There is a standard algorithm to decompose a unicode compatability character to its canonical form. The canonical form may consist of one character or more than one characters.
Here I ask for the reverse lookup. Given the canonical form, tell me if there is any possible compatability character.
I want this functionality to be a web service, a commandline, a programmable library or a plaintext database.
I want to fairly keep up with newer versions of unicode, like 4.1 or 5.0.