[ogg-dev] OggPCM version / header finalization

[Ralph Giles]

On Fri, Nov 11, 2005 at 07:03:51AM +1100, Erik de Castro Lopo wrote:

> One week is very little time to get public comment from other 
> interested parties.

Agreed, though things are becoming more sane quickly. :)

> I also haven't figured out whether this header is included in every ogg 
> packet or just the stream. From what I know of the Ogg container format
> the stream cannot be decoded if the file header is lost. Therefore adding 
> header data to each packet is just redundant.

It should be included just at the head of the stream, it's redundant to 
put it on every packet and I don't see any advantage to that.

> There is also no discussion on the wiki on how data is packed into an
> ogg packet. A sensible constrain would be that an ogg packet cannot
> be constructed with a partial frame (ie the samples for all channels 
> for a single sampling period).

Agreed. I think this is a sufficient contraint, though perhaps 
"applications SHOULD use a fixed block/packet size for predictable 
buffering." and discuss the tradeoffs of packet and page sizes between
latency and overhead.

 -r