This is a backward-looking argument which ignores the unique benefits of the web which made it inevitable that it would evolve into an application platform, regardless of how tortured the results may feel.
The web is the first truly cross-platform development environment. It is not controlled by a single vendor, and anyone implementing a new computing device must support the web (just stop for a second and consider from a historical perspective what a monumental accomplishment that is). Furthermore, it allows casual access of content and applications without any installation requirement. It comes with a reasonable security model baked-in, which, while imperfect, gets far more attention than most OS vendor sandboxing schemes. Last but not least, the web's primitive is a simple page, which is far more useful than an app as a primitive—for every app someone installs they probably visit 100 web pages for information that they would never consider installing an app for.
I agree that the web is sort of abused as an application platform, the problem is there is no central planning method which will achieve its benefits in more app-oriented fashion. No company has the power to create a standard for binary app deliverables that will have anywhere near the reach of the web. And even if one could consolidate the power and mastermind such a thing, I feel like it would run squarely into Gall's Law and have twice as many warts as the web.