The data sets were in XML, so that was easily put on to data.gov. In part, some of the data sets that we chose were the low-hanging fruit, what is easily put on to data.gov. We used XML, and that works well.
On March 2nd, 2011. See this statement in context.