Right now we're running the parser before c14n to deal with certain namespace-related issues. Find a way to avoid doing that.