Re: Mathematical Typography, Layout and Rendering

Paul,
 
It could be expensive research; it involves some R&D for the visual design components, possibly making use of a web browser or extensible IDE, and then also some R&D with regard to the machine learning components which output a heuristic or mapping between two domains of trees, outputting an abstract syntax tree describing logical processing upon the input trees and other typographical parameters.  The situation could be described as inducing tree transformations, inducing tree-based automata or transducers, or, generically, inductive logical programming. Additionally, for a constant mapping, the topic could be phrased as inducing parameters per box depending upon the layout tree's box model.
The attractiveness of it, I think, comes from considering the processes that occur in the design, development and stewarding of layout engines where feedback from web developers and end users often includes instances of MathML and rendering-related pixel motions, possibly accompanied by references to other layout systems, e.g. AMS-LaTeX, or, abstractly, discussion of aesthetics.  The use of visual design premises to easily version the layout and rendering of mathematics, and for a large number of fonts, would seemingly be appealing to implementers. 
 
Adam