Even assuming you could get people to do the work (probably the real issue here) could a single schema syntax capture the semantics of the universe of documents

Even assuming you could get people to do the work (probably the real issue here) could a single schema syntax capture the semantics of the universe of documents that exist as PDFs? PDFs succeeded because they could reproduce anything.