@John: there’s a growing number of tools out there, both toolkits (i.e. programming languages such as Flash, PHP, Python, R, Processing etc) and ready-made solutions that don’t require programming but are consequently more limited in their capabilities. When it comes to the type of visualizations we can currently do I feel like we have a chicken-egg problem: we need to combine an understanding of humanities data with a solid grasp of visualization (and, if we want to do quantitative stuff, statistics) to build meaningful visualizations.
And yes, combining subjective tagging with structural characteristics is a challenge. Kinda ties in with Hugh Cayless proposal for markup:
thatcamp.org/2010/the-inadequacies-of-markup/

@Brian: I’d definitly like to hear more about geospatial and temporal vis.

I initially considered using Processing as a framework for my own visualization project, but eventually decided on Adobe Flash for prototyping, as it’s already focused on interactive motion graphics. Of course, I’m pretty neutral about platforms — and am in fact more interested in how the development process can be thought through to take more advantage of the strengths of different languages or programs without the current significant barriers of knowledge or technology.

How does one integrate both subjective and textual data together? I think this question, along with the others you mentioned, would be very interesting discussion topics.