Gregory's ServerThe thing that I *really* want— every application is running inside of a sandbox I fully control, such that all attempted OS-level changes are accepted from the app's perspective but actually logged and tracked, and then I'm allowed pijul-style to decide which "patchsets" of sandboxes are simultaneously active for any one piece of software— might be kinda hard to implement and implement efficiently. But it seems *some* third option must be possible
|
2 comments
 @inthehands Did you know the original Mach, the experimental microkernel that gradually turned into Mac OS X, originally had a feature where userland program 1 could launch userland program 2 in a special way, and then from program 2's perspective program 1 was the kernel. Program 2 would think it's sending syscalls to the kernel but the syscalls are all going to program 1 and program 1 is allowed to freely decide whether to forward them on to the real kernel or just like… lie |
@mcc
I’ve wanted this too, and have thought about it. I pondered for a while a layered file system in which the user controls which layers are readable and writable by each app, and was halfway to implementing a nonsense version of Nix-in-the-OS before I decided this was not actually an easy problem.