How are downloads counted for datasets?
While counting unique IP addresses within 5-minute windows is a good way to avoid duplicates, it’s just a pretty rough estimate. It’d be great to have more detailed stats about repositories on the hub in general, like bandwidth usage and the most popular files.