[onetwo_threefour]

^ Yeah, Hawking and Mlodinow have an interesting discussion of model-dependant realsim in The Grand Design (just finsihed reading it) Their position is to assert that there is truth to any model as long as it agrees with observations. Kind of like what you're saying about the flat earth. On a small scale a flat earth model is useful because you can navigate on ordinary scales of speed (walking, running, driving) without worrying too much about the curvature of the earth. Of course if you want to use your GPS then you'll have to adapt your model because a flate earth model would cause massive errors in GPS. (They made a different argument about GPS and errors but I'm just making a similar example)

It was an interesting discussion that seemed quite apropos to this thread.