Apple ordered to bypass auto-erase on San Bernadino shooter's iPhone

Is that true of non-Secure Enclave phones like the 5C?

But then you go on to speak of edge cases that precedent can't deal with, but I would argue that your view of case law as essentially "algorithmic" is flawed. Obviously precedent cannot be exact, but using prior interpretations helps guide thinking.

For example, look at Katz v. US, one of the seminal cases that would inform this case. In Katz, the court starts with existing law, one that prevents illegal searches, and tries to decide whether that law can apply to electronics (such as tapping a phone). At this point the OCCSSA is in effect, but not really tested, so we've got a pretty fuzzy legal area despite the fact that phones are a well established technology. The court rules that even though law enforcement did not search or seize things, that privacy is still implied in electronic communications because of other acts a user takes surrounding the act of making a phone call (closing the door, making the phone call from home, etc.).

In fact, there's a whole legal concept at work here called lawful intrusion that is built upon every time a case is judged. These concepts, and human application of them to the case at hand, help attorneys, judges and juries deal with edge cases.

I think there's yet another point to be made here. In an imperfect world, where we can't design axiomatic systems that are corner-case free and where we can't get stuff right on the first try, the judge's flexibility seems like a necessary safeguard against the letter of the law totally missing its spirit. I like to imagine it as a grease, a lubricant - you need to apply it, or else minor imperfections will grind your machine down.

A fundamental question here is - do you want the law to be a completely trustless system? Sticking to the letter of the law and enumerating all imaginable corner cases are examples of going into that direction.

I'm not totally convinced trustless systems are a good idea for society, because of inefficiencies they introduce as a cost of not having to trust anyone.

If Apple is capable of compromising security on its devices (by using its root key to sign a custom version of iOS, or through some other method), then I see no way that they will avoid eventually being subject to a court order in some jurisdiction that compels this action. If that's true, then device security is already compromised and Apple knows this. Let's say the facts of the case were slightly different, that the FBI "knows" a terrorist attack is about to occur, and Jack Bauer-style demands that Apple assist in compromising a specific device that has the top seekrit plans on it. In that instance, do you think Apple would comply with a warrantless request for cooperation? Hm...

Reading Tim Cook's announcement in light of this thought experiment, methinks he doth protest too much! Apple does not have any objection to compromising user security at the root level, and in fact has already done so by creating a device that has some limited vulnerability to malicious action by the manufacturer signed with its root key. (By the way, no doubt every other manufacturer has done worse, so this is not to deprecate Apple vs. any other big company.)

I would speculate that Tim Cook's goals with this announcement are largely PR-based, and that the goal of Apple's legal strategy is not to avoid cooperation but rather to retain the ability to decide whether to cooperate, and/or to impose a higher perceived cost on the government for such requests. No doubt Apple is correct to say that once a precedent is established, then it will be widely used by law enforcement even in routine cases.

At the end of the day, I am not optimistic that we can avoid a world in which large device manufacturers are compelled (legally and practically) to build security flaws into their devices. Perhaps not the flaw of a back-doored crypto implementation, but other flaws such as those that have been identified in current iOS devices that allow the government (with commitment of sufficient resources) to chip away at some of the more superficial protections.

I understand this, and if I thought it were possible, I would agree.

However...

The law is essentially trying to codify a moral code, one that changes with time. We've seen how the law struggles with changes to technology, lifestyle and shifting attitudes. But sometimes the very axioms change (see slavery, suffrage, common law). Because the law cannot keep up with changes, we're stuck with messy human interpretation to smooth over some of those rough edges. Something like Brown v. Board of Education, which seems obvious in retrospect, may not have occurred in an axiomatic system (perhaps Plessy v. Ferguson may not have occurred either, though considering the primary axiom that American law is based on once said that blacks were 3/5 of a human, I find that unlikely).

I would argue that we ought to try to introduce algorithms into the enforcement of the law (policing, traffic enforcement, jury selection, public defenders, etc) rather than the interpretation of it. Of course, one could argue mass surveillance is exactly that, so I don't know.