Not buying it. The colossal attack surface of a giant monolithic kernel should not be controlling an automobile. No offense to Linux or anyone who works on it, it's great, you just can't rely on it with your life. It was never designed to be.
then remove that "colossal attack surface" by compiling a custom kernel and utilities that only includes the features the product needs. create a system tuned to the exact product to make it extremely reliable. almost everything electronic you see in commercial use is Linux because of this very fact.
Many medical devices run Linux.
Toyota, Tesla, Audi, Mercedes, and Hyundai vehicles use Linux.
you certainly can rely on it for your life and nearly every electronic device you use will use some derivative of it.
Can you please name at least one product that directly uses and relies on Linux for a safety-critical system? Those vehicles only use Linux for the ICE and not the ECU. And those medical devices are not directly controlling patient-facing instruments via Linux as far as I know.
I'll gladly provide 2 examples that I know of :
The entirety of the Tesla OS is based on Linux, meaning that their proprietary autopilot program running on their OS is directly dependent on the Linux kernel for its core functionality.
Tesla has been working to upstream support for their Full Self-Driving (FSD) SoC into the mainline Linux kernel.
Tesla's Autopilot HW3 computer is running Linux kernel 4.14.
Tesla has been enabling the Tesla FSD SoC for the upstream Linux kernel over the past year.
You can compile a small custom kernel with just the needed drivers. Also an LTS kernel. This is likely exaggerated
Yes but that small kernel is still pretty massive and all running at ring0
For sure. But it works. This is the point.
I am a total redoxOS fanboy but I have not even got that thing to launch in a FOSS VM.
There's nothing to buy, Linux is already in ECUs on the road. Elektrobit is just developing yet another option.
Do you have any proof for that claim?
What you need is redundancy
Hurd durr /s
I come from a MCU background and feel the same way. Linux kernel is for consumer level stuff. For serious machinery, I choose a real-time OS like FreeaRTOS. Less code, and more low level code makes it easier to review, maintain, and have less chance to break.
Ive been seeing Linux take a more controler of controllers kind of role. Handling updates, networking, complex logic, logging, metric, etc.
It'll be interesting to see where it goes. On one hand ASICs, FPGAs, and microcontroller are getting easier than ever to program, its still not as easier as having a full Linux OS to build on.
Not buying it. The colossal attack surface of a giant monolithic kernel should not be controlling an automobile. No offense to Linux or anyone who works on it, it's great, you just can't rely on it with your life. It was never designed to be.
then remove that "colossal attack surface" by compiling a custom kernel and utilities that only includes the features the product needs. create a system tuned to the exact product to make it extremely reliable. almost everything electronic you see in commercial use is Linux because of this very fact.
Many medical devices run Linux.
Toyota, Tesla, Audi, Mercedes, and Hyundai vehicles use Linux.
you certainly can rely on it for your life and nearly every electronic device you use will use some derivative of it.
Can you please name at least one product that directly uses and relies on Linux for a safety-critical system? Those vehicles only use Linux for the ICE and not the ECU. And those medical devices are not directly controlling patient-facing instruments via Linux as far as I know.
I'll gladly provide 2 examples that I know of :
The entirety of the Tesla OS is based on Linux, meaning that their proprietary autopilot program running on their OS is directly dependent on the Linux kernel for its core functionality.
Lastly, NASA's Mars helicopter.
You can compile a small custom kernel with just the needed drivers. Also an LTS kernel. This is likely exaggerated
Yes but that small kernel is still pretty massive and all running at ring0
For sure. But it works. This is the point.
I am a total redoxOS fanboy but I have not even got that thing to launch in a FOSS VM.
There's nothing to buy, Linux is already in ECUs on the road. Elektrobit is just developing yet another option.
Do you have any proof for that claim?
What you need is redundancy
Hurd durr /s
I come from a MCU background and feel the same way. Linux kernel is for consumer level stuff. For serious machinery, I choose a real-time OS like FreeaRTOS. Less code, and more low level code makes it easier to review, maintain, and have less chance to break.
Ive been seeing Linux take a more controler of controllers kind of role. Handling updates, networking, complex logic, logging, metric, etc.
It'll be interesting to see where it goes. On one hand ASICs, FPGAs, and microcontroller are getting easier than ever to program, its still not as easier as having a full Linux OS to build on.