How to build an OS from scratch. This page is work-in-progress set of notes, as I talk to Devine and Nina.

6th of February, 2022

The uxn/varvara concept looks beautiful as it is. Especially with paradise, it asks not to bring there any of the conceptions one learned before.

Make what you can, or what you want?

As a unix-like systems engineer, I really miss fork/wait/kill in varvara, this way I could build something unix-like out of it. Tho I'm still worried I just do things the way I learnt things should be. This approach works well in commercial world, sure. Would it work in the world of art?

In classic, monolithic, unix-like operating systems, there is always a process, and the way to create a process is to fork itself. Then the child would exec into new application, much like launcher in varvara can run another ROM. Arguments to the new app are passed inside the syscall and straight to the int main(int argc, char *argv). Operating system there is no more that the library with the interfaces, syscall, protected behind interrupts.

Most of it can be done with uxn/varvara with its devices and vectors.

You'd want to create a uxn app that handles the routing of vectors to loaded applications in memory. You'd have to devise a scheme for the application format if you're really keen on doing multi-tasking in uxn.

This, however, poses a question yet to solve:

First is how the Devine desined it, quoting them

Uxn is the CPU, think of uxn like the 6502 residing inside the NES. Varvara is only a device API spec, think of it like the Commodore 64, or the Atari. An operating system written for varvara, in uxntal, would allocate time and devices control to applications in memory.

Second is how I saw it in real life: uxn is supposed to run on top of some kind of CPU, varvara is supposed to run on top of some kind of OS.

I could do a unix-like operating system for varvara. What do I want, and what my goal here?