But what if you just want to do a simple decoding transform to get a stream of Unicode code points from a steam of bytes? If your definition of a stream is that it has UInt8 values, that simply isn't possible. And there's still gonna be waaay too many code points to fall back to an async iterator of code points.