just parsing it (once ...) ought to be sufficient. we don't need to
validate the syntax beyond that anyway.