A common way to proof De l'Hôpital's rule (for both function limits going to zero as x approaches point a) consists of "extending" the functions so they are defined in point a. Intuitively I can see why this extension is okay, but mathematically we're introducing and using a new assumption here - why is this allowed?