All,

Please find attached a conceptual new API between the audio service and 
whatever audio framework we have (ALSA, GStreamer, PulseAudio, or 
whatever). :-)

This new API would be defined under LGPL, so that guys like me can use 
it for some in house closed source projects( Luiz, and Johan,
i hope that's OK with you :-) ).

The principles of the API are the following:
  - define a set of messages (that would be the functions in 
traditionnal C programming)
  - define a set of structs, *one per message* (that would be the
list of arguments in traditionnal C programming).
  - define a reduced set of a few helper functions to bootstrap the 
message exchange process, as well as to retrieve the data fd through 
ancilliary data.

The usage of the API should be relatively straightforward : you would 
use bz_audio_service_open() to retrieve the file descriptor, and then 
integrated it in your main loop or whatever.

Then you would read fixed size data chunks that you would cast to 
t_bz_audio_msg_header to retrieve the message type. Once you have the 
message type, you would cast again the result to a structure that 
matches the message name, and obtain the list of parameters that are 
relevant for this message.

I didn't want to bring more features in the API as otherwise we would 
start to implement toolkit-dependant behaviours (such as need a 
mainloop, or a need for GLIB, or ...)

I think this API is a lot easier to maintain in the long term that what 
we have today, where structure are shared between messages, and some 
messages have more that one data structure. :-)

If we can reach an agreement quite rapidly (that would mean today :-)), 
then i could start implementing it right away and hopefully have a 
mostly working implementation on sunday night.

As always, comments are welcomed !

Cheers,

Fabien