There are upsides and downsides to implementing multiple or single file solutions. There used to be limitations on menu items for example. I believe these are long gone. One of the downsides of multi-file solutions involves complicated context. Each file has it's own relationship graph for example. Given that a file can contain a virtually unlimited number of tables it would seem a single file solution would be preferable. Using a Separation Model (one file for raw data and another for the interface) would be best in my opinion.