[MEncoder-users] ATSC closed captioning

[Reimar Döffinger]

On Sun, Nov 28, 2010 at 09:57:09AM -0800, Scott W. Larson wrote:
> On Sun, 2010-11-28 at 08:43 +0100, Reimar Döffinger wrote:
> > It doesn't really matter to MPlayer _where_ the data is, at this point of
> > the code everything is considered to be part of a frame, but there's currently
> > a problem when there's more than one data packet in what MPlayer considers
> > one frame.
> 
> Yep, that's the problem I discovered and why I had to create a separate
> linked list to reorder all of the caption data correctly. The user data
> doesn't necessarily arrive associated with a displayable frame. It can
> arrive outside of a GOP.

Again: there is no "outside of a GOP" or "outside of a frame" in
MPlayer. All data, every single bit belongs to some frame both
inside MPlayer and the ffmpeg decoders.
The problem is multiple userdata segments within (in MPlayer
terms) one frame.

> Most of the time it doesn't though. Maybe an occasional dropped
> character or two is good enough for mplayer.

But the point of "reimplementing" was to get a "proper" solution :-).
I actually think that the most reasonable approach would be to store it
in NAL-unit format, then it could also be used for "custom" formats
where other parts than 0x1B2 are used for custom data.