[MPlayer-users] different output between mplayer & ffplay/ffmpeg

[Reimar Döffinger]

On Sun, Aug 18, 2013 at 10:24:20AM +0400, Abu Abdullah wrote:
> would it be better to start all the time before the seek point (or at it
> exactly if possible) unless it exceeds certain limit e.g. 1 second where
> you will start apply the logic of nearest frame. my point is that you do
> not want to miss some of the speech.

I think you are focusing too much on your use-case and forget about
most of the real-world uses.
In many cases 0.5 seconds behind is far better than 2 minutes before
the seek point.
In addition, when the user tries to seek forward by 10 seconds, picking
a location 20 seconds earlier than desired is not just worse than
something 2 minutes ahead, it is simply broken and gives a crappy user
experience.
In principle it is possible to do exact seeking, but it will be slower.
More than a few users care more about fast seeking than accurate
seeking.
Using the API in the patch I pointed out it would not be hard to make
this configurable, but if your solutions is based on "just do ... all
the time" I am quite certain it will _not_ work well for most users.
Which I guess might be part of the reason why we're kind of stuck with
the current behaviour, even if it's not as good as it could be.