The idea is brilliant, but it isn't new, as far as I know (or maybe it is, but I had it too, independently -- I don't know).
The big problem with this, as with 4- and 5-axis milling, is software. For milling, software exists, but it's expensive. For 5-axis FDM, software does not exist yet, but it's likely to be much more complex than a classic 3-axis slicer. And I don't just mean more code because there are more axes, I mean conceptually more complex, and involving much more interesting mathematics. I'd like to point out that his thesis does not solve this problem: he generated custom G-code for his specific objects. What he did was only one step removed from writing G-code by hand: because the G-code was very repetitive and involved a lot of computations, he wrote a Java program that generated the G-code. The Java program was specific for each particular shape.