently use
different intervals for up and down stats, so the appearance will be
different.  Should we make the intervals equal, now that the up connection
stats are no longer visible by default?

> So right now the code detects no activity in Firefox, but "should" start
> to work if they eventually implement the missing parts.

Don't count on it.  I fear there's nobody working on WebRTC at Firefox.

-- Juliusz

[Galene] Re: Experimental outbound activity detection

[Alexander Dietrich]

> (WebAudio is a lot of fun, by the way, it's what's used on the login page
> for the Fourier transform.  It could be used for implementing audio
> filters, it's extremely flexible if you don't care about CPU and battery
> usage.)
That's probably the way to go, if RTCAudioSourceStats turns out to be a 
dead end. But I wanted to see first how far I could get by re-using the 
existing code.
> I haven't looked yet, but one thing comes to mind: we currently use
> different intervals for up and down stats, so the appearance will be
> different.  Should we make the intervals equal, now that the up connection
> stats are no longer visible by default?
Both stat types are updated at the same rate in my branch. The label is 
updated at the previous rate, so that it doesn't become distracting.
> Don't count on it.  I fear there's nobody working on WebRTC at Firefox.

The last comment in the Bugzilla issue is from this month, so maybe not 
all hope is lost. :)

Kind regards,
Alexander