The idea that solid state electronics work for a long time is not a big claim in any way. This is their default state. The only reason flash drives wear out at all is because of the high voltage nature of the rewrite cycle. Meanwhile the idea that a clunky mechanical device with a bunch of moving parts will work for years with no maintenance is a big claim.
MTBF only tells you half the story. It gives you the average. It does not give you the distribution. Many hard drive failures are essentially random. This means that the actual failure of an individual device can happen much earlier, or much later, than the average. Meanwhile SSDs fail almost exclusively by wearing out, on a schedule which is determined by their design, so that the actual failure will happen almost exactly as predicted.
But the real reason that SSDs and enterprise HDs have similar MTBFs is, apparently, because the HD manufacturers lie out their asses. See for example http://www.neowin.net/index.php?act=view&id=38693. Meanwhile there's no reason to believe the same of SSD manufacturers because the failure modes are well known and their claims are actually reasonable.
Your assumption is false. Just because I place a higher importance on obeying the law than on keeping my job doesn't mean I don't need a job to live or that I don't support anyone else with it.
Yes, being placed in a position where I have to break the law to keep my job would suck. But that doesn't mean that I'll choose my job. Because when the shit hits the fan (and you know it will eventually), then I'll go to jail, which is much worse for that whole support-the-family thing than merely getting fired. Meanwhile, the government has no particular mandate to protect you from situations which suck.
As for your talk about the manager asking the impossible, being a bad manager is not, and should not be, a legal offense.
I'm sure I'll catch some flack for this but... why not hold the employees who committed the illegal acts responsible?
If your choice is between breaking the law and losing your job, it should be pretty obvious that you should lose your job. If I command my employees to bring me ten million dollars by closing time, should I be liable if they then go rob a bank?
You think you're paying full price for that iphone? Get real!
Apple gets a kickback from AT&T for every iPhone subscriber. The sum of the kickback over the two-year contract amounts to roughly $400. So that $400 iPhone you purchase actually generates $800 in revenue for Apple. In countries where Apple is forced to sell the iPhone unbundled, it does indeed cost about $800 without a contract.
The iPhone is no different than any other phone in this respect, it's just more expensive than most.
That is a really interesting case, and I imagine a lot will depend on the exact details of how it was all handled. If there's a lot of wink-wink-nudge-nudge going on then it seems that they ought to be liable even if they did not explicitly give such orders. On the other hand if they were just clueless morons then they should not. I suppose it will all turn on whether you can find evidence to support the idea that this was their ultimate intent or not.
If that's how the law sees it then this pretty much lines up with my theory. Intentionally getting underlings to commit crimes is a crime in itself, not by the corporation but by the person, so that person should be personally responsible. But if the top people weren't involved then they shouldn't be personally responsible, since the whole idea of a corporation is that it takes the hits for its own actions.
I tend to agree. I think that freezing the entire corporation may be somewhat excessive, depending on exactly what the crime is, but that's the direction I think things ought to go.
If GM is convicted of some heinous environmental crime, should the CEO go to jail for it? Even if he was not involved in any way and only became aware of what was going on when he got dragged into court?
If they personally do something wrong then they should go to jail. But if their company does something wrong without their involvement then putting them in jail is simply evil. Beyond the morality of it, you will attract only pathological leadership, destroying the ability of any large company to function
This is not an MTBF measurement. Flash memory write cycles are limited by design. Writing to a flash cell involves high voltage and is inherently destructive to the material. A typical flash memory cell can only be written between 100,000 and 1,000,000 times before it breaks. This is not a mean time between failure but a total lifetime limit. Unlike a hard drive, you won't have a small percentage of flash drives which fail after only 10,000 writes, or a small percentage which survive 10,000,000 writes.
As there is basically no other mechanism which will cause failure (unless you abuse it, anyway) then stating that you can write to it continuously for 50 years seems pretty reasonable to me.
The conviction which was overturned resulted in them losing their ability to perform the audits which were their bread and butter. It's true that they turned it over voluntarily, just like Nixon resigned voluntarily; it was only done because it was better than being forced. The how and why isn't too important, the ultimate fact is that Andersen was basically destroyed directly as a result of government prosecution.
You're right that an LLP isn't quite the same thing as a corporation but the differences are not all that important for the case at hand. Andersen's status as an LLP served basically the same purposes as a regular corporation in terms of shielding its owners from liability.
I agree, that's why I mentioned the bit about personally committing crimes. If they break the law in the course of their jobs then they definitely should be put in jail. But they shouldn't be put in jail if they didn't personally break the law just because their company did.
