[Comment removed by author]
For one, the GNU ML Compiler (GMC) would have optimized functional languages a ton. It would be faster than MLton… over a decade ago… with it even faster now. Ocaml would be blazing copying and improving on its techniques in backend. The amount of research into automated analysis and theorem proving for it would’ve made SML one of easiest languages to write bulletproof code in without much effort. SPARK Ada might not even exist since they were using ML. Rust would be Linear or Affine ML extensions which happened sooner than the current ones. CakeML would’ve happened much sooner since academics like CakeML team wouldn’t have wasted decades trying to deal with C or Pascal compilers. Combined with concurrent extensions and SCOOP-like model, these things would also run on multiple CPU’s or cores with easy-to-prove safety. It would’ve been a dream come true for high-assurance security esp trying to sell to commercial sector who would benefit easily.
Nah, I’m just kidding. Linus would’ve taken one look at that, decided C was the best language for max performance or style, and forked a BSD userland into a GPL project with his kernel. Optionally rewriting a bunch of Minix if that was too hard. Anything but SML. ;)