For a real "deeper" blog they could have given us some benchmarks comparing good implementations on FUSE and kernel space. They instead gave a meaningless diagram to argue about performance. Toward closed source binary kernel blob they go, and out I jump.