Can a `&T` be read as a `T`?

Decided

D-0025, settled by the language’s author: part of spec §18. Older text cites it as 18.25.

Not yet — and the operator traits stop needing it. &T reaches through . for members and is otherwise not a value: fn f(b: &f32) -> f32 { return b } is an error, and ^ derefs a ^T only (§5.2). That made §11.2’s fn mul(&self, rhs: &Rhs) unimplementable for Mul<f32>, so Rhs is taken by value instead, which is the better signature for the copyable math types these traits exist for. Eq/Ord keep &Self, because comparing a move-only type must not consume it. Whether &T should gain a read of its own is left open — nothing needs it now. Settled by 18.35: it reads as a T where one is wanted