For the data for the flexible formats, we could do an initial population from
a) the data already in the patterns
b) the open-office data (where they don't conflict with a).
Should we do this, and if so, when?
See bug 146.
Add way to get around combinatorics with formats