[NUT-devel] NUT samples / NUT muxer?

[Rich Felker]

On Mon, Mar 27, 2006 at 10:17:36AM +0200, Oded Shimon wrote:
> > As I said, I started implementing a (de)muxer from scratch.
> 
> Ah, yes, it's even still quoted in this mail. :) BTW, you should read:
> http://wiki.multimedia.cx/index.php?title=NUT
> 
> It has a detailed explanation for keyframe exact seeking in NUT...
> 
> Your de/muxer from scratch is gonna be in lavf?..

No, we want to update the lavf implementation to (with someone else
doing that work) in order to have at least 2 independent
implementations based on the spec. This will help us catch any
bugs/mistakes and assure users of interoperability and feasibility of
implementing from the spec.

As for the avi/ogg demuxers in nutmerge... my understanding is that
they're supposed to be minimal and accept only good, fully-valid
files. We can leave handling of broken files to more advanced
implementations as long as nutmerge rejects the files..

> > >> OK, so what's with this:
> > >>
> > >> $ ./nutmerge
> > >> ==============================================================
> > >> PLEASE NOTE THAT NUTMERGE FOR NOW CREATES _INVALID_ NUT FILES.
> > >> DO NOT USE THESE FILES FOR ANYTHING BUT TESTING LIBNUT.
> > >> ==============================================================
> > >> bleh, more params you fool...
> > >
> > > read README
> > 
> > I did:
> > DO NOT USE NUTMERGE FOR _ANYTHING_ EXCEPT TESTING!!
> > IT CREATES COMPLETELY INVALID BROKEN FILES, WHICH JUST "HAPPEN" TO WORK!
> > 
> > So you're saying the files are OK after all, at least for some inputs?
> 
> Well, in a very hackish way, nutmerge actually DOES generate valid files in 
> the case of MPEG-4 ASP with vbr MP3 in AVI,

As long as the program writing the file did not decide to pack N mp3
frames into each avi packet, where N>1..

> and Vorbis in Ogg. But it's 
> bad hack and could easily fail in some cases. (like, doesn't handle zero 
> byte frames from mencoder)

Yes..

Rich