An engineer at Canonical has poured cold water <https://www.theregister.com/2020/10/15/even_2020_cannot_bring_forth/> over the suggestion by Eric Raymond that Microsoft might move to substituting a Linux kernel over its own in the Windows OS. Sure, the idea seems complicated and expensive, and very likely troublesome, too. But since when have Windows upgrades not been troublesome? Even updates between different versions of what is ostensibly all the same OS, Windows 10, have been far from smooth lately. Attempting the level of compatibility needed for every Win32 API and syscall would be a huge undertaking. Consider that the WINE folks have already provided a usable fraction of this, with resources that are only a drop in the bucket compared to those available to Microsoft. And as for the vast array of drivers enjoyed by the NT kernel, Barnes reckoned that shunting them into Linux would cost millions. Sure. But would it cost as much as having to continue maintaining them? Particularly since most of those drivers would not be needed, because Linux already has its own native ones. Furthermore, rebasing Windows on Linux could have the ironic effect of stifling competition should a kernel monoculture take hold. Interesting to discuss this use of the term “monoculture”. Linux is already adaptable to a wider range of platforms and applications than Windows can manage. This in spite of the fact that the Windows NT kernel has been fragmented into a number of different versions, including about 5 or 6 different ones for ARM, with little or no compatibility between them. Linux is able to manager better interoperability than that, even between versions for entirely different architectures.