I have moved recently to DC in order, in part, to start work at a computer science research lab at University of Maryland. This lab is funded in large part by DARPA, but also by NSF, NSA, and other relatively creepy gov't agencies.
Some of the work I will be doing (or, more accurately, some of the work that people in this lab are doing which I am familiar with) amounts to politically safe versions of TIA. So we're working with Howard University's computer science department to build a TIA-like system using all of the Freedman's Bureau material. Likewise, we're doing a project, funded by the NSA, using Holocaust Museum data.
In both cases, you have data sets which are 1) politically unobjectionable, but 2) similar enough to TIA-like datasets in that both are "whole population" sets (though much more constrained than TIA would be). The idea is to develop some of the algorithms, techniques, and tools "in the clear, in public" using politically unobjectionable datasets. Then take the lessons learned into organizations like NSA, behind the veil of secrecy, to see whether or how those lessons apply to something like TIA.
DARPA and NSA funders talk freely about the connections between these projects and TIA. I cannot speak to the larger funding issues, but these projects at UMD are *funded* and moving forward and do not rely (yet) on 2004 FY appropriations (well, near as I can tell...).