Re: mm framework

Enrico Weigelt wrote:

> * Steve Lhomme <[email protected]> [2004-01-03 18:22:58 +0100]:
> 
> <snip>
> 
>>>What we have to consider on the design of an audio api are exact latency
>>>measurements, especially if we're gonna work w/ networked audio. At least
>>>we need something like timestamps (which go against ntp-synced rtcs)
>>
>>I highly agree !!! The latency indication is one of the key of the 
>>success of BeOS for audio apps. And IMO this should apply to video as 
>>well (for a modern API).
> 
> 
> Yes, so we need something like a getLatency() method on each codec,
> device or whatever we call it. A networked media driver should implement
> some timing (or perhaps use a static configurable value).
> 
> All together we should use some timestamping for each audio and video
> frame for better synchronizing. Latency information should be used to 
> predict optimal buffering or for faking timestamps where they're not
> available. 
> 
> Does this sound okay ?

That's what is done in BeOS.
This is very usefull for predictability and efficient caching. I think 
it should make seeking even more accurate (when programmed in a stream 
like in DVDs).