> That means that the HARDWARE doesn't support variable rates.

Sure.

> To do it right you have to do sample rate conversion in software.
> This is rather expensive to do if you want high quality.

Yeah, but not playing it at all or playing it wrong is no viable solution.

We need a consensus on what to do in this case:

 a) convert (maybe without even telling userland) in the kernel
 b) tell userland this failed and fix applications to convert themself
    (or decide to use some other source)

I don't like (a) and prefer (b), maybe with some system supplied conversion
library or a set of shared libraries, each providing one special conversion.
How do other OSes handle this?

And actually I don't care for bad output quality on el-cheapo on board
soundchips.


Martin