A lot of comments here focusing on whether this is a good thing or not, but I'm worried about a more fundamental issue: how is it supposed to happen in the first place?
As I understand it, the proposal is that every bit of software running on a device that is currently considered "firmware" ought to be in the OS instead. What does the story of how we get from here to there look like?
1. We convince manufacturers that it would be a great idea to release thorough documentation for all of their hardware products, and that way open source kernel developers can write their own firmware and merge it into open source kernels. Bryan gestures in this direction several times throughout the talk, claiming that this would be good for the hardware manufacturers because it would enable us to by more of their products. But ... no one's really convinced this is going to work, right? The hardware companies don't want to release documentation for the exact same reason they don't want to open source their hardware; it's a barrier (however small) against leaking trade secrets.
2. We focus our efforts on small companies like Oxide building integrated, top-to-bottom systems designed to run without firmware. This is great because it enables BIOS-less system design from the ground up and can be every bit as open source as we want it to be. But this means waiting and hoping that a tiny number of companies manage to reverse engineer the chips (like the AMD processors discussed in the presentation), which probably means running years behind and an extremely limited selection of hardware. This is not a solution that is going to deliver my next laptop, let alone my grandma's next laptop.
3. We convince manufacturers that EFI is a broken, sucky, insecure mess and include and let them help lead a process designed to replace it. This could maybe happen, we did get rid of BIOS (in favor of EFI) after all, but what is the end result likely to look like? It looks like binary blobs running unknown code necessarily tainting nearly every Linux system under the sun. And that's if they don't just decide to integrate with Microsoft and ignore Linux completely.
That's the thing about isolation. It has problems, as this talk discusses, but it also means that (putting aside a great number of bugs) an OS like Linux can exist without needing to reverse engineer firmware for the enormous amount of hardware it runs on. It's great to be able to run an OS with few-to-no blobs active, e.g. with open source Wi-Fi and graphics drivers. Closed firmware is in theory a security nightmare, but in practice it's not that often that anyone managed to exploit these vulnerabilities. Can you imagine having to run closed source Wi-Fi firmware in the Linux Ring 0?
As I understand it, the proposal is that every bit of software running on a device that is currently considered "firmware" ought to be in the OS instead. What does the story of how we get from here to there look like?
1. We convince manufacturers that it would be a great idea to release thorough documentation for all of their hardware products, and that way open source kernel developers can write their own firmware and merge it into open source kernels. Bryan gestures in this direction several times throughout the talk, claiming that this would be good for the hardware manufacturers because it would enable us to by more of their products. But ... no one's really convinced this is going to work, right? The hardware companies don't want to release documentation for the exact same reason they don't want to open source their hardware; it's a barrier (however small) against leaking trade secrets.
2. We focus our efforts on small companies like Oxide building integrated, top-to-bottom systems designed to run without firmware. This is great because it enables BIOS-less system design from the ground up and can be every bit as open source as we want it to be. But this means waiting and hoping that a tiny number of companies manage to reverse engineer the chips (like the AMD processors discussed in the presentation), which probably means running years behind and an extremely limited selection of hardware. This is not a solution that is going to deliver my next laptop, let alone my grandma's next laptop.
3. We convince manufacturers that EFI is a broken, sucky, insecure mess and include and let them help lead a process designed to replace it. This could maybe happen, we did get rid of BIOS (in favor of EFI) after all, but what is the end result likely to look like? It looks like binary blobs running unknown code necessarily tainting nearly every Linux system under the sun. And that's if they don't just decide to integrate with Microsoft and ignore Linux completely.
That's the thing about isolation. It has problems, as this talk discusses, but it also means that (putting aside a great number of bugs) an OS like Linux can exist without needing to reverse engineer firmware for the enormous amount of hardware it runs on. It's great to be able to run an OS with few-to-no blobs active, e.g. with open source Wi-Fi and graphics drivers. Closed firmware is in theory a security nightmare, but in practice it's not that often that anyone managed to exploit these vulnerabilities. Can you imagine having to run closed source Wi-Fi firmware in the Linux Ring 0?