This doesn't have to be perfect.
We agree it doesn't need to be 100% perfect. But it needs to be at least, like, 60% perfect, right? And unless you make it at least a bit hard to bypass, it will stop virtually no one.
Installing a new browser is already a bit hard for most people. I think you are a little skewed in your thinking being online on HN.
You also aren't thinking about age. Certainly 16 and 18 year old probably can get a new browser installed. But a 14 year old? 12 year old? 10 year old? That barrier is a lot higher the younger a kid is.
To give you an example of the workarounds kids will find: Youtube was blocked on school laptops, so the kids all started embedding Youtube videos inside of Google Sheets in order to watch stuff. This isn't, like, something a few savvy kids did, it was a widespread and common practice.
To effectively keep adult content away from kids, it merely requires secure boot and closed app stores, which are already widespread. And they are only required on the devices actually given to kids, rather than every single computing device.
But this proposal has another problem: it's easy for a website to run isUserOver(n) in a loop to derive the exact age. And on a persistent account, it can be queried every day to derive an exact birthday! Which comes back to my main point that the only technical schemes we should be considering are ones where information strictly flows one way - the website/app supplies information to the browser/OS, which then [may] implement parental control policy. anything else fundamentally boils down to a mandate for identity verification.
This is unacceptable. If I own a computer, I expect to be able to build and run any program, either written by myself or others, without asking anyone for permission.
> And they are only required on the devices actually given to kids
My whole point is that this limits the blast radius, compared to any solution involving "age" (read: identity) verification which has a blast radius of every computing device!
Perhaps my other comment will show you where I'm coming from better: https://news.ycombinator.com/item?id=48645646
...I guess I don't really see the difference.
Closed app stores are widespread on some platforms but certainly not others, and I for one would really like them to not spread any further.
But in general there is a huge difference between the freedom-destroying properties of secure boot with closed app stores, and the next step of remote attestation. Remote attestation lets the server insist that you only run software fully of their choosing rather than your choosing, as a condition of interacting with them. This completely destroys the idea of protocols that mediate between two parties with diverging interests, and computationally disenfranchises users. Imagine the next generation of the Cloudflare nagwall that doesn't let you past unless you buy a new computer, and that new computer must be running MSWin/OSX and MSIE/Chrome.
(also note that my use of "secure boot" here includes systems like on Pixels where you can straightforwardly unlock the bootloader (erasing the data on the device), install whatever you want, and then relock. I still find these systems philosophically objectionable, as there is still a privileged key baked in and retained by the manufacturer - similar security properties could be provided without the backdoor. But pragmatically they've been working okay)