It's a simplification on many levels and just shows that the magnitude of the claimed amount is about right.
Computers sold in the Energy Star era (post-1992) all had some form of power stepping. In all fairness, the early forms were a nuisance and were often disabled. More recent computers, minimally, step down the processor clock when there's nothing to do. Typical desktop CPU power draw, full CPU utilization compared to idle but still running the OS, has varied, peaking several years ago and now dropping somewhat as peak CPU power draw has come down a little. You lose 5-10% in the power supply. 40 watts is perhaps a little low, as an average over the past 10 years.
There is also the additional air conditioning load, which is typically 30% for the fraction of time air conditioning is in use in the building.
There is also additional fan wear since the fans will run harder. And the additional heatsink clogging due to dust that a technician must remove. Quite possibly, the complaints of slow PCs were due to the CPU speed being stepped down due to heat limiting because of fouled heatsinks from dust combined with the CPU hog program keeping everything running hot.
There is also more rapid component failure due to heat. This affects the CPU itself, and depending on the case design, possibly other components as well.
Finally, with an IT mindset rooted in the importance of running SETI, it is unlikely that other power-saving strategies, like shutting off machines at night or on weekends, or allowing them to enter sleep or hibernate states, were seriously entertained, though to be sure such policies don't always make sense.
From TFA: "Gilbert police are now involved in the investigation and criminal charges may be filed."
How is this criminal? He had legitimate access. What's special about the scale? If someone ran a single instance of SETI@home on the PC on their desk, would that be criminal?
In an ideal world, none of the techs are on call because there are sufficient scheduled staff on site to deal with any problems. Larger organizations approach this, simply because beyond a certain point the "on call" person is busy enough that the model doesn't work anymore and you have to switch to shifts.
Small companies I've worked at have used a number of strategies to make this fair. One way to do it is limit service hours. For a while we limited service hours to 7 am to 7 pm, and if customers called outside that the expectation was that they'd get a call at 7 am the next day. I have also seen best-effort systems where there is no formal SLA, and several people get paged at once if there's trouble, which works in some cases.
The problem with being on call is that, if you have a life, you have commitments and do stuff that you can't just unwind in 10 minutes. Golf, fishing trip, reffing a kid's softball game, community orchestra rehearsal, picking up kids on the other side of town.
Chinese internet filtering is justified publicly by stating that it is done to help Chinese people avoid inadvertent violations of the law, and that is how it is seen by most Chinese. The real purpose of the censorship there is to facilitate prosecution of dissidents by making it impossible to violate laws against anti-government speech and unlawful assembly inadvertently.
In general, lone inventors don't benefit much from patents.
I used to work for $BIG_COMPANY where we had a process where a team of lawyers and engineers would evaluate everything someone thought remotely patentable, and they would consider whether the patent was worth the time and money (usually several thousand dollars even though they used in-house counsel and did thousands of these a year). The analysis was based not only on the validity of the patent and the value of the innovation covered, but also the difficulty of creating a comparable innovation that is non-infringing, and the difficulty of detecting infringement (One of the problems with software patents is that it is often difficult to detect whether infringement has occurred in a closed-source competing implementation, which makes the patent unenforceable in practice).
Most of their patents were never worth anything, even with all the vetting. The valuable ones were core innovations in emerging industries where the product gets sold in consumer quantities, not some way to make a better forklift.
Somebody explain to me how this is different from someone selling Avon, or selling at the local farmers' market, or moonlighting as a musician at the local dive bar, or any other similar wellspring of unemployment stupidity?
Well, the whole point behind a distro is that everything -- the kernel, a bunch of loadable kernel modules, the libraries, the userland, any X Window system you might choose to run, and various other stuff -- is assembled and tested by others, and some sort of effort is made to evaluate interdependencies and resolve them in a sensible way. It isn't a trivial job.
