the implementation I had in mind for solving the API problems would be designed for maybe a few hundred bytes of data, not for serving whole files.