As I see it, the shortest path would be a high-level programming language with an ORM (object relational mapper) that integrates template builders (something like mokk.me or Dreamweaver). The ORM abstracts out database specifics and allows the programmer to work with a more familiar construct, objects, while providing necessary methods like save, edit and delete. A template builder would visually handle HTML/CSS so that it feels more like design than development. Allowing components to be optimized later (replace or tweak the ORM with better SQL calls) would make the language/framework even better.
An alternative approach is an open source Flash-type language, that benefits from a standard plugin container. Sadly, in over 20 years of the web, the most integrated solution we have is Adobe Flash–a much maligned (insecure, bloated, slow, proprietary, …) language. For all its faults, Flash allowed people to create dynamic web pages that rendered uniformly–as designed–across browsers. Opening the spec may help as open specs are generally more secure and more optimized than closed versions. If the spec was open, people could write converters from various languages allowing everyone to produce the same output.
While we research an integrated web language there’s also room for people to research a new application container. The web browser replacement could use the same stack (TCP/IP, HTTP and HTML) or define new protocols.
Did I miss an option? What do you think will win?