outline.md

• the problem: hoisting

  • closures in loops: sadface

  • is this worth solving?

    • the closures in loops thing is pretty bad; ideally we want things to work the first time

    • but still, maybe not

• the problem: global names in modules (or is it a problem?)

• the solution: let is the new var

  • how does let differ from var?

    • let is block-scoped

    • global let is not a property on the global object

    • for(let) is a new binding each time through

    • even for(let;;)

    • the temporal dead zone

      • try to present this as a convenient "oh you have a bug" kind of error
      • so let-declared variables are still hoisted, but only to the start of the enclosing block. and it's an error to try and access them before the declaration is reached.
      • this is a weird compromise design and I wouldn't have gone this route myself.
      • (some languages scope each variable to the code following the declaration, to make this a compile-time error, but because functions are hoisted in JS, the design of let was constrained.)

    • the global lexical tier

    • at toplevel in modules, var acts like let.

  • let is a reserved word in strict mode code. for backward compatibility, you can still declare a variable named let in non-strict code, but please don't do this

  • bindings introduced by catch follows similar rules--- catch has been block-scoped all along.

  • class bindings are similar.

• const

  • just like let, but you can't assign to it (that's a SyntaxError)
  • only initialized at the point in your code where it's declared
  • TDZ error to refer to it before that point

• implemented?

  • in firefox by Brendan; redesigned drastically in committee; updated by Shu-yu Guo