In the context of µkernel performance, the work of Jochen Liedtke and the numerous people working on L4 should not be underestimated. For a discussion of microkernel performance, I would recommend Liedtke’s paper “µ-Kernels Must And Can Be Small”. In this paper, he discussed why microkernels have to be small and why this in turn is required for developing a fast microkernel.

Nevertheless, thanks for the interesting story on the Hurd!

Listen, the Stallman/GNU defence is: “oh, it was too hard [to take the Mach kernel and DO something with it”. And yet, Apple took it and – together with the FreeBSD userland stuff – made the best Unix out there for the end user.

It’s clear to me Stallman did not have the chops for his lofty goals. When GNU wrote GCC, the standard book was the Dragon book. ANYONE, today, can probably write the eraly GNU stuff.

And Linux is the OS, it’s not GNU. HEre I’ll prove it: even before LLVM came along, I saw people use other compilers (Intel, for example). The userland NEVER had to be GNU tools. So that leaves you…Linux as the fundamental part.

Anyways, I hate Linux and GNU: these days, I’m on Mac OS X (never looked back) and FreeBSD – do have something like Grand Central Dispatch on Linux? Nope. Dtrace? Nope. ZFS? Nope. Jails? Nope. Etc, etc, etc.)

Live with it, GNU sucks, Linux sucks (well, ok, not entirely ;-))