How the value is stored internally is not necessarily relevant to how you would access it in GLSL. Using normalised colour values (0-1) is much easier in practice. Is there some reason you want to manipulate pixel values in your pixel shaders in the range of (0-255)?