I'm sure you're right that there is no easy answer. Generally there never is, because if there were an easy answer then somebody would have tried it already. Increasing penalties seems like a good idea. Most people's lives are effectively ruined if they are convicted of a serious crime, but a large company can shrug such a thing off without too much trouble. Seems like this is the key thing to change.
Top officers should be responsible for their own actions, but not the actions of the corporation. In other words, if a top officer gives an order to perform something illegal then they should be held responsible for that order. But if the company does something illegal but their involvement cannot be proven then they should not be held legally responsible merely because they run the place. The consequences should be placed upon the company itself, not the people in charge.
As to why this should be done, consider what would happen in the opposite case, where the top company officers can be held legally, personally responsible for illegal actions of the company they run. Consider a company like General Motors. Merely because of their size, they probably break the law in many ways constantly; any group of human beings that size will, it's just how humans are. If the top officers can be held accountable for this then no sane person will want to run the company, because it's practically a guaranteed ticket to jail. You'll only have two kinds of person willing to take over the top jobs. One kind will be absolutely risk averse. He believes the company can be run properly but will do anything and everything necessary to make sure there is no chance whatsoever that any of his underlings break the law. This will come at the expense of actually accomplishing things and making money, and thus the company will come to a halt. The other kind is even worse; you'll attract psychopaths who think that the law can't catch them, that they can maneuver the company to avoid it, or whatever. In other words, the stereotypical amoral CEO will become the actual common truth. Not a good thing.
The threat of punishment rests upon the corporation itself, and that is where it should rest. The only problem is that the threat is currently too small to really keep things in line. The answer to this is to increase the punishment, not to put the leaders in jail for things they didn't instigate.
Please don't do that. "Fixing" people's words while leaving them in quotes just makes you look like a jerk.
In any case, no organization can be completely controlled. If you force personal risk as a consequence for the actions of the organization as a whole you will completely destroy the ability of large organizations to properly function. Large companies will either cease to exist or will be captained by psychopaths, neither of which is particularly good for anybody.
Well, the government did basically nuke and pave Arthur Andersen as part of the Enron fallout, so it can happen. You're right that penalties are frequently too light, though.
My point is that agitating for direct consequences to the top officers is generally a bad idea. If they have personally committed crimes then of course they should answer for them, but they shouldn't be forced to be personally responsible for the whole organization. Many slashdotters may find this difficult to believe (not including you, I think, but lots of others), but the limitation of liability is a hugely beneficial aspect of corporations which has greatly enriched humanity and sped human progress. I'd definitely agree with harsher penalties or reduced rights, but making top officers personally responsible for the actions of the corporation will just destroy incentives to take risk and innovate legally.
My math is only "off" because you're assuming a worst case in write cycle lifetime and a completely absurd optimal case in write speed.
Write cycles on modern flash are more like one million rather than 100,000. It varies depending on the drive, though.
No way will you ever even approach filling up a 3Gbit SATA connection talking to a flash drive. More realistically you may get one tenth that speed.
Thus, ten times as many write cycles being hit at one tenth the speed you assumed gives us 100 times more lifetime, or around 100 years.
This number is of course only an estimate, done with very approximate numbers. But the point remains that even with the most pessimistic possible workload you'll still outlive the rest of your equipment.
networkBoy already explained this, but I thought it would be useful to offer a really brief, to-the-point summary.
You will always write to the whole drive. If you constantly overwrite a single logical byte, the wear leveler in the controller hardware will juggle things around so that this byte gets moved throughout the whole device so that your writes are evenly distributed.
Well, that makes sense. I would have thought it would be the opposite, though, at least for common use patterns. Flash has no seek time, so for random access writes it should be significantly faster. If you do a lot of long linear streaming writes then it could end up being much worse, though. I don't really know the numbers.
I really don't understand why everyone treats SSDs as being so fragile when writing to them. Yes, they have a limited write cycle. But so does your regular hard drive. The difference is that your SSD's cycle is guaranteed by the manufacturer, whereas your HD could blow up at any moment.
With modern wear leveling algorithms, you can write to an SSD continuously at its maximum write rate for about fifty years before you wear it out. They are, if anything, much more suitable for rapidly changing data than a regular hard drive.
Yes, it's perfectly equivalent to compare penguins, which anyone can go see at the zoo as you point out, and of which there is a great deal of evidence, which would require an enormous conspiracy by millions of people if it were made up, and which nobody really doubts, with your imaginary friend for which there is no evidence whatsoever.
I've never met anyone who's had their car NOT stolen due to a car alarm. That's kind of meaningless, because unless you reacted very fast and didn't just ignore the alarm like most people do, you'd never know that it had just scared away a thief instead of being set off by thunder or whatever.
