|
Authored by wayneborean on Sept 23, 2015 16:29:56 GMT
|
|
Ian Al
Guest
|
Authored by Ian Al on Sept 24, 2015 5:32:45 GMT
I've always tended to consider Eben Moglen as an Open Source fundamentalist; too extreme for real world IT. SystemD and the VAG scam show me that we lose if we don't stick to the principles he advocates.
|
|
|
Authored by wayneborean on Sept 24, 2015 6:41:04 GMT
I've always tended to consider Eben Moglen as an Open Source fundamentalist; too extreme for real world IT. SystemD and the VAG scam show me that we lose if we don't stick to the principles he advocates. And I wonder how many other things are being hidden that we don't know about because we can't see the code. Wayne
|
|
|
Authored by wol on Sept 24, 2015 8:29:33 GMT
In defence of Volkswagen, do the regulators actually understand the regulations?! How do you define what the emissions "should" be? WAS THERE ANY *DELIBERATE* INTENT TO DEFRAUD? Or is this whole thing a storm in a teacup because the regulators don't understand the tech, and nor do the PHB's?
Let's look at the facts. The Engine Management Software engaged a profile that was intended to minimise emissions, because it (allegedly) was programmed to recognise the fact that it was on a test-bed. Or was it? Given that the system is DESIGNED to choose the MOST APPROPRIATE profile for the car based on the situation the car was in, how on earth is the test bed *ever* going to return a "valid" reading? How do you define "valid"?
And it would not surprise me in the slightest if there was a profile meant for "boy racer revving the engine at the lights", which is quite likely going to be a close approximation to "car sitting on a test-bed". Surely it makes sense, if the system detects that the car is doing nothing, to select a profile that minimises emissions? No intent to deceive whatsoever. Every intent to comply with the regs. And at the end of the day, what f*** is the testbed designed to test for? Anybody with half a brain knows that's it's a completely artificial environment that returns results that meaningless results! The "mpg in urban cycle" and "mpg at 56mph" are useful figures, sure, but anybody who expects to actually GET that mpg is an idiot. Likewise, anybody who actually expects to get the emissions as reported on a test-bed is an idiot.
To my mind, the car was in all likelihood programmed to detect that the engine was not doing anything much, so it selected the highest eco profile it should. A lot of newer eco engines would have switched the engine off! That would really have messed up the test results! :-)
Cheers, Wol
|
|
|
Authored by wol on Sept 24, 2015 8:40:17 GMT
SystemD and the VAG scam show me that we lose if we don't stick to the principles he advocates. What SystemD scam? To my mind, the fuss over SystemD is driven by a bunch of conspiracy theorists with a "we hate Lennart" agenda. The trouble with Lennart is he doesn't agree with the sendmail philosophy of "be liberal in what you accept, and strict in what you send". Quite the opposite, in fact - "if upstream is broke, don't band-aid about it. Get them to fix it!". That's ruffled far too many feathers in the past, and people are now out to get him. Obligatory disclaimer - my systems don't run SystemD, because it's not the distro default. They run OpenRC, and I'm not going to risk switching a production system. Oh - and yes SystemD is more complicated than init. But the SystemD boot setup is a lot simpler that an init boot setup. I wish they'd fix the niggles in OpenRC that are biting me - nfs and mysql (although the mysql thing is probably mysql, not nfs). Cheers, Wol
|
|
Ian Al
Guest
|
Authored by Ian Al on Sept 24, 2015 14:50:13 GMT
To my mind, the fuss over SystemD is driven by a bunch of conspiracy theorists with a "we hate Lennart" agenda. I actually use SystemD (I didn't actually call it a scam) because it is the line of least resistance in order to run my favourite Linux distro, Kubuntu. My understanding was that SystemD comprised binary blobs for which the source was not available and that it also includes an increasing array of features like a mini-database function. Because it is not truely open, no-one can stick a fork in it (I think that's the expression) and dig out the superfluous functions. If it is open such that anyone (with the skills) can fork it, then I agree with you and would tell the anti-Lennart brigade to fork up or shut up.
