Just being able to get 1/2, 1/4 or 1/8th of the image resolution would already cover most of the srcset use case and the results are good enough for many existi

Just being able to get 1/2, 1/4 or 1/8th of the image resolution would already cover most of the srcset use case and the results are good enough for many existing image decoders to already use this optimization for the decoding if not the network part.