Not that easy The reason is that so far the fits format is the only one that allows incremental reading (you don't have to read the whole file to get the pixel values). So, if we want to add new image formats to a integration algorithm, it should be re-engineered completely, and it would be less efficient than it is right now.