|
|
|
Authored by cpeterson on Sept 24, 2015 15:34:19 GMT
In defence of Volkswagen, [...] ... WAS THERE ANY *DELIBERATE* INTENT TO DEFRAUD? [...] And it would not surprise me in the slightest if there was a profile meant for "boy racer revving the engine at the lights",[...] To my mind, the car was in all likelihood programmed to detect that the engine was not doing anything much, so it selected the highest eco profile it should. [...] Cheers, Wol I think that if Volkswagen's actions were innocent in the manner you propose, they would have defended it as such. I don't think they would have said "We lied" if they hadn't actually lied.
|
|
|
Authored by wol on Sept 24, 2015 15:37:17 GMT
My understanding was that SystemD comprised binary blobs for which the source was not available and that it also includes an increasing array of features like a mini-database function. Because it is not truely open, no-one can stick a fork in it (I think that's the expression) and dig out the superfluous functions. If it is open such that anyone (with the skills) can fork it, then I agree with you and would tell the anti-Lennart brigade to fork up or shut up. The old init system was written in bash shall etc - ie an interpreted language anyone could read. SystemD (to an extremely crude approximation) simply took all those bits of shell that got repeatedly cut-n-pasted from one init script to another and converted them into C helper progs (GPL'd). It did a few other things as well, admittedly. udev, for example, causes grief on gentoo (my system) because SystemD and udev are now one codebase. You can build udev without SystemD no problem, but because the source is so tightly intertwined they have to develop in lock-step. And the other stuff it's done is to standardise all those little "quirks" that distros have - like where the local host name is stored. That's put a lot of noses out of joint. Basically, think of SystemD as a collection of helper tools, all written in C. And a SystemD resource file is circa 10 lines telling the boot system which helper functions to call. As opposed to traditional SysVinit which is all those helper tools, written out longhand, hundreds of times, with unfixed bugs in many of the implementations, etc etc etc. The classic bug I came across was in a Bind shutdown script. Somebody reported it on lwn, but basically, the bind script would first kill the network, then loop until some service shut down. If for whatever reason, that service FAILED to shut down, your system was now in an infinite loop with no way of contacting it except by going to the console. And for the person in question, that console was 2000 miles away. Oh f***! SystemD, in its init program, simply has a timeout that will kill ANY and ALL processes to enable shutdown to complete. (Oh, and when he looked, he found the same bug in several other program init scripts ...) (And I think the service that failed to shut down was NFS, which should be shut down *before* the network, not after. Oh - and which is easy to get into an "infinite mount loop".) If you want to audit SystemD, go ahead. It's all Free/Libre code. The binary blobs you refer to are probably the same binary blobs that you find in the kernel - device drivers that must be loaded into hardware and without which the hardware won't function - and which you'll find in the equivalent SysVinit system too. Unless of course, those binary blobs are the product of the shell gurus moaning that they can't read object code and can't be bothered to learn C ... Oh - and SystemD IS growing Topsy-like and pulling loads of stuff into its orbit - because it makes sense. Like specifying a standard location for hostname. Like setting system time. Like taking over fstab. Like lots of sysadmin type stuff. Thing is, all those things are *modules*. That can be replaced if you are sufficiently motivated. Like most people can't be arsed - if it ain't broke don't fix it. That also annoys the "everything should be a shell-script we can tweak" brigade, but most of us are glad we don't have to. There have been repeated calls from the SystemD folk to the effect of "if you don't like it, do something about it, don't ****". Problem is, very few people have done anything about it, and SystemD has pretty much steam-rollered those who have. That's why my systems run OpenRC - someone did something about it. And frankly, I suspect I'd have been better off if they hadn't - my systems aren't complicated but some stuff is broken, and I expect it gets worse with more complicated setups ... Cheers, Wol
|
|
|
Authored by wol on Sept 24, 2015 15:43:50 GMT
