“CHERI” is a project at Cambridge University based on a very old concept for implementing memory/process protection on computer hardware: that of “capabilities”. Instead of having specially-trusted kernel code run in a privileged mode that is allowed to do pretty much anything, capabilities allow for a more decentralized model, with distribution and separation of different areas of trust. The capability concept fell out of favour, because its implementation overhead was greater than the processor-privilege-level (also often described as “privilege rings”) concept that is so commonplace today. But with the increase in security vulnerabilities like Meltdown and Spectre, there is increasing interest in alternative approaches that might reduce complexity, hopefully with less of a performance cost than the fixes that were needed to mitigate those vulnerabilities. ARM Ltd’s experimental “Morello” chip is an implementation of the CHERI concept. And it has got as far as running CheriBSD, which is a specially-adapted variant of FreeBSD. This now includes Wayland as the GUI server, and it provides a KDE Plasma desktop on top of that, according to <https://www.theregister.com/2022/07/26/cheri_computer_runs_kde/>. All this code has been tweaked, where appropriate, to remove the assumptions of arbitrary access to process memory allowed by conventional C pointers, in favour of the more restricted access controlled by capabilities. The CHERI project started with MIPS-based hardware, and besides being on ARM now, it looks to be coming to RISC-V as well. By the way, the Linux kernel has already (mis)used the term ”capability” to refer to an entirely different concept <https://manpages.debian.org/bullseye/manpages/capabilities.7.en.html>. Those familiar with the process-privilege system on old VMS from DEC may notice some resemblance ...