Sure, you could take an older Debian distro or one of the other distros, remove Linux, add FreeBSD, and you might get it to work. But I can guarantee that at least *some* of the packages will be broken. What Debian are doing is treating those problems as release-critical, so that they will fix them either on the kernel side or in the affected packages themselves before issuing the distro.
News Flash: 10,000 Slashdot accounts compromised in phishing scam. Most common passwords were 31415 and 0xdecafbad.
Affected users have been placed on an isolated network where they can't do anything but post whinges about Microsoft and Apple to a web server that runs SSL using a self-signed certificate and actually follows the RFCs.
I read TFA (more specifically the speech transcript) and I don't believe that the Rocky Mountain News would have been much better off even if they did everything right. There aren't any city-specific news web sites out there that are making anything like the kind of money that newspapers made in their heyday. Like the buggy whips, the telegraph industry, and home coal delivery the business is gone and the new industry that is replacing it is too far removed for a transition to be possible.
The glory disappeared when we got what we thought we wanted: compilers that worked, operating systems that don't crash, source level debuggers, enough memory, enough disk, fast networking, source code control that actually works, and the ability to ssh in reliably from home.
I think that the "glory," the absence of which we bemoan in this thread, is best understood as a metaphor for the era when success in IT and its predecessor fields was mainly about being smart. If you were in what was called the data processing business in 1960, intellect was, by and large, what made you successful. That situation persisted until the mid-1980s or so (depending on where you worked), and it gradually became more important to have knowledge than intellect, and the non-technical skills (writing, teamwork, getting along with your boss, dealing with politics) became more important, too. The change was because of the drastic increases in complexity of the systems we worked with, and because the tools were so much more reliable.
In 1980, it was not an unreasonable objective to read every word of every document printed, and every line of source code, for Bell Labs UNIX. Something that could easily be done in a few months. It's not a reasonable goal, anymore, for any of the major desktop releases, and so you have to specialize, and rely on having things just work. And by and large, they do, and even people who don't specialize in technology can use computers and write Excel macros these days. They for the most part do quite well unassisted, and so the panache that came with restoring the boss's spreadsheet from a floppy disk with a bad sector isn't there anymore.
There are still good gigs out there, but they can be hard to find, and you have to make your tradeoffs among technical challenge, funding continuity, salary, management quality, coworker quality, and the extent to which the technology is strategic from a career perspective. And once in a while you still get to work around a compiler bug.
... or aircraft control and navigation, or banking, or encryption, or much of anything besides consumer products where it's OK to fail once in a while. Different situations require different approaches.
I agree, but even in those cases complexity doesn't mean better or more reliable performance. Reliability and predictability is often easier to ensure in a simple solution where there are fewer potential failure modes. Better is truly the enemy of good enough in many cases.
Yes, but you probably don't want to skip the unit test, and having a carefully thought-out design might not be a bad idea.
... or aircraft control and navigation, or banking, or encryption, or much of anything besides consumer products where it's OK to fail once in a while. Different situations require different approaches.
Spolski is smart and I usually like what he writes, but it's important to remember that he spent his formative years at Microsoft.
I'm not planning on taking my Thinkpad X301 with the hardware-encrypted SSD over any borders for exactly this reason.
But if it were more commonplace, they would lose interest. Border patrol operate like cops setting up speed traps. They don't care how many smart people slip through, they care about finding the technique that nets them the largest number of arrests. If it becomes pointless, they'll change it at a policy level.
How long will it be until freedom-loving, consumer-supporting manufacturers start making devices that are resistant to searches like these? With today's technology there's no reason I shouldn't be able to have strong encryption of any nonvolatile storage and a means of locking down the device so that nothing is left in RAM or cache and the key is sequestered or destroyed (presumably pending manual reentry after the checkpoint is cleared). Fine, the law says they can conduct a forensic search, but there's no reason I have to make it easy for them.
Hopefully whatever comes of this will help out groups like IMSLP that are working on books and other media outside the text-centric Google mold. Orphaned copyright, and excessive copyright terms in general, are too large a problem to let an almost "good enough" solution like Google Books carry the day.
