False dichotomy. A user-in-control sandboxing model is perfectly possible, e.g. https://sandstorm.io/

I agree that mainstream desktop OSes have failed to give us this yet. Fuchsia sounds like an opportunity to do better.

Agreeing with you, mini rant time:

Why in the fuck do desktop computers have user-based permission models? Don't answer that, I know why, it's because they inherited them from the server-OS ancestors, but it's retarded. Desktops are personal computers, there is no reason to stop the user from doing what they want because it is their computer. Some might say, "oh, well, we can protect the system from malicious software by limiting user access to it" and they're correct, but who gives a shit? If my OS gets destroyed I'll reinstall it. It's a bit of a pain, but whatever. What's really important to me are my documents, my work, and the user-permission model does jack and shit for protecting those as the ransomware wave proved.

Application-oriented permissions are the obvious and correct way to do things in personal computing.

There are desktop computers which are in shared environments (computer labs, public libraries, kiosks, teller lines, etc). So user-based permission models make a lot of sense for desktops.

Not really. There are better solutions than user permissions for those situations, we used them all the time in the 90s when multi-user didn't even exist in the Windows or Mac Desktop world.

Network resources, hosted on servers, need user permissions. Not desktops.

In those environments the same local user account is usually used. Or, a network account and in that case files are usually stored on a server.

Thanks for sharing that link with me.

Never heard of Sandstorm before. That's a really impressive product. I'll definitely be using it in the future.

I'm not super a fan of typing on glass, especially when next-keystroke prediction is as poor as iOS has been lately managing. I am super a fan of having a device with the size of display that eliminating a physical input device achieves.

My old Palm TX did the same thing, showing the graffiti area when a text input had focus and otherwise giving applications the use of all but a status bar's worth of glorious 320x480 screen space. Granted that Palm OS did less to help applications really leverage the extra space than iOS does. But most of what I did on that device was reading anyway.

Of course, Graffiti via stylus suffers less from a soft UI than keyboard input via touch, so it's not quite apples to apples. Still, though, as annoying as a soft keyboard so frequently can be, I wouldn't give up half my phone's display for a hard one.

  Did people really think the arms race of heuristic
  malware detection in bolt-on security software would
  be be sustainable?

No, people thought user education and decreasing software bug counts would be sustainable.