Their datadump strategy is questionable. You end up needing perhaps 1 book for each shard. I guess they are made by sorting the ISBN or some other numerical field and then grouped until they reach a certain size. It is a bit annoying to put the shard of interest to download in your torrent client, pause it, figure out which book is the one you are after by looking at the ids, instruct your torrent client to only download that book.