I wonder how much space the entire contents of SciHub would occupy, if they extracted just the plaintext from all articles and packed it. It would be great being able to grep every scientific paper ever published.