Is OpenML suitable for low latency audio io?

The way I read the specification there is no way for an application to use a buffer size that is used by the hardware also, so it won't be possible to do e.g. double buffered io using select() to wait for buffer completion.