Just got around to digesting Tom Coates' essay on "Amazon, excess, and the future of navigation...". Sparked by Amazon's recent deployment of, per user, freeform labeling on items, Coates riffs on how to deal with navigation when there are copious objects to deal with. Overall a good piece that captures a rapidly developing trend. I'll just add a couple of cents worth of my own.
Navigation in an environment of excess is a starting point. People want to get things done as they're noodling about on Amazon. While tagging is a good way to socially find new stuff, it's not much help in executing a particular task other than generic browsing. There's a lot of room in web systems for deploying and improving task specific interface mechanisms. As an example, for Amazon making the shopping cart smarter would be a win, but they also have authoring tools in there for making lists, writing reviews, writing guides, and managing registries. How could these new navigation mechanisms make those authoring tools better?
One other thing about dealing with excess. I was just scanning Tufte's "Envisioning Information" again and as usual was jazzed by the sections on Micro/Macro Readings and Small Multiples. Those general ways of thinking seem particularly appropriate for navigating excessive corpuses. Unfortunately, it's really hard to apply these techniques in web based systems. The graphical control needed to pull such visualizations off is well beyond current browsers' capabilities. Also, interaction can provide a major assist here, but even pushing the limits of AJAX and DHTML the best infoviz techniques can't be applied. Maybe this is the appropriate role for Flash and Java applets.
In short, all of these systems that are providing access to huge piles of stuff will force the further development of web-centric interface techniques.