Or another way to measure would be to look at the first X Mb worth of packages which are most commonly installed. Debian provides this info: http://popcon.debian.org/
All values of X would be arbitrary. Maybe continue down the list for as long as the packages can fit on one CD?
Another measurement which might be even better (but is surely overly complicated) would be to use the popularity of each package as a weighting, and measure the whole list.
...but until there's a way to factor in GNU's contribution via catalysts, legal infrastructure, and organisation, all these measurements will undervalue GNU and will be jumped on and mislabelled as "empiric" by anti-GNU people (they're empiric in that they measure something repeatably, but they measure the wrong thing).