Always happy to see new input sanitization libraries, but most don't deal with the especially difficult problem of safe subsets of HTML. i.e. How to accept html from an untrusted source, apply whitelisted tag/attribute combinations, and ensure safe output.
