Source: https://www.google.com/search?q=os+x+root+exploit&gws_rd=ssl
Source: https://www.google.com/search?q=os+x+root+exploit&gws_rd=ssl
This protects against a whole host of issues. It safeguards against garden-variety incompetence[1]. It provides some defense against the large number of badly-intentioned people who can write an Objective-C app, but don't have the expertise necessary to weaponize a typical root escalation exploit. It prevents apps from accessing your contacts, reading your emails, determining your location, and accessing the webcam and mic without your knowledge, amongst other things.
Does it protect against a motivated, highly technical attacker? No, not really. But that hardly makes it useless.
[1]: http://www.macobserver.com/news/98/december/981229/bungierec...
The exploits tend to be trivial, often trivial enough to fit into a single tweet. (https://twitter.com/i0n1c/status/623727538234368000) They require no competence to use.
As for protecting against incompetence and mistakes, that is far too an extreme of a measure solely to protect against that. Some decent QA will fix that.
So what is the point, really, of sandboxing if it does not thwart highly technical attackers? It severely limits the functioning of apps, makes it far more difficult for app developers (myself included), and for what benefit that could be worth the trade off?
https://www.google.com/search?q=developers+leaving+%22mac+ap...
It thwarts the attackers who aren't highly technical, and frustrating the script kiddies could have flow on effects when beginner attackers don't get the reinforcement to motivate themselves to refine and build their skills.
Secondly, exploits can be patched over time. Ten years from now, is OS X going to be better off for having the sandbox? Do you expect a lot of trivial exploits to be discovered after another century of person-hours are invested in the sandbox?
And yet we do those things anyway. The idea is defense in depth, such that if one mechanism fails then hopefully another will mitigate the damage. Sandboxing isn't perfect, but it's another layer of security and I'd rather have it than not.
It's in the best interest of the hacker that broke into your system that your system continues to work flawlessly for both you and the hacker. This is why Mac OS X "rootless" is just yet another obstacle for the power user, yet another obstacle when compiling and installing POSIX code from source, and yet another step closer to locking down OS X to be an appliance like iOS.
And if you really want to disable rootless anyway, you can do so. Boot into the recovery partition and there's an option there to turn off rootless.
I'm also completely baffled by the claim that, just because no security solution is 100% perfect, that we shouldn't even try. That makes no sense at all. Yes, security is hard. But protecting you from 99% of all malware, even if there's the rare case of malware that gets past you, is still extremely useful. Besides, it's awfully cynical to declare that SIP is an impossible goal before you've even looked at it.
But, you can disable SIP so not sure how much it really matters.
Good catch on finding something that breaks with SIP, but even if you philosophically disagree with the idea of rootless, you should still agree with the notion that library interposing is a serious security threat and should welcome the changes to block interposing of system processes[1].
[1] From the What's New In El Capitan docs[2], the specific aspect of SIP that applies here is "Code injection and runtime attachments to system binaries are no longer permitted".
[2] https://developer.apple.com/library/prerelease/mac/releaseno...
So no, I don't welcome "SIP". There are better ways to solve that problem.
I recommend that you tally up the revolutionary technologies produced over the last 30 years that allowed us to get to this point (including the web), and then consider how many of them could be invented on Mac OS or iOS today.
Yeah, cracking is asymmetric warfare that we have no hope of winning, I think anyone with any knowledge of computers realizes that is true. It doesn't mean we should smugly shoot down anything that makes it incrementally harder.