Back in 2012, in response to the Cost Of Knowledge declaration, Elsevier made all articles in “primary math journals” free to read, distribute and adapt after a four-year rolling window. Today, as David Roberts points out, it seems they have silently withdrawn some of those rights.