I favor the microkernel philosophy, which means removing code en masse from the kernel even at the expense of typical programmer convenience and possibly performance. I appreciate that TFA presents a counterargument that acknowledges the tradeoffs of the abstractions we choose, particularly in this case where a kernel file system is familiar and largely desirable. Is the increased flexibility and robustness of the microkernel design beneficial in the end? I say so, but even if I'm right, the alternatives should always be present for examination. Times change, and comprehending a time is hard enough. Operating system development subsumes software development, even as the reverse is true: it is the craft of abstractions and implementations, given means and ends.