==============================================================================
= define generic event struct (probably Pd-ized input_event )
  something like:
  
struct input_event {
	struct timeval time;
	t_int type;
	t_int code;
	t_int value;
};


The question is whether the timeval is needed at all.  Linux and Darwin
support it.  Currently, I can only think of UPS PWR events actually using
timevals.



===============================================================================
= HID Manager Type/Usage/UsagePage -> Linux Type/Code mapping

 UsagePage


Misc Input/Generic Desktop X  == ev_rel/rel_x
Button Input/Button #1 == ev_key/btn_left


LED UsagePage => ev_led
LED Usages == Linux ev_led codes



==============================================================================
= device selection 

by # (1,2,...), generic name (mouse1, joystick2, tablet3...), or device name
("Trackpad", "Microsoft 5-button Mouse with IntelliEye(TM)", etc.)

 first get # working, that's probably the easiest

by #
------------------------------
GNU/Linux
    sprintf(x_devname->s_name,"/dev/input/event%d",deviceNum + 1);

Darwin
	 prHIDBuildDeviceList();
	 currentHIDDevice = discoveredDevices[gNumberOfHIDDevices];


==============================================================================
= figure out how to store device ID in obj struct

(in SC_HID.c its locID and cookie)

it should probably just store the Pd arguments



==============================================================================
= raw values vs. calibrated

- relative axes should probably be raw data, since its pixel data, but then
  this causes problems with sensitivity across different mice.  The mouse
  sensitivity would probably best translate as resolution, ie calibrated data,
  rather than sensitivity, ie raw data.

- absolute axes should be calibrated, so that the same positions on different
  devices map to the same value

- but then, I could just create an [autocal] object in Pd, and have [hid]
  output only raw values.  This would probably be best in keeping with the
  idea of having [hid] giving as low level access as possible, then using
  [mouse], [joystick], [tablet], etc. objects providing a nice, consistent
  interface to the respective devices.



==============================================================================
= pollfn for mouse-like devices

- determine whether using a pollfn is actually better than using a t_clock

- any device that acts like a system mouse can be used with a pollfn, since
  the mouse data will go thru Pd's network port, triggering the pollfn.

- this is probably unnecessary since the t_clock seems to run well at 1ms delay



==============================================================================
=  make generic functions for the basic actions

    releaseDevices()
    buildDeviceList()
    buildElementList()
    getEvent()

 - make Darwin HID Manager -> Linux input event convertor functions



==============================================================================
= [close( message might be totally frivolous.

- what does Max's [hi] do with that?

- when would you need to close a device, yet still have the object there?
  [open( closes the previous device anyway.



==============================================================================
= function return values

- most functions probably do not need return values

- return (1) seems to be the default on many functions


==============================================================================
= control input messages

- the [delay( message should be replaced by the [poll( msg

- should [poll( also start things, or should it just set polling time?

- are [start( and [stop( needed?  is 0/1 enough?