"As a second example, take the object plane and the image plane to once again both be at x = 1 and define T(1, u, v) ≡ (1, −αu, αv). As with the flat mirror of the previous example, the transformation scales the image plane, but be- cause of the relative minus sign in the y and z coordinates, the solution surface will not reverse an object as a conven- tional mirror does.

The relative minus sign also means, alas, that no exact solution surface exists. I obtained an approximate solution surface confined to the rectangular volume x = 34 ± 1 cm, y = z = 0 ± 3 cm, and constructed a prototype mirror. With α = 160, the mirror’s field of view is wide enough that I could see myself in the mirror when it was held at arm’s length, as shown in figure 1. To achieve the appropriate ray paths, the mirror is saddle-shaped."

So, this mirror only does its magic in a small volume; if you move your eye closer to the mirror or sideways, or even when you look at it with two eyes, the effect breaks.

From cursory reading, I even get the impression that this mirror is also tailored to the positions and shapes of the objects being photographed.

But as I said: cursory reading. Corrections welcome.