I tend to agree there. You have to strike a balance in the middle. Languages which are too pure tend to become unusable. But languages which just do whatever looks the most useful at any given moment also become unusable.
This is why I'm a big Python fan at the moment. It strikes a nice balance between purity and practicality. If only it weren't so godawful slow....
I didn't miss it. The common C++ attitude is that much of the pain is necessary. That C++ misses the idea of "A good design is not complete when there is nothing left to add, but when there is nothing left to take away." If this isn't your position then obviously I misinterpreted, but it is the common position of C++ users in my experience.
Lisp is generally a compiled language, and with a good compiler it will be competitive with C/C++ in terms of performance. And of course its metaprogramming facilities eat C++'s for breakfast. You know you're in trouble when your language's metaprogramming facilities end up being Turing-complete by accident. In Lisp you get to write your metaprograms in exactly the same language you use for everything else, and you don't get pages full of errors when you make a minor mistake.
I also love the part where you say, "The inherent difficulty always increases with increased ability." This is such a C++ attitude! Most powerful languages are easy. Smalltalk can be learned in hours. Basic Lisp in hours. C++ is the only language in popular use which requires years of hard study to become proficient in, and even after all that effort you get nothing remarkable.
C++ is a great language if you think that effort equals results. It forces you to go to great strides to produce anything of worth, which in turn makes you feel great because you slew the beast. Meanwhile people using reasonable languages are just quietly getting things done and have no opportunity to feel good about themselves in this way.
Well, not everyone's car alarm is a twitchy SOB which constantly goes off without cause. To my knowledge, the only time my alarm has done this is when I've been foolish and accidentally triggered it myself, by setting it while I was still inside or by trying to open it without the key when I left the window open. It's possible that I'm just one of these ignorant doofuses whose alarm pisses everyone else off but I don't think so, it doesn't seem particularly twitchy.
As for a negative value of the alarm, I have a hard time believing in a careful thief who inconveniences himself in order not to cause damage to my vehicle. If he were that considerate he probably wouldn't be out stealing people's stuff in the first place.
If some insurance companies give a discount then I'd say this is a pretty good indicator that it has some useful value. After all, they're in it for the money, and they have mass statistics you and I don't have access to. That discount wouldn't be there if they got fewer claims from people who had them.
Slashdot Mirror
The idea that solid state electronics work for a long time is not a big claim in any way. This is their default state. The only reason flash drives wear out at all is because of the high voltage nature of the rewrite cycle. Meanwhile the idea that a clunky mechanical device with a bunch of moving parts will work for years with no maintenance is a big claim.
MTBF only tells you half the story. It gives you the average. It does not give you the distribution. Many hard drive failures are essentially random. This means that the actual failure of an individual device can happen much earlier, or much later, than the average. Meanwhile SSDs fail almost exclusively by wearing out, on a schedule which is determined by their design, so that the actual failure will happen almost exactly as predicted.
But the real reason that SSDs and enterprise HDs have similar MTBFs is, apparently, because the HD manufacturers lie out their asses. See for example http://www.neowin.net/index.php?act=view&id=38693. Meanwhile there's no reason to believe the same of SSD manufacturers because the failure modes are well known and their claims are actually reasonable.
Your assumption is false. Just because I place a higher importance on obeying the law than on keeping my job doesn't mean I don't need a job to live or that I don't support anyone else with it.
Yes, being placed in a position where I have to break the law to keep my job would suck. But that doesn't mean that I'll choose my job. Because when the shit hits the fan (and you know it will eventually), then I'll go to jail, which is much worse for that whole support-the-family thing than merely getting fired. Meanwhile, the government has no particular mandate to protect you from situations which suck.
As for your talk about the manager asking the impossible, being a bad manager is not, and should not be, a legal offense.
I'm sure I'll catch some flack for this but... why not hold the employees who committed the illegal acts responsible?
If your choice is between breaking the law and losing your job, it should be pretty obvious that you should lose your job. If I command my employees to bring me ten million dollars by closing time, should I be liable if they then go rob a bank?
You think you're paying full price for that iphone? Get real!
Apple gets a kickback from AT&T for every iPhone subscriber. The sum of the kickback over the two-year contract amounts to roughly $400. So that $400 iPhone you purchase actually generates $800 in revenue for Apple. In countries where Apple is forced to sell the iPhone unbundled, it does indeed cost about $800 without a contract.
The iPhone is no different than any other phone in this respect, it's just more expensive than most.
