[FFmpeg-devel] [PATCH 3/3] lavfi/formats: document the negotiation process

[Soft Works]

> -----Original Message-----
> From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of
> Nicolas George
> Sent: Thursday, 19 August 2021 17:31
> To: ffmpeg-devel at ffmpeg.org
> Subject: [FFmpeg-devel] [PATCH 3/3] lavfi/formats: document the
> negotiation process
> 
> Signed-off-by: Nicolas George <george at nsup.org>
> ---
>  libavfilter/formats.h | 85
> +++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 85 insertions(+)

Hi Nicolas,

I finally came to review your proposed changes as documented in this
patch and I also watched your presentation:
https://www.normalesup.org/~george/articles/format_negotiation_in_libavfilter/#0

From a general point of view, most of your ideas make sense to me - probably
even all of them when viewed individually.

But in the moment when you are putting this all together, I'm getting a bit
scared. The one concern I have about this is the possibility of loosing

- control of negotiated formats and
- deterministic filter behavior

The whole process (e.g. a single ffmpeg run) will probably remain to be 
deterministic  (same input will always produce the same output), but that
would no longer apply to individual filters in a graph.
What I mean is that - at least as far as I have understood the proposal,
every single change that you make to a filter graph could completely change
the formats that are negotiated between individual filters.

This can surely provide better results than the current implementation,
but it can also make things worse in cases like mine, where I am building
filter graphs in a way that relies on precisely predicting/knowing the
negotiated formats of all filter connections in advance.

I don't want get too far at this point; maybe I misunderstood something,
or you already have a plan for addressing these concerns?

PS: I'll respond separately on the subtitles subject 

Kind regards,
softworkz