The hyperbole over this kind of reasoning threatens the very fabric of the Web. Snowden did the world a service in revealing all of the NSA hacking going on, but the paranoia that is resulting from this is breaking the original spirit of the Web.
It is, after all, a Web of links, and those links were intended to be not just between siloed content, but between different sites owned by different people. Links by their nature, permit tracking. All you need for for sites to pool their web logs and collude, you don't even necessarily need fancy JS tracking.
When Web 2.0 was ushered in, there was an early euphoria in the community, of everyone offering transparent data and APIs to their sites, and people being able to easily compose content and services between multiple actors to make new sites and services.
It is one of the things that makes the Web better than native -- the ability to compose parts of the Web. No need for stuff like OpenDoc, or other notions of document composition, all you need are URLs and semantic elements that import or interface with external resources.
In the pursuit of paranoia levels of "privacy", what will we lose? Will we balkanize everything into content silos?
I'm not against trying to make things more "private by design", e.g., proxying to scrub requestors before the CDN sees the hit, or replicating resources locally. But if we take this to the extreme, we end up making local copies of everything, and the Web loses some of the semantic information from it's graph that I think is valuable to retain.