That is a really interesting case, and I imagine a lot will depend on the exact details of how it was all handled. If there's a lot of wink-wink-nudge-nudge going on then it seems that they ought to be liable even if they did not explicitly give such orders. On the other hand if they were just clueless morons then they should not. I suppose it will all turn on whether you can find evidence to support the idea that this was their ultimate intent or not.
If that's how the law sees it then this pretty much lines up with my theory. Intentionally getting underlings to commit crimes is a crime in itself, not by the corporation but by the person, so that person should be personally responsible. But if the top people weren't involved then they shouldn't be personally responsible, since the whole idea of a corporation is that it takes the hits for its own actions.
I tend to agree. I think that freezing the entire corporation may be somewhat excessive, depending on exactly what the crime is, but that's the direction I think things ought to go.
If GM is convicted of some heinous environmental crime, should the CEO go to jail for it? Even if he was not involved in any way and only became aware of what was going on when he got dragged into court?
If they personally do something wrong then they should go to jail. But if their company does something wrong without their involvement then putting them in jail is simply evil. Beyond the morality of it, you will attract only pathological leadership, destroying the ability of any large company to function
This is not an MTBF measurement. Flash memory write cycles are limited by design. Writing to a flash cell involves high voltage and is inherently destructive to the material. A typical flash memory cell can only be written between 100,000 and 1,000,000 times before it breaks. This is not a mean time between failure but a total lifetime limit. Unlike a hard drive, you won't have a small percentage of flash drives which fail after only 10,000 writes, or a small percentage which survive 10,000,000 writes.
As there is basically no other mechanism which will cause failure (unless you abuse it, anyway) then stating that you can write to it continuously for 50 years seems pretty reasonable to me.
The conviction which was overturned resulted in them losing their ability to perform the audits which were their bread and butter. It's true that they turned it over voluntarily, just like Nixon resigned voluntarily; it was only done because it was better than being forced. The how and why isn't too important, the ultimate fact is that Andersen was basically destroyed directly as a result of government prosecution.
You're right that an LLP isn't quite the same thing as a corporation but the differences are not all that important for the case at hand. Andersen's status as an LLP served basically the same purposes as a regular corporation in terms of shielding its owners from liability.
I agree, that's why I mentioned the bit about personally committing crimes. If they break the law in the course of their jobs then they definitely should be put in jail. But they shouldn't be put in jail if they didn't personally break the law just because their company did.
I'm sure you're right that there is no easy answer. Generally there never is, because if there were an easy answer then somebody would have tried it already. Increasing penalties seems like a good idea. Most people's lives are effectively ruined if they are convicted of a serious crime, but a large company can shrug such a thing off without too much trouble. Seems like this is the key thing to change.
Top officers should be responsible for their own actions, but not the actions of the corporation. In other words, if a top officer gives an order to perform something illegal then they should be held responsible for that order. But if the company does something illegal but their involvement cannot be proven then they should not be held legally responsible merely because they run the place. The consequences should be placed upon the company itself, not the people in charge.
As to why this should be done, consider what would happen in the opposite case, where the top company officers can be held legally, personally responsible for illegal actions of the company they run. Consider a company like General Motors. Merely because of their size, they probably break the law in many ways constantly; any group of human beings that size will, it's just how humans are. If the top officers can be held accountable for this then no sane person will want to run the company, because it's practically a guaranteed ticket to jail. You'll only have two kinds of person willing to take over the top jobs. One kind will be absolutely risk averse. He believes the company can be run properly but will do anything and everything necessary to make sure there is no chance whatsoever that any of his underlings break the law. This will come at the expense of actually accomplishing things and making money, and thus the company will come to a halt. The other kind is even worse; you'll attract psychopaths who think that the law can't catch them, that they can maneuver the company to avoid it, or whatever. In other words, the stereotypical amoral CEO will become the actual common truth. Not a good thing.
The threat of punishment rests upon the corporation itself, and that is where it should rest. The only problem is that the threat is currently too small to really keep things in line. The answer to this is to increase the punishment, not to put the leaders in jail for things they didn't instigate.
Please don't do that. "Fixing" people's words while leaving them in quotes just makes you look like a jerk.
In any case, no organization can be completely controlled. If you force personal risk as a consequence for the actions of the organization as a whole you will completely destroy the ability of large organizations to properly function. Large companies will either cease to exist or will be captained by psychopaths, neither of which is particularly good for anybody.
Well, the government did basically nuke and pave Arthur Andersen as part of the Enron fallout, so it can happen. You're right that penalties are frequently too light, though.
