reliable, and it's certainly simpler. The speaker ALSA support would
probably benefit from using the same backend at some point, and with a
similar backend for OSS we'd be in a good position to more closely
unify the support for the various different sound APIs.