Crosscut

As I said yesterday, Caterpillar is currently very confusing and hard to use. I've run into this problem with many previous prototypes. It seems you just need a critical mass of language features that promote clarity, until you cross some threshold where the language becomes productive.

But what if I have tooling that tells me exactly what's going on at runtime? Will that, by itself, make an otherwise confusing language practical to use? This is not a ridiculous thought, I think. People write non-trivial stuff in assembly languages. Presumably, at least in some cases, with the help of debuggers.

I'm wondering if developing a debugger first is a more practical path to productivity, than reaching this critical mass of features. If so, I could have a usable language from early on, and could take my time expanding it. Only adding features incrementally, as I deem it necessary and appropriate. But I need to figure out if this approach can work, and that's the focus of the current prototype.