Things I'd like to improve:

* Have better handling of suprasegmentals, rather than the hackish,
unintuitive way it's done now. Ideally, we should be able to specify which
segment is the head of a suprasegmental group (like a syllable or foot),
and let the suprasegment have properties of its own unrelated to the
segments beneath it.

* Provide a library of functions to be used in rules. These should provide
simple implementations of common linguistic processes, like assimilate(),
dissimilate(), etc.

* Someday in the far future, add Optimality Theory to the module.