Any kind of security system that provides a limited lifetime or constrained redistribution rights for messages is, fundamentally, DRM. Therefore, it's subject to the same kinds of attacks that cause DRM to fail. Ultimately, unless you can build a trusted platform module with remote attestation that is tamper proof, there are gaps. This particular attack is, at a more abstract level, really about producing counterfeit trusted nodes. Without a TPM at each node and some way to authenticate independence through a trust hierarchy, there's no way for this to work.
Blind tests of violins and bows are notoriously difficult to conduct effectively. Much of the problem is that players become accustomed to particular instruments and unconsciously adjust their playing, and indeed their artistry, to the response of a particular instrument. Instruments have off days due to changes in humidity or string wear. The bow has to match the instrument and the performer. Differences among great violins are subtle. Selection of music to be played has a role. Performers, too, are variable, and rarely give three or more great performances of a work in a row.
Nonetheless, this is promising work. A modern violin by the best makers is typically a $25,000 instrument, while professional players in major orchestras are expected to spend several times that for an older instrument. It's like having an extra house payment. If the quality of the modern instruments starts to rival and surpass those of lesser makers in antiquity, it will help young players immensely as well as giving speculators in such instruments a well-deserved comeuppance.
There are tradeoffs involving proliferation risk. If you are willing to accept the presence in a commercial power reactor of fissile materials that could be converted to weapons use, then there's plenty of fuel.
Back then, we didn't have Windows. Now we do, and we can use Windows and Windows technologies to control our systems. Stuff like OPC (OLE for Process Control, yes, that OLE...).
And plant management can open up a nifty Excel worksheet, pulling out the numbers from the plant immediately...
</joke>
Analog control loops and things like PDP-8s weren't necessarily a whole lot more reliable than Windows. I don't know what the plant actually had for controls, but if you wanted a digital computer, the PDP-8 was fairly typical of the era. An analog meter with a d'Arsonval movement and optical sensors for the trip points was just tickety-boo in those days. Sometimes even the good ones stick.
In a sick sort of way, Chernobyl had more effective public relations -- the public belief was that the accident wasn't nearly as bad is it indeed was, while at TMI the public belief was that the accident was considerably worse than the facts showed.
A close reading of the accident narrative at TMI doesn't support the idea that the failure was "solved correctly." It shows that despite a grave combination of equipment failures, human factors problems, and bad judgment, the plant was eventually shut down without any leakage of radioactive material into the environment, due to a combination of conservative design and a little luck.
The main technical differences between TMI and Chernobyl were that a) Chernobyl had an intrinsically less safe graphite moderator while TMI had heavy water and b) TMI had better containment.
Slashdot Mirror
Computers sold in the Energy Star era (post-1992) all had some form of power stepping. In all fairness, the early forms were a nuisance and were often disabled. More recent computers, minimally, step down the processor clock when there's nothing to do. Typical desktop CPU power draw, full CPU utilization compared to idle but still running the OS, has varied, peaking several years ago and now dropping somewhat as peak CPU power draw has come down a little. You lose 5-10% in the power supply. 40 watts is perhaps a little low, as an average over the past 10 years.
There is also the additional air conditioning load, which is typically 30% for the fraction of time air conditioning is in use in the building.
There is also additional fan wear since the fans will run harder. And the additional heatsink clogging due to dust that a technician must remove. Quite possibly, the complaints of slow PCs were due to the CPU speed being stepped down due to heat limiting because of fouled heatsinks from dust combined with the CPU hog program keeping everything running hot.
There is also more rapid component failure due to heat. This affects the CPU itself, and depending on the case design, possibly other components as well.
Finally, with an IT mindset rooted in the importance of running SETI, it is unlikely that other power-saving strategies, like shutting off machines at night or on weekends, or allowing them to enter sleep or hibernate states, were seriously entertained, though to be sure such policies don't always make sense.
