raph, I too was underwhelmed by the results of the paper. I don't share the authors' confidence in their approach but I see how seductive it might be to a non-type-designer... the same intellectual trap that mars METAFONT results, as you say: the conceit that letters can be dissected into mathematical objects at a high level. The research seems to show that this is not the case. I did think the paper had some interesting aspects but these were tangential (haha) to its main premise.
Re: UI frameworks, yes, it's a tragedy that we have to wrestle with several mutually incompatible UI models. Sigh. Guess each player doesn't really want cross platform software (excepting the open source toolkits of course).