I think that different visualizations of the internet aren’t so successful, because internet does not have a notion for ‘location’, and a notion of ‘time’ is somewhat different from the real world one, too.
So, firstly, human instincts about the land mapping are not applicable to cyberspace (it’s difficult even to define the concept of ‘distance’), and automatic visualizations aren’t based on any real (perceptible) facts, so they don’t bring in any useful information and are considered of little use.
If we could do search using successive approximations approach, then an average web user could hold a significant part of the Net in his brain in the form of associations “here i need to turn this way”. But now we’re tied to search engines 😦
In this sense the ancient Gopher and current FTP protocol are better than HTTP, because they provide at least tree-like object arrangement, but unfortunately internet is far from being tree-like…
Hyperlink graphs seem to describe the Net structure pretty precisely, but they aren’t reversible and opaque. Also, they are vulnerable to spammers, outdated and poorly reflect the ‘non-page-like-objects’ (user profiles, emails, instant messages, etc)
But there should be some kind of a solution! 🙂