[gst-devel] gst-python and GstXOverlay

[Ronald S. Bultje]

On Wed, 2004-11-10 at 12:42, Martin Soto wrote:
> IMO, the whole combination mpegparse/mpegdemux/dvddemux is a mess. It
> needs serious refactoring. It is good to know that you guys are putting
> some effort into it. 

*nod*. However, there's many problems, we can't just put some monkey in
front of a computer and expect him to fix it. mpegdemux works around a
lot of problems with MPEG files correctly, it's a really good element,
even though it's somewhat complex at first sight. I know only few
problems with mpegdemux:

* seeking and timestamp discont handling doesn't work together nicely
yet. However, we need that because MPEG files can and are allowed to
have random breaks in time continuity.
* it's sad that handling of dvd-spec code was moved upwards.

I'm not convinced of dvddemux, however. Seriously, it might be a nice
idea at first sight to separate the two, but does it really buy us
anything? It has many problems in my opinion, some of which I cannot fix
easily.

* it used to disable discontinuity handling, which gives horror effects
when feeding it such movies.
* it highly depends on dvdnavsrc interaction
* the current_* pads are workarounds for scheduler bugs in non-official
schedulers. Opt handles renegotiation correctly.
* DVDs can seek internally and switch language via menus, we need a
better way to support this than those hacks, imo.

Anyway, I have your email somewhere and will come down to it in more
detail, but not now, people want me to fix ALSA first.

Ronald

-- 
Ronald S. Bultje <rbultje at ronald.bitfreak.net>