[mythtv-users] FFmpeg releases version 0.5

[Nick Rout]

On Wed, Mar 11, 2009 at 12:04 PM, Robert McNamara
<robert.mcnamara at gmail.com> wrote:
> On Tue, Mar 10, 2009 at 2:57 PM, gonzalo diethelm <gdiethelm at dcv.cl> wrote:
>
>> Straight, honest to God question: why is it necessary to import the
>> code? Would it be possible to use FFmpeg as an external library?
>>
>
> Not that thorny-- a couple of reasons as I understand them.  Firstly,
> Myth has a fair number of local modifications for the libav* source,
> particularly the Transport Stream and Program Stream demuxers, which
> makes the upstream versions unsuitable without some hand massaging.
> Secondly, importing the source at compile time is very very tricky--
> ask mplayer devs, who are *constantly* fixing compilation of their
> player because of their decision to do so.  The ffmpeg code changes a
> *lot*, including the API, and it would be a lot more work than the
> relatively small group of core devs could easily manage.
>
> Myth as a project is maybe 10ish core devs and 20-30 regular
> contributors-- just keeping a dynamically imported version of libav*
> working adds up fast in manpower and developer frustration.

Note: avidemux does much the same.

In short, linking to the movable feast that is ffmpeg would be a
nightmare. You basically would need to ensure that every myth system
had a compatible version of ffmpeg, compiled with the same options.
And with distros being what they are, that wouldn't be easy at all.

Its compounded by lack of releases from ffmpeg, most distros use an
svn version, and they change like the wind. I've even seen command
line options change, breaking scripts that rely on an earlier version
(like the need to put k at the end of a bitrate so a parameter that
was previously 384 becomes suddenly 384k).

So by importing the code en masse, you lose the speed with which the
ffmpeg devs incorporate fixes or new features, but gain stability and
control, as well as your own myth-tweaks.