This strikes me as odd (from section 5.4):

> "Acquisti and Varian [2001] examine a simple model with two types of consumers: high-value and low-value, in which a monopolist can commit to a price plan. They find that although a monopolistic seller is able to make offers conditional on previous purchase history, it is never profitable for it to do so, which is consistent with the earlier analysis of intertemporal price discrimination by Stokey [1979] and Salant [1989]."

Isn't this what OS companies do? They charge less for upgrades? According to this paper, this would "never be profitable".

Am I misunderstanding this?

EDIT: added section number for reference and context