Crosscut

Earlier prototypes had dynamic type systems, and this is rather easy to do. In the previous prototype, I had started to implement a static type system. And while I never got the type inference fully working before I switched to the new prototype, I think it would be practical to continue with explicit type annotations. So both static and dynamic are viable options.

The eventual goal is to have a static type system with full type inference. From an implementation perspective, a static type system with explicit annotations has the advantage of providing a smoother transition to that, compared to a dynamic type system. But it's not as easy to implement. Which could turn into a distraction.

In the short term, I am much more confident in my ability to implement a dynamic type system. And while I used to think that this would be a detour, I'm starting to change my mind. Because while a static type system provides a smoother transition to the eventual goal from an implementation perspective, a dynamic type system might present a smoother transition for the user.