Perhaps microkernels WERE the future, after all. Let's face it: companies still DON'T GET IT that they are more secure opening their code for others to see and modify than doing the opposite. They still don't get, either, that they can actually make money buy doing so.
Let's recognise, also, that MBA's WILL NOT, at least in the short term, recognise these benefits.
So, sadly, proprietary code will be with us for a long, long time...

So now perhaps we should check again the microkernel ideal. Yes, they are way slower than monolithic kernels - there's an inavoidable performance penalty for the way in which they work. But you gain in security when you just CAN NOT TRUST THE CODER!!!

Just as in this case: bugish and proprietary (these words are almost synonims, to me) drivers run in user space, not in kernel one - so there's a limit in the damage inflicted if it has a remote hole in the code or if it's buggy.

And, after all, hardware speeds up continually... so in a few years the performance penalty may be just negligible!! Can anybody ask kindly RMS if he can just stop experimenting with OS design and just FINISH CODING HURD NOW? The world does need an operational and libre microkernelish OS.