3 ms·

My Wayland performance experience was that it broke Nvidia at a time when Nvidia was supposed to be working. The reason it didn't work was down in the weeds of

My Wayland performance experience was that it broke Nvidia at a time when Nvidia was supposed to be working. The reason it didn't work was down in the weeds of this version of Mesa and that version of Nvidia, both of which were recent, and the bottom line was that X performance still exceeded Wayland's. Security has always been a hard sell to end users because the costs are immediate and tangible while the benefits are in the future.

DarkUranium 3y ago
I think one thing that people forget is that security is always an overhead[]. Mind, I'm not saying it's not sensible* overhead --- that would depend on the use-case. It's an overhead in the same sense on how designing a bridge to be able to carry loads N% over its designated capacity is an overhead: it's an additional cost. This is also part of why so many companies are terrible with security, and consider it only an anti-liability "minimal required measures" thing (a big part, of course, is companies simply not understanding security). In the case of a bridge, in terms of material, time, and money. In software, it depends on the program, but it's usually some combination of performance (e.g. encryption), cognitive overhead (remembering passwords, access policies, ...), accessibility, development time, and a few other things. And notably (for Wayland), features --- locking down a system naturally reduces what a system can do. Note that this overhead can be a blocker. Accessibility is an obvious one (e.g. take those numeric keypads that randomize the order of digits presented on the display each time you want to unlock them --- blind people will not be able to use those), but features can also be a problem: if part of my work requires commonly taking screenshots (e.g. because I work as a technical writer, and need screenshots for documentation reasons), then I won't be able to do my job. In fact, the corporate world is chock-full of (maybe even sensible!) security policies reducing efficiency of workers, or sometimes outright blocking work. What I'm trying to get at is, security requirements depend heavily on each user and use-case. And I think Wayland's fallcy was in trying to shoe-horn everyone into this vision of a fully-sandboxed world without considering the practicalities and what people actually want or need. For example, I keep my personal desktop (which I mostly use for gaming & FOSS work) far less secure than my work computer (which is also in my apartment, as I work from home). Because I've deemed the overhead of trying to better-secure my desktop, unlikely to get stolen from my apartment, as unnecessary overhead. Meanwhile, my laptop, which I also use for gaming & FOSS work, is more secure, as I tend to take it with me when I go on a trip. Because while the severity of a breach is the same as for my desktop, the probability is much higher. So I have three situations, and I'll argue that no single security policy would be a good fit for all 3: (or even two) - Home desktop => low risk of breach, moderate impact if breached[*] => low security - Home work PC => low risk of breach, high impact if breached => high security - Laptop => moderate risk of breach, moderate impact if breached => medium-to-high security That, I think, is part of the issue people have with trying to sell Wayland's flaws as "security". It's basically a one-sided "I'm changing the security policy, and you better pray I don't change it further". When said security might be more than just undesirable, but an active obstacle. At the end of the day, Wayland is a tool. It has a job and a purpose, and yet it can't even properly do some things of the tool it's meant to replace. Because "security". ([] If you disagree that it's an overhead, then I'm going to expect that every single door in your home, even the fridge & closet doors, are packed with as many heavy-duty locks as will fit on the physical door. And that each door is always locked unless you're currently actively interacting with the door. For starters.) ([*] Yes, I realize the issues with an attacker being able to access all my accounts and such, but I'm putting it in relative terms compared to the work PC.)

DarkUranium 3y ago

I think one thing that people forget is that security is always an overhead[]. Mind, I'm not saying it's not sensible* overhead --- that would depend on the use-case. It's an overhead in the same sense on how designing a bridge to be able to carry loads N% over its designated capacity is an overhead: it's an additional cost. This is also part of why so many companies are terrible with security, and consider it only an anti-liability "minimal required measures" thing (a big part, of course, is companies simply not understanding security). In the case of a bridge, in terms of material, time, and money. In software, it depends on the program, but it's usually some combination of performance (e.g. encryption), cognitive overhead (remembering passwords, access policies, ...), accessibility, development time, and a few other things. And notably (for Wayland), features --- locking down a system naturally reduces what a system can do. Note that this overhead can be a blocker. Accessibility is an obvious one (e.g. take those numeric keypads that randomize the order of digits presented on the display each time you want to unlock them --- blind people will not be able to use those), but features can also be a problem: if part of my work requires commonly taking screenshots (e.g. because I work as a technical writer, and need screenshots for documentation reasons), then I won't be able to do my job. In fact, the corporate world is chock-full of (maybe even sensible!) security policies reducing efficiency of workers, or sometimes outright blocking work. What I'm trying to get at is, security requirements depend heavily on each user and use-case. And I think Wayland's fallcy was in trying to shoe-horn everyone into this vision of a fully-sandboxed world without considering the practicalities and what people actually want or need. For example, I keep my personal desktop (which I mostly use for gaming & FOSS work) far less secure than my work computer (which is also in my apartment, as I work from home). Because I've deemed the overhead of trying to better-secure my desktop, unlikely to get stolen from my apartment, as unnecessary overhead. Meanwhile, my laptop, which I also use for gaming & FOSS work, is more secure, as I tend to take it with me when I go on a trip. Because while the severity of a breach is the same as for my desktop, the probability is much higher. So I have three situations, and I'll argue that no single security policy would be a good fit for all 3: (or even two) - Home desktop => low risk of breach, moderate impact if breached[*] => low security - Home work PC => low risk of breach, high impact if breached => high security - Laptop => moderate risk of breach, moderate impact if breached => medium-to-high security That, I think, is part of the issue people have with trying to sell Wayland's flaws as "security". It's basically a one-sided "I'm changing the security policy, and you better pray I don't change it further". When said security might be more than just undesirable, but an active obstacle. At the end of the day, Wayland is a tool. It has a job and a purpose, and yet it can't even properly do some things of the tool it's meant to replace. Because "security". ([] If you disagree that it's an overhead, then I'm going to expect that every single door in your home, even the fridge & closet doors, are packed with as many heavy-duty locks as will fit on the physical door. And that each door is always locked unless you're currently actively interacting with the door. For starters.) ([*] Yes, I realize the issues with an attacker being able to access all my accounts and such, but I'm putting it in relative terms compared to the work PC.)