thought that i should finally work on my cool OS architecture idea and then i quickly found out that the design i sketched is literally just Zircon
the high-level decisions are essenitally identical, and the main differences are like "how exactly the syscalls get executed" and "i can't be bothered add a handle array to the IPC mechanism so capabilities are just HMAC-signed global indexes"
it could still be potentially something to build because a Zircon-style OS kernel that is even more minimal and aimed at amateur OS research seems valuable, but it's a lot harder to justify spending my time on it now
@whitequark have you looked at our trustworthy saviour seL4 at all? It definitely checks the ‘minimal’ box even regardless of its whole proof system
@whitequark signed capabilities… how much of a performance overhead would come from that? A signature check every syscall?
@whitequark Zircon is a pretty good kernel. The VMO concept is well done. The two things it was missing when I was working with it were: 1) priorities and priority inheritance, and 2) a good mechanism on top of VMOs to implement things like file descriptors. The latter was being worked on last I heard. It was called streams at the time.