[gst-devel] [RFC] invalid buffers

[Xavier Bestel]

Le sam 15/06/2002 à 10:19, Joshua N Pritikin a écrit :

> Generally speaking, i would expect the seek code to somehow:
> 
> 1. scan backward to the most recent I frame
> 
> 2. decode it
> 
> 3. skip the following P frames until exactly the desired P frame is found
> 
> For single-stepping forward and backward by one frame, i expect that a
> local I-frame cache or index would be essential for efficiency.

Yeah, but if the src can only seek forward ? Or if it can not seek, but
you start start at an unknown location ? You have no cache, no apriori
knowledge of frame boundaries, can only seek to an estimated byte
location. You'll certainly decode some useless buffers before building a
valid one.
I admit this is not the case that interests me, but perhaps in this case
we would want to avoid invalid frames too. If you want to store an
incoming rtp stream on-disk, and use it afterwards you'll have to clean
the invalid parts. Better have it done automatically.

	Xav