[FFmpeg-devel] [RFC] New library for shared non-generic libav* utils

[Stefano Sabatini]

On date Friday 2010-07-09 17:09:44 +0200, Michael Niedermayer encoded:
> On Fri, Jul 09, 2010 at 05:04:32PM +0200, Michael Niedermayer wrote:
> > On Fri, Jul 09, 2010 at 09:54:11AM -0400, Ronald S. Bultje wrote:
> > > Hi,
> > > 
> > > On Thu, Jul 8, 2010 at 7:07 PM, Stefano Sabatini
> > > <stefano.sabatini-lala at poste.it> wrote:
> > > [.. cut ..]
> > > > This new lib will contain all code/utils which need to be shared
> > > > between more libav* libs, and are not enough generic to deserve a
> > > > place in libavutil, which is to be considered a collection of
> > > > generic/non-multimedia-related utilities.
> > > 
> > > Disregard me if majority says otherwise, I just wanted to
> > > bikesheddishly note that my personal humble opinion is that less libs
> > > is good, so I'd not have any problems with media-related stuff going
> > > into libavutil. I think the chance that people use a FFmpeg lib for
> > > something unrelated to multimedia is relatively small and should not
> > > be our main focus. Reminds me of not allowing media-specific stuff in
> > > libgstreamer.so. It only causes headaches and distractions. There is
> > > no practical advantage.
> > 
> > as maintainer of libavutil i object. We can have a seperate lib for
> > common code. Iam not stopping people from having their common lib
> > which prior to libavfilter was libavcodec. But now due to libavfilter
> > not depending on libavcodec this is no longer possible.
> 
> also if people dont want a new lib, we can make libavfilter depend on
> libavcodec
> are there any projects that use libavfilter that do not use libavcodec?

Not that I know, but this is something which I would like to avoid if
possible, filtering and encoding/decoding look like different problems
and I feel that introducing a dependency on libavcodec (which is
*big*) in lavfi is the wrong way to go.

I'm concerned about the fact that adding a new libav* would increase
complexity, on the other end I like the fact that:
* libavutil would stay as minimal as possible, with code re-usable in
  other non-multimedia related projects
* we would avoid to add unnecessary dependencies (e.g. lavfi <- lavc)

So I consider that the pros outhweigh the cons in this case, thus I'm
in favor of such a new library.

Regards.
-- 
FFmpeg = Frightening & Fundamentalist Meaningful Prodigious Ecletic Geisha