40 watts x 24 hours x 365 days x 10 years x 5000 machines x $.06 /kwh = $1,051,200
Only a school district or the government could have taken 10 years to find a CPU hog running on 5,000 computers.
From TFA: "Gilbert police are now involved in the investigation and criminal charges may be filed." How is this criminal? He had legitimate access. What's special about the scale? If someone ran a single instance of SETI@home on the PC on their desk, would that be criminal?
Small companies I've worked at have used a number of strategies to make this fair. One way to do it is limit service hours. For a while we limited service hours to 7 am to 7 pm, and if customers called outside that the expectation was that they'd get a call at 7 am the next day. I have also seen best-effort systems where there is no formal SLA, and several people get paged at once if there's trouble, which works in some cases.
The problem with being on call is that, if you have a life, you have commitments and do stuff that you can't just unwind in 10 minutes. Golf, fishing trip, reffing a kid's softball game, community orchestra rehearsal, picking up kids on the other side of town.
Chinese internet filtering is justified publicly by stating that it is done to help Chinese people avoid inadvertent violations of the law, and that is how it is seen by most Chinese. The real purpose of the censorship there is to facilitate prosecution of dissidents by making it impossible to violate laws against anti-government speech and unlawful assembly inadvertently.
I used to work for $BIG_COMPANY where we had a process where a team of lawyers and engineers would evaluate everything someone thought remotely patentable, and they would consider whether the patent was worth the time and money (usually several thousand dollars even though they used in-house counsel and did thousands of these a year). The analysis was based not only on the validity of the patent and the value of the innovation covered, but also the difficulty of creating a comparable innovation that is non-infringing, and the difficulty of detecting infringement (One of the problems with software patents is that it is often difficult to detect whether infringement has occurred in a closed-source competing implementation, which makes the patent unenforceable in practice).
Most of their patents were never worth anything, even with all the vetting. The valuable ones were core innovations in emerging industries where the product gets sold in consumer quantities, not some way to make a better forklift.
Somebody explain to me how this is different from someone selling Avon, or selling at the local farmers' market, or moonlighting as a musician at the local dive bar, or any other similar wellspring of unemployment stupidity?
Sure, you could take an older Debian distro or one of the other distros, remove Linux, add FreeBSD, and you might get it to work. But I can guarantee that at least *some* of the packages will be broken. What Debian are doing is treating those problems as release-critical, so that they will fix them either on the kernel side or in the affected packages themselves before issuing the distro.
Affected users have been placed on an isolated network where they can't do anything but post whinges about Microsoft and Apple to a web server that runs SSL using a self-signed certificate and actually follows the RFCs.
Sounds like a copyright violation to me. After all, it's not a parody, it's a remake.
I read TFA (more specifically the speech transcript) and I don't believe that the Rocky Mountain News would have been much better off even if they did everything right. There aren't any city-specific news web sites out there that are making anything like the kind of money that newspapers made in their heyday. Like the buggy whips, the telegraph industry, and home coal delivery the business is gone and the new industry that is replacing it is too far removed for a transition to be possible.
I think that the "glory," the absence of which we bemoan in this thread, is best understood as a metaphor for the era when success in IT and its predecessor fields was mainly about being smart. If you were in what was called the data processing business in 1960, intellect was, by and large, what made you successful. That situation persisted until the mid-1980s or so (depending on where you worked), and it gradually became more important to have knowledge than intellect, and the non-technical skills (writing, teamwork, getting along with your boss, dealing with politics) became more important, too. The change was because of the drastic increases in complexity of the systems we worked with, and because the tools were so much more reliable.
In 1980, it was not an unreasonable objective to read every word of every document printed, and every line of source code, for Bell Labs UNIX. Something that could easily be done in a few months. It's not a reasonable goal, anymore, for any of the major desktop releases, and so you have to specialize, and rely on having things just work. And by and large, they do, and even people who don't specialize in technology can use computers and write Excel macros these days. They for the most part do quite well unassisted, and so the panache that came with restoring the boss's spreadsheet from a floppy disk with a bad sector isn't there anymore.
