I think it's useful to think about this question from the angle that "web browsers" are really a bunch of things in one bundle. A universal application platfor

I think it's useful to think about this question from the angle that "web browsers" are really a bunch of things in one bundle. A universal application platform with "download and go" deployment semantics is obviously a winner. A distributed information publishing/viewing system is nice too (HTML/CSS/HTTP). Merging that and the application platform kinda compromises both, but also avoids the need for a distinction. If you were doing it from scratch though, building the information viewer on top of the application platform would be sane (rather than the other way around like we have now). Sockets are good (We kinda already knew that). WebSockets are not as good a real sockets, and it'd be nice to have the real thing sometimes ... A universal text-based marshaling mechanism is useful. JSON is maybe not the best it could be, but perfect is the enemy of good enough, I guess. A universal RPC machines is too. Running XHR over HTTP is kinda broken, but ... Half of what's done in browsers duplicates (often poorly) what's done in the underlying OS. All that said, the vast majority of what you would change is basically pointless to think about because the deployed base is so large. Legacy code is always what gets you.