Yes, that’s right. And just to be clear: > (for/list ([s (in-list (instance-variable-names l))]) (define str (symbol->string s)) (if (symbol-unreadable? s) (eq? s (string->unreadable-symbol str)) (eq? s (string->symbol str)))) '(#t #t #t)

Well, there can also be unreadable symbols due to macro expansion. But there shouldn’t be anything but interned and unreadable symbols as long as a macro doesn’t use gensym.

@mflatt Are all unreadable symbols “interned” in that sense, or only some of them? And if only some, is there a predicate to tell?

All of them.

@cadr I think this suggests that our problem is just serialization of unreadable symbols (which we are currently writeing like regular symbols)