Shu-yu Guo

@littledan I admit I'm somewhat surprised to hear the dependence on Annex B.3.3. If web content forces our hand, then it is what it is. I would like us then...

> "random accessible" means that the reader doesn't have to iterate over all statements to reach the function, but just by following the NodeLink with 1 step, right? Yep! >...

> the current change requires it to be an exhaustive list (no duplication, no unmatched items), and in that case I think it's fine. Ah, I think I misunderstood what...

> if not, that doesn't sound right. if actual inner function is missing from .innerFunctions list, the error is thrown in the following step: Ahh, oops, you are correct. The...

The current PR doesn’t handle surface encoding errors like linking to nonexistent functions. I should put a note. The checking at delazify time is about verifying that the functions are...

@littledan Sounds like a good idea to me!

@ljharb The conservative option is as you say: to treat the presence of directives as an annotation, like the proposed solution for storing scope info. That is, the extra bit...

JSON should be a very aspirational thought at this point of the proposal.

In practice, I expect that compilers are free to transform early error cases in text JS into a single throw, or just refuse to emit an AST. However, implementations must...

Indeed, they are very different now. The proposal defers all early errors to "first invocation of function". In the proposal world, partial graph evaluation is always possible.