> Early Linux was exactly like this, too. I remember running Red Hat Linux 4.2 on my SPARCstation and having kernel panics right, left and centre. Several a day

> Early Linux was exactly like this, too. I remember running Red Hat Linux 4.2 on my SPARCstation and having kernel panics right, left and centre. Several a day, every day. No, not exactly like this. Even early versions of Linux used separate address spaces for (each) user process and system memory (like all, but the earliest Unix systems), preventing unprivileged user space processes to clobber system memory. A wayward pointer in an unprivileged application is strictly speaking undefined behavior, but on Unix systems typically causes a segmentation fault signal (by default terminating the offending application), not a crash of the whole system. That doesn't mean that there were no bugs in the kernel and crashes resulting from those, but even already mid-nineties, Linux, while perhaps not yet comparable with the likes of Solaris and Interactive Unix, wasn't any worse than SCO Unix and much more stable than the 16bit offerings from MS. Linux kernel crashes were rare (not as rare as today, thanks to continuous effort of hundreds of contributors and the prioritizing of the fixing of regressions), more often however users experienced out-of-memory situations, which before the addition of the oom-killer could effectively freeze a Linux system for a long time or even indefinitely. Also the X11-server or rather some graphics card driver (part of the X11 server then) wasn't quite of the same quality and when it crashed, terminated the user's session (all of the user's application started during that session). I.o.w., used as a small server, Linux was reasonable stable early on, used as a GUI desktop system not so much.