In the last hint (#4) of Chapter 5, the pseudocode applies a normalization to the origin ray:

• r ← ray(ray_origin, normalize(position - ray_origin))

normalizing in this case unnecessary, since with or without, the intersections detected are going to be the same.

My supposition is that this it was an accidental addition, or that it’s a common practice, so it’s been added even if it’s not necessary.

I think it should be either removed, or, if the case is the second, some explanation should be given, in particular, because at this stage of the book, it’s not entirely clear when normalization should be performed (there is only a generic explanation at page 30), so I find it (at least, personally) confusing.