17:00 <Putonlalla> Is `else` supposed to be rebindable? I thought not, but the behavior of `cond` says otherwise, as the expression `(let ((else #f)) (cond (#f 'if-branch) (else 'else-branch)))` evaluates to an (unspecified) empty value.

17:19 <samth> Putonlalla: yes, `else` is rebindable.

17:20 <samth> consider: `(let ([else #f]) (cond [else 1] [#t 2]))` which produces 2

17:22 <Putonlalla> So, syntax rule literals are somehow defined and also not defined.

17:24 <samth> Putonlalla: they are recognized by their binding. So `cond` recognizes the `else` that was bound (or not) where `cond` is defined, and other bindings of `else are not the same.

17:27 <Putonlalla> If I declare `(module m racket (provide c) (define-syntax c (syntax-rules (e) ((_ e) 'yes) ((_ _) 'no))))`, then `e` is somehow implicitly exported, although I cannot refer to it without getting an `unbound identifier` error.

17:38 <samth> Putonlalla: no, e is not exported, it's unbound, and so the `c` macro recognizes an _unbound_ `e`

17:39 <Putonlalla> I see.

17:43 <samth> the traditional scheme practice of using unbound identifiers as keywords has some drawbacks, and racket has moved away from that, but `cond` is very old

18:17 <Putonlalla> It's indeed a bit strange that `else` and `=>` are free variables, especially since this problem could have been avoided with `(define else #t)` and forgetting about `=>`.