Well. The circle I am in is: Think random thoughts until I've got a reasonable grasp on all invariants and the compilation scheme I want to support, work up some appetite to translate it to code, code a bit.
It's mind-numbing. Bosch was right, I feel like the bloke above.
I am not sure yet to support a naming scheme which interfaces with C object files right. I need something where if a combinator 'foo' is declared in namespace 'baz.baf' in some file 'qux.hi' it unambiguously 'knows' the symbol name of the C function representing 'foo' such that it will link...
You'ld expect a compiler is all about parsing and typing but, honestly, that's only because these are the trivial aspects academics can concentrate on easily to publish their greeklish.
The mundane world of naming stuff. Tss.