Re: [webrtc-stats] Request to add neteq waiting time to stats report (#427)

Yeah, it is possible to poll jitterBufferDelay.  The downside is computational cost involved.  Given the current implementation of RTCStatsReport, I have to poll the entire report each time just to compute one or two values.  Whereas bringing those two variables into report can help to avoid all the hassle.  An alternative option is to have a few light-weight polling method focusing on smaller set of values.  It doesn't exist today.  I am doubtful if it matches webrtc roadmap and strategy as well.

BTW, is jitterBufferDelay an average value or an instantaneous value?  Does it include time consumed in computing PLC, CNG, PLCCNG, etc?

Max is very useful to understand how badly the service has experienced.  If it deviates from mean much, there is indication that there may be severe voice quality degradation in addition round trip latency effect.  It is desired to be a short term moving value.  It gives information on user experience over time.  Right now, it is implemented as max over an entire call.  It can be very useful when detailed call logs are not available.  One can still use snippet of it at end of the call to infer what had happened.  

GitHub Notification of comment by wssjwl
Please view or discuss this issue at using your GitHub account

Received on Friday, 12 April 2019 16:52:06 UTC