Yesterday I decided to switch to using HTML + RDFa (using BIBO) as the
native/internal model. So you feed the processor a CSL style + some
JSON data, and it spits out HTML + RDFa (with other output modes
possible).
There will be details to work out, but I think it should work nicely
(aside from issues with RDFa; like I don't think I can represent a
bibo:authorList correctly).
<http://github.com/bdarcus/citeproc-py/>
Bruce