[FFmpeg-devel] libavdevice: PTS stability and risks of drift

[Marc Mason]

Michael Niedermayer wrote:

> some people might run ntpdate once a day or week this might lead to a few
> second stepping of the time when its run, I don't think an hour is realistic.

I had a PC whose clock was 100 ppm (0.01%) too fast. I was told by the 
NTP devs this was not uncommon.

Stepping the clock once a week meant setting the clock back 60 seconds 
once a week.

AFAIR, NTP refuses to handle clocks outside +/- 500 ppm

-- 
Regards.