We observe how a slight and even natural change in the Kurzweil-Henstock integral leads to a characterization of derivatives. We will also argue that the only way to characterize derivatives is by using some object or procedure which is at least as complicated as an integral.