>>33797Mach has been an outdated research design since the 90s, even moreso when there is close to thirty additional years of academic and industry experience behind us. The only reason Apple keeps their implementation around is, that they're intent on holding onto XNU, and the fact it's mostly used to emulate Unix doesn't put much stress on the slow IPC. Linux has for the time being absolved the GNU project of the need to ship it's own kernel, so HURD has been allowed to limp along the same dogged development path, that lead to its irrelevance in the first place.
The GNU model has often been described as cathedral-like, yet even their best software tends to be somewhat overengineered. In practice the fast-moving "bazaar" software in use by Linux today is usually propped up by Redhat employees or software engineers employed by other companies with a stake in the ecosystem. If the absolute state of modern GCC is any indication, then too many GNU projects move at a pace that's simply not conducive to high-quality software with their current pool of developers.
This state of things feels extremely frustrating, when it has already been shown there is another way, from enduring if mismanaged corporate products like Solaris or QNX, over the BSDs and other venerable communities, to smaller software projects carefully cultivated by lone developers. Many of these have almost passed into the realm of public infrastructure (recall Minix3 use in IME), precisely because they are reliable and don't require a near-constant stream of development to keep functioning.
The scope of the GNU project in combination with its license has always seen itself as the real movement, that would enable access by the public to high-quality software and ideologically it is still the closest to this in the current software ecosystem. However the last few decades force us to soberly regard it as a net failure. Few notable GNU projects are currently on par with the corporate-driven project landscape, permissively-licensed or not. Projects like the almquist shell, busybox, libarchive or libedit, which were started by individual developers, have routinely shown even their most reliable and widely-used code to be comparatively subpar in execution. Unfortunately i can't help but conclude, that up to now both small-scale, "artisinal" software development and the limited cooper
Post too long. Click here to view the full text.