Re:Spectre mitigations murder userspace performance
by staticassertion on Saturday June 19, 2021 @03:50PM
(#27563817)
I don't think this would solve anything. This is like putting Word macros behind a "do you want to run this macro?". Yes, yes they do want to run it.
If people were willing to put every webpage behind a "do you want to run this webpage" we'd see them all using noscript already - virtually no one wants that experience.
> largely under the theory that we can do it safely with correct design
I blame operating systems and hardware vendors tbh. It's their job to make this safe, and they do a pretty bad job of it. Browser vendors have had to pick up a massive amount of slack to try to compensate, to the extent that browser teams have to make major patches to the Linux kernel.
If people were willing to put every webpage behind a "do you want to run this webpage" we'd see them all using noscript already - virtually no one wants that experience.
> largely under the theory that we can do it safely with correct design
I blame operating systems and hardware vendors tbh. It's their job to make this safe, and they do a pretty bad job of it. Browser vendors have had to pick up a massive amount of slack to try to compensate, to the extent that browser teams have to make major patches to the Linux kernel.