I'm always worried when someone wants to build an OS and builds a UI layer on top of it. It's a lot of work just to make a good kernel that's compatible with real hardware and is easy enough to port software for it.
While I agree an AR-first UI would greatly benefit from a real-time kernel, I think developing both separately would be the wisest path.
He can team up with a hw-manufacturer and together they develop the machine that was needed for so long.
Why not?
I once worked for Tuxedo, the beginning of that company was the question if they sell hardware to the books as well?
It’s a huge investment of resources. Remember Apple failed so many times to make a successor to MacOS classic they had to let Steve Jobs take over Apple. And Tuxedo didn’t make an OS from scratch.
Real innovation in OS is somewhere between impractical and impossible.
You might think POSIX is boring and bloated but if if you really change the relationship of applications and the OS you will need all new applications. If you like microkernels, replace POSIX with Mach or maybe L3. Hypothetically you could make a POSIX implementation which beats Linux for some particular special case but Linux has so much investment in it to deal with difficult problems that I don't see how you beat it.
Steam Deck is Linux with an implementation of the Win32 API. Hypothetically you could do better but Linux works and there are zillions of Win32 games on steam.
Meta Quest 3 is basically an Android phone you wear on your face, Apple Vision Pro is basically a Macbook you wear on your face. Is this optimal? No. But they can draw from a pool of existing applications as well as developer skills and tooling. Even if these platforms fail it's not a complete waste of time to learn how to develop for them.
Linux is very very different from POSIX. I've had to target actual POSIX and it's missing half the stuff you'd expect. It's really a minimum viable OS API.
Linux has new things like Uring and weird things like it's graphics API.
I still don't see what this has to do with the form factor though. That should be 100% user space maybe with a better scheduler or some process pinning. The thing is anything new eventually ends up in Linux. That's why it's kernel.org and not Linux.org.
Previously discussed here: https://news.ycombinator.com/item?id=44240265
I'm always worried when someone wants to build an OS and builds a UI layer on top of it. It's a lot of work just to make a good kernel that's compatible with real hardware and is easy enough to port software for it.
While I agree an AR-first UI would greatly benefit from a real-time kernel, I think developing both separately would be the wisest path.
He can team up with a hw-manufacturer and together they develop the machine that was needed for so long. Why not? I once worked for Tuxedo, the beginning of that company was the question if they sell hardware to the books as well?
It’s a huge investment of resources. Remember Apple failed so many times to make a successor to MacOS classic they had to let Steve Jobs take over Apple. And Tuxedo didn’t make an OS from scratch.
I'm guessing the custom kernel is because manufacturers don't want the GPL giving customers an escape hatch from their value add?
I really can't imagine any other reason you couldn't just use Linux.
Real innovation in OS is somewhere between impractical and impossible.
You might think POSIX is boring and bloated but if if you really change the relationship of applications and the OS you will need all new applications. If you like microkernels, replace POSIX with Mach or maybe L3. Hypothetically you could make a POSIX implementation which beats Linux for some particular special case but Linux has so much investment in it to deal with difficult problems that I don't see how you beat it.
Steam Deck is Linux with an implementation of the Win32 API. Hypothetically you could do better but Linux works and there are zillions of Win32 games on steam.
Meta Quest 3 is basically an Android phone you wear on your face, Apple Vision Pro is basically a Macbook you wear on your face. Is this optimal? No. But they can draw from a pool of existing applications as well as developer skills and tooling. Even if these platforms fail it's not a complete waste of time to learn how to develop for them.
Linux is very very different from POSIX. I've had to target actual POSIX and it's missing half the stuff you'd expect. It's really a minimum viable OS API.
Linux has new things like Uring and weird things like it's graphics API.
I still don't see what this has to do with the form factor though. That should be 100% user space maybe with a better scheduler or some process pinning. The thing is anything new eventually ends up in Linux. That's why it's kernel.org and not Linux.org.
>Real innovation in OS is somewhere between impractical and impossible.
This explains Steve Jobs' saying that you have to work on hardware. The next OS will at first run on GPU/FPGA only
Guess PowerPC wasn't enough of a difference (the OS innovation was thus a layer on top of BSD) & therefore neither is RISC-V, it would seem
> The next OS will at first run on GPU/FPGA only
Modern CPUs co-evolved with the software we run on them. If we really want to rethink software, we need to rethink the CPU itself.
That's what I meant.. probably I will use your phrasing next time, it's less kooky. Thanks in advance!