>>13there was once a time when people considered Unix to be the unstable, buggy of the bunch.
It still is. Compare it to VMS, VM/CMS, z/OS, MULTICS, MCP, NonStop, VME, OS/400, TOPS-20, etc. These operating systems all take advantage of the hardware as much as possible (usually the OS and hardware were designed for each other), but Unix ignores the hardware's special capabilities and works around it. AMD even neutered x86 segmentation so you can't run a MULTICS-like OS in 64-bit mode.
It's also weird to think that we've strayed so far from the true Unix model in this post-Unix era that there are Unix traditionalists cropping out of the woodwork in the form of BSD hobbyists and Suckless advocates.
They have been taught that Unix is the solution, but it's actually the problem. I read some years ago that there is an estimate of 8000 man-years of work for the Linux kernel. Instead of applauding them for their work, you should be asking ``Why is everything taking so much work and so much time even though they cut so many corners (like the OOM killer)?'' It's because their work is built on a bad foundation.
These are just pedestrian observations, from an outsider, mind you, but isn't it interesting (and, yes, I know I'm preaching to a choir) how the ivory tower plan9 project failed?
Plan 9 offers no improvements in productivity, time saved (programmer and computer), reliability, security, quality of life, user friendliness, modularity, speed, memory efficiency, consistency, OS design, or anything else. It's Unix without the compatibility or improvements added by non-Unixers (e.g. dynamic linking), which means there's absolutely no reason to use it unless you think Unix doesn't suck enough. The commands are mostly the same as Unix (slight alterations of the same source code), with no attempt to improve them. It still uses C, so no improvements there. Plan 9 today is still worse than commercial Unix from the 90s (i.e. the Unix the UNIX-HATERS hated). Solaris with CDE was more usable and better than Plan 9.
Plan 9 was too much like Unix to interest people who dislike Unix, but not compatible enough to interest people who want a better Unix-like OS (so they can run their Unix applications). In many ways, it was
worse than Unix. Linux does the compatibility better. If you have the chance to make something incompatible, make it an actual improvement. Anyone who is told that Plan 9 is ``OS research'' or ``latest and greatest'' (even if you only include OSes written in C) will end up believing that no improvements are possible, which is of course not true.
Plan 9 is smaller than other Unix-like OSes simply because it doesn't follow the standards for Unix-like OSes. The Unix weenies want academic operating systems to have POSIX layers, making them take too long to write and run too slow to be usable. They want everyone focusing on ``compatibility'' (with Unix, but not with other systems) before they even get it working. The first thing they ask when someone builds a CPU or OS is whether it can run C and Unix-like programs. They want new CPUs to be optimized for C and Unix-like memory models. It has nothing to do with the market because most of these projects never make it to the market. They are more interested in distorting the results of academic research, making things seem slower, harder to use, or more bloated by forcing POSIX and Unix-isms (like fork()) to be supported. Choosing C doesn't help things either. A better language would probably be 10 to 100 times more productive.
I'm sure a lot of people would want a Lisp machine if they knew it could run JavaScript and Python code (and other popular dynamic languages) just as fast as Lisp, but nobody is making them because the weenies say there is no market. They do not want there to be a computer where the lowest level languages are higher level than C, because everyone will start wondering why they still use C. It would make Linux, GCC, Plan 9, etc. look like counterproductive bloatware and a huge waste of time and money (which it is).