- Need to add to dspctl
	- SNDCTL_DSP_SYNC
	- SNDCTL_DSP_CHANNELS
	- SNDCTL_DSP_SUBDIVIDE
	- SNDCTL_DSP_SETFRAGMENT
	- SNDCTL_DSP_GETISPACE
	- SNDCTL_DSP_NONBLOCK
	- SNDCTL_DSP_GETCAPS
	- SNDCTL_DSP_GETTRIGGER
	- SNDCTL_DSP_SETTRIGGER
	- SNDCTL_DSP_GETIPTR
	- SNDCTL_DSP_GETOPTR
	- SNDCTL_DSP_MAPINBUF
	- SNDCTL_DSP_MAPOUTBUF
	- SNDCTL_DSP_SETSYNCRO
	- SNDCTL_DSP_SETDUPLEX
	- SNDCTL_DSP_PROFILE

- Create sockets to emulate "opening" /dev/audio. Create libosscat,
  analogous to artscat or esdcat, such that "libosscat foo" is
  equivalent to "cat foo > /dev/audio" on Linux.

- Handle format conversion using AudioConverter (see
  /Developer/Examples/CoreAudio/Services/DefaultOutputUnit). Respond
  to format ioctl's. Should we add support for other conversions,
  eg: mu-law? (not too hard to do).

- Use fragments to minimize time mutexes are locked. Respond to
  fragment ioctl's.

- Respond to flow-control ioctl's.

- Will we implement all the mixer stuff? We should see what
  applications will need liboss, and see what they use--if they
  never use the mixer, or midi, we don't need it.

- Add a oss_real_time(int delay_usec) function, since so many
  audio-related packages will need it.

- Are there any flags and such that are needed for eventual support
  of other platforms (pure darwin, darwin x86, etc.)?

- Add liboss-config, or an alternate method of adding '-Dopen=___sys_open -Dioctl=___sys_ioctl -Dclose=___sys_close -include @prefix@/include/liboss/soundcard.h -Wno-precomp' to clients.
