I think there should be three modes, one mode for documents and two modes for apps. Right now the document "mode" is too overpowered while the "app mode" is under-powered. If the user switches to "app mode" it should give the website access to the web API's, while in document mode it should only be possible to manipulate the DOM. The second app mode should make the app more "native" like and unlock more API's. Or we should just fork the browser into two programs, one for documents and one for app runtime.