Got mild food poisoning from the honours BBQ again (every year...). You'd think I'd learn :)
My cobweb chapter has been annoying me for a long time (about 5 years) because I know it's wrong, but I couldn't explain exactly why the idea (which so many people have tried, and all have failed) doesn't work. I had been distracted by the concept of computability and more importantly, the difficulty in understanding code. As tempting as this line of argument was, it was not satisfying, and I couldn't convince Kim of its truth.
I realised in our meeting that the problem is not the fact that one-way constraints can compute anything (for sure that is useful), but that it doesn't perform optimisation. All our later (more successful) work has modelled page layout as a minimisation problem with hard constraints. It is simply not practical to design an adaptive layout that is computed without minimisation (perhaps I should add this to my conclusion), because minimisation is the only way we know how to handle tradeoffs.
So I'm feeling a lot better about the flow of my thesis. On the topic of thesises, Jiri got his accepted with corrections and enhancements.