But in the article, in the comments, he says that you should only use this process for one offs on large amounts of data