search

cta-standards / R4WG20-QoE-Metrics

Issue tracking repository for the R4-Wg20 QoE Initiative
9 stars 2 forks source link

Consider adding "Average wait time percentage" (or replace "average stalled time percentage") #6

Closed tonihei closed 5 years ago

tonihei commented 5 years ago

For video services playing very short videos, the "average stalled time percentage" is likely to be much less than the percentage of watched time spent waiting for the startup time.

A potentially better and more robust metric is the "average wait time percentage" (name TBD) which includes all wait times (startup, stalls, seeks) as a percentage of the watched time (wait times+play time).

gheikkila commented 5 years ago

Yes, in principle you are correct. Especially for short clips there is a clear tradeoff between the start time and any rebuffering during the clip. With long enough start time there will never be any mid-content rebufferings.

mlevine84 commented 5 years ago

WG believes the viewer impact of initial start up and re-buffering time is different and there are metrics defined in the spec to calculate the requested information. By combining these metrics, the total waiting time can also be analyzed. Hence WG recommends not to replace or add a new metric.