The test data is big.  Really big.  Too big to store it in a
source code repository.  This poses the problem that the test
has to either download it or to rely on the data being in a
defined location that is always on and always accessible from
everywhere.  Both solutions don't work in all cases.

Bottom line:  we have to figure this out!  For the time being the
test will just fail when the data is not in the expected location.