The problem is that the OSS model of needing to call read/write for audio I/O is pretty hard to fit with the expected pro audio model of "get a callback called

The problem is that the OSS model of needing to call read/write for audio I/O is pretty hard to fit with the expected pro audio model of "get a callback called when data is ready", likely through some hardware interruption.