All existing python API's which manipulate audio data are already interoperable, provided they use some generic python type to represent the data (strings, arrays, etc...). There are several advantages and disadvantages to this approach.
FIXME: Do some real research and update this section. (This is currently my out-of-date impressions on the matter.)
So far I've only looked at a few modules in the standard library which deal with audio (audioop, sunau, wave, and sndhdr), and that was a while back (python 1.5, perhaps?). At that time, those modules made use of strings to expose data.re regular expression
module, etc... The array module provides a better
solution; arrays are mutable, allowing more efficient processing of
data, and they can represent data of different sizes.
Even more useful are Numeric Python matrices.
Like array objects, these are mutable, but better yet can represent
multi-dimensional data (which has applications in audio, say for
multi-channel signals, or signal processing transforms). Even more
important, there is a wide range of critically important routines for
signal analysis, synthesis, and manipulation which operates on
Numeric matrices. Although the Numeric
routines are perhaps at a lower-level than average audio developers
may want, they provide a powerful and flexible set of linear algebra
primitives which are the basis for most audio effects and analysis.