Collective SSL FAIL a symptom of software's cultural malaise
Apple, WhatsApp and Belkin show that if you ask for bad software, you get bad software
In the 19 years that have passed since the first implementation of SSL, you could be forgiven for expecting that the industry could do it right by now: and yet last week, not one but three SSL vendors were discovered to have implementation problems.
Belkin was caught not checking SSL certificates; WhatsApp was discovered to have overlooked certificate pinning in its SSL implementation; but Apple's SSL woes outshone them both, exposing hundreds of millions of iOS and OS X Mavericks users to man-in-the-middle attacks.
Given the Snowden revelations, it's almost axiomatic that conspiracy theorists would find the fingerprints of the NSA in such events.
However, Vulture South is moved to wonder if the conspiracy we need to worry about isn't one of shadowy spooks subverting vendors or developers, but the cultural conspiracy of sleepless software development, which is almost guaranteed to deliver bad code.
It's apparent that Apple suffered a double slip: someone's command-V slipped, and it wasn't caught in any code review.
It's a mistake so simple that Vulture South is tempted to wonder whether the gung-ho no-sleep bring-coffee culture that pervades software isn't at fault. Yes, we know that the software industry has collectively spent a lot of time and energy creating and implementing processes to more-or-less bulletproof code creation. But there's also not a company or government department so boring that staging a Hackathon won't bring it some hipster credibility. Throw in a trophy and a boxed set of steak knives as first prize, and coders will show up, work ruinous hours and happily turn over their output for free.
Reward mechanisms inside companies are, I suspect, no better: it's easy to imagine that eyes bleary and bloodshot simply didn't notice the error. It's easier to see it that way than to dive into conspiracy theory.
Over at Columbia University, Steven Bellovin notes that as a piece of spook sabotage, “goto fail;” is thunderingly ham-fisted: it appears in a piece of the Apple source code base that's released under an open source license, rather than being hidden in a proprietary binary.
Bellovin adds: “There's another reason to think it was an accident: it's not very subtle. That sequence would stick out like a sore thumb to any programmer who looked at it; there are no situations where two goto statements in a row make any sense whatsoever.”
Google's Adam Langley doesn't quite agree with Bellovin about the subtlety of the error, but he points out that the bug isn't caught by the two compilers he tried it on – GCC 4.8.2 or Xcode's Clang 3.3, when run with warnings enabled. Only if the code was compiled under Clang with the -wunreachable-code flag would the bug have been noticed.
“Code review can be effective against these sorts of bug. Not just auditing, but review of each change as it goes in,” Langley writes.
But that's the kind of eagle-eyed operation that needs eyes that aren't held up by matchsticks. ®
Sponsored: Hyper-scale data management