Crosscut

As I'm talking about separately optimizing input code and rendered code for their respective purposes, I realize that what I'm describing is basically WYSIWYG. Except that there is no printed form that would unambiguously define "what you get". Instead, deciding what the "what you see" part should be, is a big part of the design challenge.

I'm not sure what to make of this realization. Ironically, I've always preferred plain text formats over WYSIWYG. WYSIWYG editors always seem to have hidden state that I don't understand, and can easily mess up. While plain text formats might be a bit more clunky, but easier to understand.

Maybe that's just a matter of implementation quality, and it's possible to do better. I don't know. But either way, how different the rendered code is from traditional programming language representations, and how different input code is from rendered code, those are spectrums. I could always decide to land on the conservative end of those, instead of making something novel and ambitious.