This is an interesting idea. Although I am afraid that publishing a dataset, even a good one, will not be considered "real science" by our (broken) institutions.