It is by far the cheapest option. Remember it was invented back in the day when RAM was expensive. “Spawn a process” is actually “replace this process”. The idea is you don’t want the old process hanging around wasting RAM, the old process memory is freed and replaced with new executable code, which carries on running. What if you do actually want the old process to hang around and wait for the new one, for example what if you’re a shell? This is where you add fork(). Fork splits a process in two, both processes share the same memory and if all the process does is read things you are fitting two programs in the space of one. Fork thousands of times are you have loads of workers all sharing the same memory. What stops them stomping on each other? As soon as they write something - it triggers an interrupt that quickly makes a copy of that tiny bit of memory, which is then written over. Now you know what “shared” means in a process - memory that hasn’t been written to. If you leave out the copy on write bit you get threads. Back to the shell though. You want to run another program and have the shell still run. So you do both - first you fork() into two identical processes, then in the second process you execve() to replace the second copy of the shell with the program you want to run. Shell in the first process will wait for the first process to be done, or maybe it won’t - that’s “putting it in the background”. Now, what happens when you are resource constrained or want to be fast, you can skip the execve() step and just have the code inside the shell itself. Well done, you just invented busybox. You get identical behaviour, you can wait for the second bit of your own code to finish, or “put it in the background” and have it do two things at once. Lots of options, super cheap.