6 ms·
How does the distribution mechanism in itself affect the quality? It's just a way of getting new software on the car. You can have the exact same standards of
by peeters 13y ago
How does the distribution mechanism in itself affect the quality? It's just a way of getting new software on the car. You can have the exact same standards of quality for an update installed over the air as one installed at a service centre.
- smackfu 13y agoThe easier it is to update software, the less important that it is bug-free.
- npizzolato 13y agoUntil that bug causes a car to drive unexpectedly into a wall in the interim time until you've sent out a patch. The importance of being bug-free is much more driven by the potential damage caused by a bug than how quickly you can update the software.
- GenerocUsername 13y agoFear monger. What if what if what if... I'm not saying it couldn't happen, I'm just saying that possible new issues are no reason to persist the old issues
- throwaway2048 13y agoThis is utter nonsense, "possible new issues" are every reason to carefully evaluate and weight new options.
- sukuriant 13y agoNo. It's a car, and has the power to kill people on software malfunction, making it a critical system. Critical systems are tend to go through a much greater and more robust form of testing.
- stcredzero 13y agoThe destructive potential of a car exceeds that of a lot of handguns one can buy. There have long been techniques for writing software as formal proofs of correctness. As far as I can tell, they haven't been popular with mainstream programmers because they are simply too different.
- sukuriant 13y agoThat is a really neat bit of fact; but, what does that have to do with what I was talking about? I really would like to know, it sounds like a really important insight, I'm just not sure what train of thought led to that.
- Fomite 13y agoNo, but "Cars do uncomfortable things at high velocity" is not exactly an edge case.
- sliverstorm 13y agoRight. So, what happens when you are the one in that car, the car that teaches them better?
- zobzu 13y agodo you think that the gasoline cars don't have software updates? their engine is nowadays all software controlled and theres a fw update plug, generally under your driving wheel. and they'll update it when they service the car. how is that software any safer than the OTA? its not.
- npizzolato 13y agoI never said that. I was responding to "The easier it is to update software, the less important that it is bug-free."
- MartinCron 13y agoI would say that the easier it is to update software, the more likely it is to be bug free. It's easier to test one change in isolation than to try to figure out which of these 20/200/2000 commits broke things.
- ssmoot 13y agoExcept at a service center there's presumably the opportunity to replace the firmware chip or board if they brick it during the update whereas a bad OTA update leaves you little option. OTOH I'd risk a bad OTA update 1% of the time if it meant avoiding the service center the other 99% of the time. And I'm sure a bad update is just a call and a tow away from a fix at a Tesla service center. Unless you're on a road-trip I guess.
- stcredzero 13y agoWith the march of Moore's Law, it should currently be feasible for automobile controllers to actually run under a kind of hypervisor and keep a copy of the old system so that it can roll back. (Including by request of the owner.)
- cjfont 13y agoYou might want to turn off updates during road trips.
- stcredzero 13y agoHow about registering your "garage coordinates" with your car's computer so that it can detect when you're on a road trip and give you the option of installing an update or not? (When you are stopped and have parked the car, of course.)
- samstave 13y ago
- rst 13y agoFor one thing, an OTA update mechanism is obviously more subject to third-party spoofing than an update delivered by physical ethernet from the dealership. (BTW, do Teslas apply OTA updates while the car is underway, or store them up for application when parked? If the former, then applying even a correct update to a car in motion could have all sorts of nasty consequences if it resulted in a sudden change to, say, response characteristics of the suspension or brakes.)
- peeters 13y agoIs that true though? I mean, in my mind OTA makes it a given that the firmware is going to be fully signed and the channel will be fully authenticated. Whereas if a dealer is installing it, they could dismiss concerns with "we trust our service technicians." Seems the latter opens the door to a lot of social engineering possibilities. But I was responding to a claim of the update being "poorly tested", not insecure. That's orthogonal to it being over the air. A poorly tested update delivered through a service center can also brick your car.
- GauntletWizard 13y agoI'm sorry, but no. No, most manufacturers update processes aren't actually secure in any meaningful way. Tesla's OTA process seems to use (I've not checked it myself) web standards, encryption, and signatures to verify that the OTA update is correct. The "Physical ethernet cable" that is used by traditional manufacturers can be done by any hobo on the street, in pretty much the same time. It's really only a matter of time before a script-kiddie-ready device that you plug-in and starts the car shows up; There's no security, no signature verification, etc. on the traditional manufacturers updates.
- oh_sigh 13y agoSource please regarding teslas OTA process.
- secabeen 13y agoHere's a pretty good summary from someone who sniffed the WiFi traffic from a Tesla doing an OTA update: http://www.teslamotorsclub.com/showthread.php/22294-Sniffing-the-WiFi-traffic http://www.teslamotorsclub.com/showthread.php/22294-Sniffing...