Thing is, those performance issues matter. In user interfaces, they mean a lot. And they're also important when you're trying to do actual work. I'm yet to see a web app that wouldn't choke on the amount of data a power user may need. Yes, that includes Google's office suite, which is so horribly laggy when you put more than a screenfull of content into it that the very experience justifies switching to Windows and buying Microsoft Office.

What we would need, if the browser is to become a platform for actual productivity tools and not shiny toys, is a decent persistent storage interface - one that would be controlled by users, not by applications, and that could be browsed, monitored. And most importantly, one that would be reliable. And then, on top of that, a stronger split between what's on-line and what's off-line. Because some data and some tasks should really not be done through a network.

The problem with the web is not one specific missing feature, or even performance. Man-centuries of effort by browser makers have been able to make performance not-quite-competitive instead of just hopelessly uncompetitive.

The problem with the web is that the developer experience is nightmarish. The fact that native apps don't suffer XSS should be a hint about where to start looking, but it's really just a house of horrors in there.