The author is talking about the difficulty of typesetting mathematics, and refers to TEX as an approach to solve this problem. eqn and troff preceded TEX. Heck, even Microsoft's and Open Office's approach to typing in equations are direct descendants of eqn.
I think the author is only interested in the box model that TeX uses insofar as it influenced the typeface design. He’s contrasting TeX’s “box and glue” setup with the rigid matrix design needed by Monotype faces, and the much more expressive non-box layout model used by Cambria.
Very true! Side note: I always admired the idea of eqn, loved reading about it from Kernighan and Cherry, but never saw it as powerful (or pretty) enough for serious use.
