W3C home > Mailing lists > Public > public-webrtc-logs@w3.org > October 2017

Re: [webrtc-stats] jitterBufferDelay and concealed samples, DTX/CNG samples

From: Harald Alvestrand via GitHub <sysbot+gh@w3.org>
Date: Wed, 25 Oct 2017 09:36:16 +0000
To: public-webrtc-logs@w3.org
Message-ID: <issue_comment.created-339273094-1508924175-sysbot+gh@w3.org>
Hm. One problem here is that the text "from the time the first packet belonging to an audio/video frame enters the jitter buffer" is implementation dependent, and inconsistent with the previous sentence's "from the time it is received". An implementation that places packets into the jitter buffer on arrival, and reassembles the frame within the jitter buffer, will give different results than one that separates the jitter buffer from the reassembly step.
Suggestion: Change "enters the jitter buffer" to "is received". Implementation-wise, this would involve timestamping the packet at reception (we already do, I think), mark the frame with the timestamp of the lowest-timestamped packet contributing to it, and measure the difference between "current time" and the timestamp when a frame is taken out of the buffer to be displayed.
Since only frames taken out of the buffer are counted, none of the current counters quite fit; we should define new counters for "framesPlayed" and "totalSamplesPlayed".

-- 
GitHub Notification of comment by alvestrand
Please view or discuss this issue at https://github.com/w3c/webrtc-stats/issues/246#issuecomment-339273094 using your GitHub account
Received on Wednesday, 25 October 2017 09:36:32 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 9 October 2019 15:14:56 UTC