That's not what Kerckhoff said. The principle is that it should still be secure even if all details (other than the key) are public. He never said that the details should be public.
Denying an adversary access to the means of encryption is a valuable tool. It raises the bar considerably. For example, Bletchley had Enigmas fairly early on, but they never had a Lorenz machine. Until the Germans made an operational mistake Bletchley hadn't enough information about the workings of the machine to be able to attack it. The mistake allowed Bletchley to infer the crypto scheme used by the machine from just one intercept (it contained the same message twice). Once they had that they realised the scheme was pretty good, but slightly flawed. And then Tommy Flowers built Colussus.
By extension, a very good trick to pull off is to arrange for the adversary to be unaware of the communication in the first place. If he's not looking, you've already won. It's worse than security through obscurity (how do you know they're not even looking!?!?). But if achieved, it's a real result. Steganography anyone?
Anyway, researchers can now look at Apple's machine. If they've been paying attention to Kerckhoff they'll be OK.