edit: My point is that it was not one lone actor, who would have made that change.
It’s just worded to make this sound sketchy. I bet ten bucks “unlinked” just refers to the standard POSIX call for deleting a file.
Also this is not like some process crash dump where the computer keeps running after one process crashed.
This would be like an plane black box uploading its data to the manufacturer, then deleting itself after a plane crash.
Deleting the data on the server is totally sketchy, but that’s not what the quoted section is about.
How Tesla could say that detecting a collision and not locking all/any of the data is normal is just insane.
The real issue is that Tesla claimed the company doesn't have the data after every copy was deleted. There's no technical reason to dispose of data related to a crash when you hold so much data on all of the cars in general.
Crash data in particular should be considered sacred, especially given the severity in this case. Ideally it should be kept both on the local black box and on the servers. But anything that leads to it being treated as instantly disposable everywhere, or even just claiming it was deleted, can only be malice.
Exactly. The issue is deleting the data on the servers, not a completely mundane upload-then-delete procedure for phoning home. This should have been one sentence, but instead they make it read like a heist.
My money is on nobody built a tool to look up the data, so they have it, they just can't easily find it.
This data is yours. You were going the speed limit when the accident happened and everyone else claims you were speeding. It would take forever to clear your name or worse you could be convicted if the data was lost.
This is more of "you will own nothing" crap. And mainly so Tesla can cover its ass.
Maybe this is not appropriate for a car, but that doesn’t excuse the ridiculous breathless tone in the quoted text. It’s the worst purple prose making a boring system sound exciting and nefarious. They could have made your point without trying to make the unlink() call sound suspicious.
The rogue engineer defense worked so well for VW and Dieselgate.
The issue of missing crash data was raised repeatedly. Deleting or even just claiming it was deleted can only be a mistake the first time.
So either the problem is Tesla engineers are fucking stupid (doubtful) or this is a poor business/product design.
My money is on the latter.
So we just shrug because software boys gotta be software boys? That’s completely insane and a big reason why a lot of engineers roll their eyes about developers who want to be considered engineers.
Software engineers who work on projects that can kill people must act like the lives of other people depend on them doing their job seriously, because that is the case. Look at the aviation industry. Is it acceptable to have a bug in the avionics suite down planes at random and then delete the black boxes? It absolutely is not, and when anything like that happens shit gets serious (think 737 MAX).
The developers who designed the systems are responsible, and so are their managers who approved the changes, all the way to the top. This would not happen in a company with appropriate processes in place.
I'd argue that this data is far less important in cars. Airline safety has advanced to the point where crashes are extremely rare and usually have a novel cause. Data recorders are important to be able to learn that cause and figure out how to prevent it from happening again. Car safety, on the other hand, is shit. We don't require rigorous training for the operators. Regulations are lax, and enforcement even more lax. Infrastructure is poor. We're unwilling to fix these things. Almost all safety efforts focus on making the vehicles more robust when collisions occur, and we're just starting to see some effort put into making the vehicles automatically avoid some collisions. What are we going to learn from this data in cars? "Driver didn't stop for a red light, hit cross traffic." "Driver was drunk." "Driver failed to see pedestrian because of bad intersection design which has been known for fifty years and never been fixed." It's useful for assigning liability but not very useful for saving lives. There's a ton of lower hanging fruit to go after before you start combing through vehicle telemetry to find unknown problems.
Even if you do consider it to be life-critical, uploading the data and then deleting the local copy once receipt is acknowledged seems completely fine, if the server infrastructure is solid. Better than only keeping a local copy, even. The issue there is that they either have inadequate controls allowing data to be deleted, or inadequate ability to retrieve data.
I worked a year in airbag control, and they recorded a bit of information if the bags were deployed - seatbelt buckle status was one thing. But I was under the impression there was no legal requirement for that. I'm sure it helps in court when someone tries to sue because they were injured by an airbag. The argument becomes not just "the bags comply with the law" but also "you weren't wearing your seatbelt". Regardless, I'm sure Tesla has a lot more data than that and there is likely no legal requirement to keep it - especially if it's been transferred reliably to the server.
This is like saying "maybe nobody has recently looked at the ad-selection mechanism at Google." That's just not plausible.
The performance ship sailed, like, 15 years ago. We're already storing about 10000000 more data than we need. And that's not even an exaggeration.
It's very easy to imagine a response to this being (beyond "don't log so much") an audit layer to start automatically removing redundant data.
The externalities of the company are such that people want to ascribe malice, but this is a very routine kind of thing.
> Tesla later said in court that it had the data on its own servers all alongPerhaps if there is some sort of crash.
Which of these is evidence of a conspiracy:
tar cf - | curl
TMPFILE=$(mktemp) ; tar cf $TMPFILE ; curl -d $TMPFILE ; rm $TMPFILEThe requirements should have been clear that crash data isn't just "implement telemetry upload", a "collision snapshot" is quite clearly something that could be used as evidence in a potentially serious incident.
Unless your entire engineering process was geared towards collecting as much data that can help you, and as little data as can be used against you, you'd handle this like the crown jewels.
Also, to nit-pick, the article says the automated response "marked" for deletion, which means it's not automatically deleted as your reductive example which doesn't verify it was successfully uploaded (at least && the last rm).
Any competent engineer who puts more than 3 seconds of thought into the design of that system would conclude that crash data is critical evidence and as many steps as possible should be taken to ensure it's retained with additional fail safes.
I refuse to believe Tesla's engineers aren't at least competent, so this must have been done intentionally.
Assuming its not intentionally malicious this is a really dumb bug that I could have also written. You zip up a bunch of data, and then you realize that if you don't delete things you've uploaded you will fill up all available storage, so what do you do? You auto delete anything that successfully makes it to the back-end server, you mark the bug fixed, not realizing that you overlooked crash data as something you might want to keep.
I could 100% see this being what is happening.