There are still good gigs out there, but they can be hard to find, and you have to make your tradeoffs among technical challenge, funding continuity, salary, management quality, coworker quality, and the extent to which the technology is strategic from a career perspective. And once in a while you still get to work around a compiler bug.
... or aircraft control and navigation, or banking, or encryption, or much of anything besides consumer products where it's OK to fail once in a while. Different situations require different approaches.
I agree, but even in those cases complexity doesn't mean better or more reliable performance. Reliability and predictability is often easier to ensure in a simple solution where there are fewer potential failure modes. Better is truly the enemy of good enough in many cases.
Yes, but you probably don't want to skip the unit test, and having a carefully thought-out design might not be a bad idea.
Spolski is smart and I usually like what he writes, but it's important to remember that he spent his formative years at Microsoft.
But if it were more commonplace, they would lose interest. Border patrol operate like cops setting up speed traps. They don't care how many smart people slip through, they care about finding the technique that nets them the largest number of arrests. If it becomes pointless, they'll change it at a policy level.
How long will it be until freedom-loving, consumer-supporting manufacturers start making devices that are resistant to searches like these? With today's technology there's no reason I shouldn't be able to have strong encryption of any nonvolatile storage and a means of locking down the device so that nothing is left in RAM or cache and the key is sequestered or destroyed (presumably pending manual reentry after the checkpoint is cleared). Fine, the law says they can conduct a forensic search, but there's no reason I have to make it easy for them.
Hopefully whatever comes of this will help out groups like IMSLP that are working on books and other media outside the text-centric Google mold. Orphaned copyright, and excessive copyright terms in general, are too large a problem to let an almost "good enough" solution like Google Books carry the day.
Any kind of security system that provides a limited lifetime or constrained redistribution rights for messages is, fundamentally, DRM. Therefore, it's subject to the same kinds of attacks that cause DRM to fail. Ultimately, unless you can build a trusted platform module with remote attestation that is tamper proof, there are gaps. This particular attack is, at a more abstract level, really about producing counterfeit trusted nodes. Without a TPM at each node and some way to authenticate independence through a trust hierarchy, there's no way for this to work.
Nonetheless, this is promising work. A modern violin by the best makers is typically a $25,000 instrument, while professional players in major orchestras are expected to spend several times that for an older instrument. It's like having an extra house payment. If the quality of the modern instruments starts to rival and surpass those of lesser makers in antiquity, it will help young players immensely as well as giving speculators in such instruments a well-deserved comeuppance.
There actually are pleasant, smart, capable people out there in tech jobs and tech management.
There are tradeoffs involving proliferation risk. If you are willing to accept the presence in a commercial power reactor of fissile materials that could be converted to weapons use, then there's plenty of fuel.
And how is that different from the coal plants that nuclear power replaces?
Back then, we didn't have Windows. Now we do, and we can use Windows and Windows technologies to control our systems. Stuff like OPC (OLE for Process Control, yes, that OLE...).
And plant management can open up a nifty Excel worksheet, pulling out the numbers from the plant immediately...
</joke>
Analog control loops and things like PDP-8s weren't necessarily a whole lot more reliable than Windows. I don't know what the plant actually had for controls, but if you wanted a digital computer, the PDP-8 was fairly typical of the era. An analog meter with a d'Arsonval movement and optical sensors for the trip points was just tickety-boo in those days. Sometimes even the good ones stick.
A close reading of the accident narrative at TMI doesn't support the idea that the failure was "solved correctly." It shows that despite a grave combination of equipment failures, human factors problems, and bad judgment, the plant was eventually shut down without any leakage of radioactive material into the environment, due to a combination of conservative design and a little luck.
The main technical differences between TMI and Chernobyl were that a) Chernobyl had an intrinsically less safe graphite moderator while TMI had heavy water and b) TMI had better containment.