[FFmpeg-devel] Project orientation

[Jean-Baptiste Kempf]

On Sat, Jul 4, 2020, at 19:44, Michael Niedermayer wrote:
> Another area as we are already on the subject is stuff falling a little
> outside what FFmpeg covers.
> Not every filter that anyone wants to write should have to be in FFmpeg

Thank $deity that you are saying this!

FFmpeg is used more and more and numerous people want to put everything in it.

The project needs to learn to say no to features and define the scope of each library.

For libavcodec, the scope is quite clear, even if there are some debates about experimental codecs.

But for the other libraries, including avformat and avfilter, there are dubious things in.

For example and in my opinion, I don't see why there is support in avformat for things that are not standards, like AMQP/ZMQ, who are neither a multimedia format nor a multimedia protocol.
In the same way, if you start to need tensorflow or openvinno for a filter, the code has no place in FFmpeg: you can use the api and use that externally. Importing tensorflow is probably more lines of code, than FFMpeg.
Or speech-recognition (sphinx) or OCR (tesseract)...
And now comes the debate with synthesizers or complex filters that are almost reinventing GStreamer inside libavfilter ("let's use libavcodec as a filter, since everything is a filter")...

When you look at the FFmpeg dependency graph on distributions like Debian, this is getting scrary, and this increases a lot the binary size and resources usage.

At some point, the project needs to decide what is in and what is out, and since FFmpeg has numerous APIs, people can plug them externally. It's not a problem to say "No" to a feature, especially when, the number of people using that feature is under 0,01% of FFmpeg users, and when people can build that externally anyway.


--
Jean-Baptiste Kempf - President
+33 672 704 734