Hi,
> > > Then the midi output layer is not called directly by the midi device > > > drivers. The midi output is responsible for parsing > > > the blocks and do the deltatime waiting. Right now, waiting is done by the sound server, because this is the layer implementing the communication mechanisms. Using the draft architecture, communication would become unidirectional, with bidirectional communication being a feature to enhance synchronity. We'd still have to decide whether to do timing in the sound server or in the driver- and, considering a simple example (the proposed SDL PCM callback driver and the UNIX sound server using ossraw), we can see that we cannot handle timing in the MIDI output layer- ossraw wouldn't know whether to return immediately (SDL PCM callback) or to wait (possibly polling for events on the input pipes, something I always considered quite convenient in the current implementation). And I don't see why EVERY MIDI output layer should contain code to distinguish between these uses. Of course, I have to admit that the SDL PCM callback server is only a proposal at this point... llap, Christoph
