4 ms·

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 ot

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.

PaulHoule 1y ago
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.
- mathiaspoint 1y ago
  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.
- gsf_emergency_2 1y ago
  >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
  - rbanffy 1y ago
    > 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.
    - gsf_emergency_2 1y ago
      That's what I meant.. probably I will use your phrasing next time, it's less kooky. Thanks in advance!
      - rbanffy 1y ago
        I'd go further - we'll need to rethink the CPU, the memory (we'll need lots of memory regions with exclusive access, a way to manage scratch memory, etc) and IO (because there will be a lot going on at any given time).
        gsf_emergency_2 1y ago
        (Aside from the scratch or IO maybe this isn't all that heretical :) there seems to be some nascent interest and work out there: https://en.wikipedia.org/wiki/In-memory_processing#Application_of_in-memory_technology_in_everyday_life https://en.wikipedia.org/wiki/In-memory_processing#Applicati... XenevaOS could at least have moved in that direction..
        rbanffy 1y ago
        As I said before, even a ground up OS for current hardware is a colossal endeavour. Coupling that to a rethink of the hardware as well is clearly better suited to academia for now. I already find it intriguing that the pre-boot hardware bring up for RPi boards is done by the GPU before the ARM cores are enabled.
        gsf_emergency_2 1y ago

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.

PaulHoule 1y ago

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.

mathiaspoint 1y ago
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.
gsf_emergency_2 1y ago
>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
- rbanffy 1y ago
  > 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.
  - gsf_emergency_2 1y ago
    That's what I meant.. probably I will use your phrasing next time, it's less kooky. Thanks in advance!
    - rbanffy 1y ago
      I'd go further - we'll need to rethink the CPU, the memory (we'll need lots of memory regions with exclusive access, a way to manage scratch memory, etc) and IO (because there will be a lot going on at any given time).
      - gsf_emergency_2 1y ago
        (Aside from the scratch or IO maybe this isn't all that heretical :) there seems to be some nascent interest and work out there: https://en.wikipedia.org/wiki/In-memory_processing#Application_of_in-memory_technology_in_everyday_life https://en.wikipedia.org/wiki/In-memory_processing#Applicati... XenevaOS could at least have moved in that direction..
        rbanffy 1y ago
        As I said before, even a ground up OS for current hardware is a colossal endeavour. Coupling that to a rethink of the hardware as well is clearly better suited to academia for now. I already find it intriguing that the pre-boot hardware bring up for RPi boards is done by the GPU before the ARM cores are enabled.
        gsf_emergency_2 1y ago