I enjoy using mongo / other document stores, but this article is lame. The contrast is between a single document storing an order, with sub-documents containing details, versus a table per sub-doc type. They claim you have to denormalize data in the relational version, which isn't true, and use that claim to assert that updates are complicated.

Guess what? Denormalizing data can make things complicated in mongo-land, too. This isn't just a relational db problem.