[XviD-devel] [Fwd: Data partitioning for encoder]

[Skal]

	Gutten Tag Sigdrak,

On Thu, 2005-04-21 at 01:17, sigdrak wrote:
> Bonsoir/Good evening,
> 
> Skal wrote:
> > 	Anyway, this step toward data partitioning raises few questions:
> > 	a) we'll have to add DP support in decoder too, sure. Or people
> > 	won't understand a file created with XviD... can't be played with
> > 	XviD.
> 
> Indeed, as with time tools may expose such a feature. But then, most 
> users, except with streaming in error-prone environments in mind, will 
> have no use for such a feature. As for video packets and reversible VLCs 
> in fact.

	Don't under-estimate the appetite for weird features ;) As soon
	as there's any new option popping up, people *will* use it. Even
	just for the fun of trying it.

> 
> But to quote my very unclear self:
> "I have decoder-side pending code, but unfortunately, it is merged with 
> my current work on video packetization."

	ok.

	Anyway, i've tried your patch, and it seems you forgot to
	include some updated header file(s). Bitstream.c isn't
	compiling at line 706. Somehow, some of your decoder-side
	work slipped through in the patch. Nevermind. I'll just
	comment out the offending line.
	Otherwise, the patch seems ok. I'll later commit it
	protected with some '#ifdef EXPERIMENTAL_DATA_PARTITIONING',
	so it won't interfere with the rest until proven ok.

> 
> > 	b) What about reversible VLCs?
> 
> This is indeed the next thing in DP. But they aren't mandatory. 

	IIRC, the standardized Profile where DP is required also 
	makes support of RVLC mandatory...

> In fact, what I really see as missing is error concealment. As far as I 
> can see, XviD has none, and nothing has been designed for it.

	Indeed. This is rather historical: XviD's main use
	is storage, not really an error-prone channel.

Skal, "je passe sous un tunnel, ca va coup.." ;)