[0] ftp://ftp.stlinux.com/pub/stlinux/2.4/
[1] http://www.nxp.com/products/microcontrollers-and-processors/...
We're doing this wrong.
The root of the problem is: you buy a part, and to use the part you need some software to talk to it. The vendor, along with the part, supplies you with some sample code, which uses some ancient kernel and ancient GPLv2 utilities. If you use their code, then you perpetuate the problem. If you choose not to, then they won't help you when you have issues.
Why do vendors use such ancient software? Vendors care deeply about their IP, and to some extent tivoization does happen with individual products, but more usually through obfuscation. Not a whole lot of open firmware for wireless chipsets, is there?
Customer demand can change things, but vendors are afraid of their customers stealing their technology, and taking it to a cheaper competitor, or building it themselves. This stuff happens all the time. So they give you some code that talks some foreign language to their part or give it a seemingly black box blob of code to run and suddenly it's a black box to you too.
Patents (for hardware), and China's lack of concern for IP theft are the root of all of this.
Vendors might at some time chose to say "freeze" and now we will just deliver this same binary blob, or at best generate new binary blobs from an ancient source stack we have under version control.
Making changes means, at best, going through an expensive testing procedure. At worst, there is no such procedure and all change is an anethema.
So they only make changes when they have to, and as the cost of staying obsolete increases, so does the cost of becoming up to date.
Many times I've spent late nights babysitting an employee from $VENDOR debugging/fixing/rewriting their driver to play nice with our stack.
Indeed, quite true.
Though none of this excuses any negligence, no matter if it is the vendor of the manufacturer.
I hear you, but isn't this one of the reasons for wanting compliance around the (L)GPL? It allows us, the owners of the vehicle and users of the software to make that inspection, and hopefully, patch the vulnerabilities in the software, right? (I believe one of the provisions of the LGPL is that it must be possible to replace the LGPL'd software with a different version — this is why most software dynamically links to LGPL libraries: it allows the .so/.dll to be "simply" swapped out for a different one.)
I mean, obviously some people at Volvo thought that actual emissions compliance was far too costly to bother with --and that didn't even result in a vehicle accident or personal injury. Certainly their bottom line has had a major impact as a result. The potential for PR disaster with Internet-connected vehicles is certainly much more concerning.
The inevitability of some such future tragedy and resulting media-fueled PR disaster for one or more vehicle manufactures is a certainty. It's only a matter of when and which vehicle manufacturer(s) will be liable. When that happens, I’m certain that such sizable companies can figure out how to streamline the approval process with whatever regulatory bodies might be involved, or they’ll just learn to live with the pain… for being for less than the much greater pain of a PR disaster.
It's the infotainment system according to one of the original blog posts. And in fairness the previous post was all about the updates they were providing - albeit unencrypted.
What do you mean by this? The car itself is a single unit so you can assume the electronics inside communicate at some level (CAN networks usually), but the electronics are separated. You have a control unit for the engine, another one for the ABS, another for the gearbox and many others for various body control functions. Some of them communicate with each other over the CAN, but only transmit/receive messages of particular interest. For example, the ABS unit in some cases tells the engine control unit the vehicle speed. The engine control unit transmits the engine RPM to the body control unit for obvious reasons. The same for the anti-theft mechanism.
To the subject: Having outdated drivers in production level software is common practice. During final validation of the SW, before start of production, if some drivers are found outdated (even declared invalid by the driver/package owner), people do not jump to upgrade immediately. It undergoes a process of analysis: is there any critical bug which the new version fixes? Does it impact the current project configuration? If not, why upgrade? You run the risk of introducing a regression => requires extra validations which can take weeks or months. Nobody will approve that just to fix some bug without functional impact in that particular system. So when you claim that a certain package has to be updated to the latest version, you have to come up with arguments. "Newest is always better" does not count. There are countless examples when newer versions of some piece of SW fixed some bug but broke something that was working. SW updates after the majority of the validation process is finished happens only on a NEED TO HAVE basis.
All of this is just crying wolf so far.
But sure - best practices should be followed.
Look at the graph. http://finance.yahoo.com/echarts?s=VOW.DE+Interactive#{"allo...
Michael Hastings, anyone?
It wouldn't surprise me if the software had a similarly long development and lock-in cycle.
It is after all an Internet-connected vehicle.
Despite the article author's recognition given for BMW's OSS compliance, I am dismayed to see yet another example of likely inconsideration given toward implementation of security in the engineering of an Internet-connected vehicle.
This sort of negligence is eventually going to result in some rather ugly outcome.