It’s because you don’t want to reinvent the wheel all the time. It sucks doing it. Lots of effort. It’s much better to build on existing stuff and maybe improve it for your needs.
But that’s the thing: is there only one wheel? Maybe wheels are a bad metaphor here, but isn’t it weird, that there aren’t any fundamentally new concepts? Unix was developed basically during the preschool years of computing and we all just kind of stuck with its concepts.
If the underlying concept is good and was well thought out, it’s better to build upon it instead of reinventing it.
Look at the 4 stroke engine (and engines in general) many of the design concepts date back to the 1880s!
There’s other engine designs (ex:rotary engine) but the 4 stroke has over a century of testing, improvements, and refinements. A new design can adapt some of the refinements, but would have to catch up on decades of innovation and testing just to catch up!
On the Unix side, there’s the evolution of the Posix standard (which was based on Unix).
I would point out, by comparison, that piston engines are effectively obsolete for certain applications. Most aircraft operate on some type of jet engine, which involves the same core concepts of thermodynamics and aeronautics, but are still fundamentally different. They also optimize for different criteria, which is why neither jet engines nor piston engines hold a monopoly on any class of vehicle.
This is really stretching the computer metaphor. I think my point is that there will be room for rethinking paradigms as our applications of computers grow to include things that weren’t originally planned for. But in a mature technology there’s a lot of established precedent, and that’s not easily overcome. It takes something that can improve the field like jet engines made new aircraft possible.
Depends on the level of abstraction you’re looking at. Operating systems today are vastly more capable of organizing different provesses, distributing work amongst multiple CPU cores, CPU caches, etc. I guess the von Neumann architecture has just proven really successful in practice. And von Neumann machines require a certain set of capabilities in their OSes.
Maybe look at embedded systems, where we find a bit more variety. Things like DSPs or microcontrollers.
