[FFmpeg-devel] [PATCH]lavf/mxfdec: Export EIA608 Closed Captions by default

[Tomas Härdin]

sön 2019-02-17 klockan 16:28 +0100 skrev Marton Balint:
> 
> On Sun, 17 Feb 2019, Marton Balint wrote:
> 
> > 
> > 
> > On Sun, 17 Feb 2019, Carl Eugen Hoyos wrote:
> > 
> > > > > > 2019-02-17 1:40 GMT+01:00, Marton Balint <cus at passwd.hu>:
> > > > 
> > > > On Sat, 16 Feb 2019, Carl Eugen Hoyos wrote:
> > > > > 
> > > > > I am not sure why there is an option to disable Closed Captions export,
> > > > > but disabling the export by default seems like a bad idea to me.
> > > > 
> > > > SMPTE 436M can be any kind of ancillary data, not only closed captions.
> > > > That is why the default behaviour (pass all the data to userspace so a
> > > > userspace app can parse it properly) makes sense.
> > > > Some applications might depend on this.
> > > 
> > > Wouldn't it still make more sense to change the default?
> > 
> > Sorry, no. We should not change existing behaviour for this.
> > 
> > > The application that knows it needs the ancillary data, well, knows
> > > it while the average user who reads the mxf file has no idea that
> > > there is a subtitle stream and has no idea that FFmpeg can read
> > > the subtitle stream.
> > > 
> > > I would bump micro in any case and the option also works fine
> > > with older FFmpeg versions so I don't really see the issue.
> > 
> > Issue is that you are removing a data stream which was previously 
> > provided. Also CC extraction in MXF decoder is a hack. It should be 
> > deprecated after the API will be capable of bitstream filtering between 
> > different codecs. And the VANC data can be parsed to multiple 
> > subtitle/data streams. Why eia608 has the perference? Why not teletext, or 
> > SCTE messages? All in all, it is better to keep hacks under the 
> > option until we come up with better API.
> 
> One more thing is that the fact that a file has VANC data does not mean 
> that it actually has Closed Captions data. It might, or it might not, so 
> unconditionally advertising a CC stream just because there _might_ be CC 
> data would also be bad.

Isn't this stuff also used for teletext and whatnot? I generally agree
that we shouldn't assume it's subtitles

/Tomas