J Bostock comments on The Geometry of Linear Regression versus PCA