For a Dynamic Model of textual variation: What Do We Need? and How to Build a Textual Data Model?