"Copernicus stuck very closely to the facts, but in Kepler I invented freely, and it's a much better book because of that."