What's the "sharp edge of the knife" here? For use of web sites, the assumption has been that the user agent is responsible for preventing websites, no matter h

What's the "sharp edge of the knife" here? For use of web sites, the assumption has been that the user agent is responsible for preventing websites, no matter how malicious, from performing harmful operations. Even before there was a formal concept of "sandboxing" this was the basic assumption, and the biggest violation of it always came from "hey, download this executable and run it on your machine". Maybe someday we'll have sufficient app isolation that there will be no worries about the idea of running completely untrusted native-code apps on your device. But this is not that day.