My point is that agitating for direct consequences to the top officers is generally a bad idea. If they have personally committed crimes then of course they should answer for them, but they shouldn't be forced to be personally responsible for the whole organization. Many slashdotters may find this difficult to believe (not including you, I think, but lots of others), but the limitation of liability is a hugely beneficial aspect of corporations which has greatly enriched humanity and sped human progress. I'd definitely agree with harsher penalties or reduced rights, but making top officers personally responsible for the actions of the corporation will just destroy incentives to take risk and innovate legally.
My math is only "off" because you're assuming a worst case in write cycle lifetime and a completely absurd optimal case in write speed.
Write cycles on modern flash are more like one million rather than 100,000. It varies depending on the drive, though.
No way will you ever even approach filling up a 3Gbit SATA connection talking to a flash drive. More realistically you may get one tenth that speed.
Thus, ten times as many write cycles being hit at one tenth the speed you assumed gives us 100 times more lifetime, or around 100 years.
This number is of course only an estimate, done with very approximate numbers. But the point remains that even with the most pessimistic possible workload you'll still outlive the rest of your equipment.
networkBoy already explained this, but I thought it would be useful to offer a really brief, to-the-point summary.
You will always write to the whole drive. If you constantly overwrite a single logical byte, the wear leveler in the controller hardware will juggle things around so that this byte gets moved throughout the whole device so that your writes are evenly distributed.
That's pretty much the whole idea of having a corporation.
Well, that makes sense. I would have thought it would be the opposite, though, at least for common use patterns. Flash has no seek time, so for random access writes it should be significantly faster. If you do a lot of long linear streaming writes then it could end up being much worse, though. I don't really know the numbers.
I really don't understand why everyone treats SSDs as being so fragile when writing to them. Yes, they have a limited write cycle. But so does your regular hard drive. The difference is that your SSD's cycle is guaranteed by the manufacturer, whereas your HD could blow up at any moment.
With modern wear leveling algorithms, you can write to an SSD continuously at its maximum write rate for about fifty years before you wear it out. They are, if anything, much more suitable for rapidly changing data than a regular hard drive.
Yes, it's perfectly equivalent to compare penguins, which anyone can go see at the zoo as you point out, and of which there is a great deal of evidence, which would require an enormous conspiracy by millions of people if it were made up, and which nobody really doubts, with your imaginary friend for which there is no evidence whatsoever.
I tend to agree there. You have to strike a balance in the middle. Languages which are too pure tend to become unusable. But languages which just do whatever looks the most useful at any given moment also become unusable.
This is why I'm a big Python fan at the moment. It strikes a nice balance between purity and practicality. If only it weren't so godawful slow....
I didn't miss it. The common C++ attitude is that much of the pain is necessary. That C++ misses the idea of "A good design is not complete when there is nothing left to add, but when there is nothing left to take away." If this isn't your position then obviously I misinterpreted, but it is the common position of C++ users in my experience.
If what you say is true, why aren't EVDO and other wireless data services censored?
Lisp is generally a compiled language, and with a good compiler it will be competitive with C/C++ in terms of performance. And of course its metaprogramming facilities eat C++'s for breakfast. You know you're in trouble when your language's metaprogramming facilities end up being Turing-complete by accident. In Lisp you get to write your metaprograms in exactly the same language you use for everything else, and you don't get pages full of errors when you make a minor mistake.
I also love the part where you say, "The inherent difficulty always increases with increased ability." This is such a C++ attitude! Most powerful languages are easy. Smalltalk can be learned in hours. Basic Lisp in hours. C++ is the only language in popular use which requires years of hard study to become proficient in, and even after all that effort you get nothing remarkable.
C++ is a great language if you think that effort equals results. It forces you to go to great strides to produce anything of worth, which in turn makes you feel great because you slew the beast. Meanwhile people using reasonable languages are just quietly getting things done and have no opportunity to feel good about themselves in this way.
Well, not everyone's car alarm is a twitchy SOB which constantly goes off without cause. To my knowledge, the only time my alarm has done this is when I've been foolish and accidentally triggered it myself, by setting it while I was still inside or by trying to open it without the key when I left the window open. It's possible that I'm just one of these ignorant doofuses whose alarm pisses everyone else off but I don't think so, it doesn't seem particularly twitchy.
As for a negative value of the alarm, I have a hard time believing in a careful thief who inconveniences himself in order not to cause damage to my vehicle. If he were that considerate he probably wouldn't be out stealing people's stuff in the first place.
If some insurance companies give a discount then I'd say this is a pretty good indicator that it has some useful value. After all, they're in it for the money, and they have mass statistics you and I don't have access to. That discount wouldn't be there if they got fewer claims from people who had them.