This, like many articles, seems written by a person who hasn't read much of the existing thought on the topic, and is therefore blindly bumping into old issues that have already been covered in much more detail elsewhere. Read Celko, read Date, read about all the reasons for the surrogate key optimization, then come back and write something if there's more to say than has been said already.

FWIW: surrogate keys are both evil, and necessary for scalability. Unless/until the database vendors come up with an internal engine optimization that avoids this problem (the issue of the width of keys / the cost to compare them / the cost of On Update Cascade), an issue which seems still to be a blind spot for them, we will have surrogate keys with all the extra joins, logic problems and headaches they create.