Engineering to fail is a art
Given the history of Govermental projects in IT and the fact nobody can name a succeful project unless they actualy worked on it. One can't help but fail that this is a somewhat shortsighted mentality and the approach is along the lines of underpant gnomes - release data, ????, Magic applications.
Now given that we all read about alot of failed projects and TAX and benifit system being one and remarkably enough systems that other countries have as well though all biased to there own rule sets in process. Why don't they release a good set of `TEST DATA` and what is needed and seed of some open source projects to do this processing. You could even wrap this up into the education system. This gives real world projects for people to intereact and learn from as well as eventualy give us some solid applications that are adaptable world-wide. Be nice to right the tax laws in a form of XML configuration file that enables it to handle any TAX laws that might come out.
Only sensitive issues is the data and thats why all largescale projects that have such data have a pile of test data (usualy morphed from real-data with the usual hand-crafted gotch'a and extreeme's, but still not real data). The govermental paid IT bod's can then remove the bod down aspects and help/work on true project managment.
Having data is all great, but people sometimes need direction , whilst others do not. All very well dumping a load of data, but some form of wish list would not go amiss and if some of those wish lists are managed in that you have online contact and feedback, then it can only go alot better.
Remember, you can lead a coder to data, but you have no idea you even wanted a swing across it.