[Theora-dev] Ogg spec

[Timothy B. Terriberry]

Ralph Giles wrote:
>>Yes, that was the intent, as I couldn't imagine what a true 'nil' page 
>>should do, as opposed to the 0 lacing value, which is required by rfc if
>>a packet ends exactly on page end.
> 
> 
> Zero length packets are possible. Theora for example uses them as
> a shorthard for 'repeat the last frame'.

Well, that's distinct from a _page_ that contains no data. While one use 
for such a page would be to mark the end of a packet which exactly 
filled up the previous page's lacing table with 255's, another use would 
be to mark the current granule pos of a logical stream even when there's 
no packet data to store. When we were considering sorting by start 
times, one of these could've been added to the end of the stream to mark 
its total length.