Re: [sc-dev] SC_CoreMIDI questions

[stefan kersten <steve@xxxxxxxxxx>]

On Sat, Nov 29, 2003 at 07:58:05AM -0500, James Harkins wrote:
> > is the pitchbend range [-8192;8192[ or [0;16384]? 
> > 
>  
> [0;16384] - divide by 16384 to make it compatible with ControlSpec.

ok, thanks.

> > MIDIOut has a field uid (destination), which is passed to the 
> > sendmidi 
> > primitive. i don't fully understand the purpose of this 
> > parameter. does the destination id represent the inport of another 
> > client in CoreMIDI? 
> > 
>  
> It's a port number assigned by the OS. If you have a multiport 
> interface, each midi cable gets a different uid. You can get the 
> destinations by: 
> 
> MIDIClient.destinations.do({ |midi| [midi.device, midi.name, 
> midi.uid].postln }); 
> MIDIClient.sources.do...  // for sources

thanks, james, that makes it clearer.

the ALSA 'sequencer' on linux works like a virtual MIDI patchbay; each
application/device reqisters with the 'sequencer' and allocates a
number of i/o ports. events can be delivered either directly to a
client/port destination, or be broadcast to all ports subscribed to a
particular outport. connections can be made by external commandline
tools or gui interfaces.

therefore i'd like to extend the semantics of MIDIOut to behave as
follows:

* when uid is equal to a certain (negative) constant, MIDIOut
broadcasts to all connected ('subscribed') ports. this is the default
mode of operation with ALSA on linux. AFAICT CoreMIDI doesn't provide
a subscription system, so the implementation should trigger an error
or broadcast to all endpoints (generally undesirable) in this case.

* otherwise MIDIOut works as before, sending to a port id previously
retrieved from a MIDIEndPoint.

comments?

<sk>