I just finished spending an hour or so starting on a Sudoku solver in written in Twelf. I didn't quite finish, but I would have to say that the complexity introduced by the lack of polymorphism seems to outweigh any of the benefits of logic programming. Plus the lack of a standard library means I needed to write custom list and matrix manipulation code. I'll probably finish it at some point just for the hack value, but perhaps not before the new Twelf module system becomes available.

I was pretty terrible on Tuesday and spent a significant portion of the day thinking about unification, anti-unification, and type inference for InforML. I am almost certain that I want to just stop working on the higher-rank polymorphism extension to AspectML and start directly on InforML. At least my hope is that if I stop working on AspectML and start directly on InforML, I'll have a rough implementation of InforML ready by WGP, which will be far more expressive than higher-rank AspectML anyway. Especially because while thinking really deeply I concluded there are a few things that we did when designing the AspectML type-inference algorithm that unnecessarily complicate it. I guess I'll have to see if I can sell Stephanie on this plan.