In defence of Volkswagen, [...] ... WAS THERE ANY *DELIBERATE* INTENT TO DEFRAUD? [...] And it would not surprise me in the slightest if there was a profile meant for "boy racer revving the engine at the lights",[...] To my mind, the car was in all likelihood programmed to detect that the engine was not doing anything much, so it selected the highest eco profile it should. [...] I think that if Volkswagen's actions were innocent in the manner you propose, they would have defended it as such. I don't think they would have said "We lied" if they hadn't actually lied. That's basically my point about both the regulators and PHB's not understanding the tech. The car was designed to detect scenarios and load the appropriate profile. As an engineer, I would have no qualms whatsoever about detecting a situation where the car was stopped with the engine racing and loading a profile designed to minimise emissions. I *would* have a problem with "detecting a testbed and loading a profile to deceive". That's why I find it hard to believe the intent was to deceive. It could so easily be innocent - engine racing when stationary isn't unusual. But the car DID load a low-emissions profile. And it DID detect a test-bench environment. But, in the developer's mind, was it LOOKING for a test-bench environment. That I find hard to believe. But I also find it easy to believe that the regulators and PHBs could assume it was looking for it. Cheers, Wol
|
|
|
Authored by cpeterson on Sept 24, 2015 16:24:39 GMT
I think that if Volkswagen's actions were innocent in the manner you propose, they would have defended it as such. I don't think they would have said "We lied" if they hadn't actually lied. That's basically my point about both the regulators and PHB's not understanding the tech. The car was designed to detect scenarios and load the appropriate profile. As an engineer, I would have no qualms whatsoever about detecting a situation where the car was stopped with the engine racing and loading a profile designed to minimise emissions. I *would* have a problem with "detecting a testbed and loading a profile to deceive". That's why I find it hard to believe the intent was to deceive. It could so easily be innocent - engine racing when stationary isn't unusual. But the car DID load a low-emissions profile. And it DID detect a test-bench environment. But, in the developer's mind, was it LOOKING for a test-bench environment. That I find hard to believe. But I also find it easy to believe that the regulators and PHBs could assume it was looking for it. Cheers, Wol In order for that to fly, you'd have to also assume that the engineers didn't understand that the test-bench situation was supposed to simulate a "steady driving on the road" situation. Once you reach that point of ignorance, there's no longer any dichotomy between dumb and evil. Stupidity and malice merge into a single, indistinguishable continuum; trying to parse them out from each other only proliferates the false notion that one of them is more forgivable than the other.
|
|
Bubba
Guest
|
Authored by Bubba on Sept 24, 2015 16:46:25 GMT
Talking with a fellow engineer who used to work in emissions years ago, his statement was that every car company "plays" with the emissions methods in one way or another, and everyone in the industry wondered how VW was meeting the emissions requirements without urea injection.
They knew what they were doing, they knew that it broke the spirit of the law, and they did it anyway.
|
|
|
Authored by wol on Sept 24, 2015 18:50:43 GMT
everyone in the industry wondered how VW was meeting the emissions requirements without urea injection. They knew what they were doing, they knew that it broke the spirit of the law, and they did it anyway. Just heard it on the news - apparently they WERE using urea injection, and they did specifically detect a test-bench, and they only used it for that. So it seems I was misguided in blaming incompetence rather than malice ... NOT GOOD! Cheers, Wol
|
|
|
Authored by wayneborean on Sept 24, 2015 22:47:48 GMT
In defence of Volkswagen, do the regulators actually understand the regulations?! How do you define what the emissions "should" be? WAS THERE ANY *DELIBERATE* INTENT TO DEFRAUD? Or is this whole thing a storm in a teacup because the regulators don't understand the tech, and nor do the PHB's? Yes, they do understand the regulations. I worked in the industry for years, dealing with Environment Canada, the Environmental Protection Agency, and the California Air Resources Board. All three agencies employ competent engineers. I know a bunch of them. The engine companies also understand the rules. Yes, we designed engines to pass the tests. But the regulations made it clear that the engine had to run that way in real life. As Bubba said, there's been rumours about some companies. I hadn't heard the ones about VW, but I've been on disability for the last five years. In at least one case involving heavy trucks, the engine companies were found guilty, and fined. So i'm really surprised that VW was stupid enough to do this. Wayne
|
|