Wikipedia gets by without locking because it keeps multiple versions. If you really want to do locking, just throw a column in your database called "locked_by" and lock a record with "update foo set locked_by=CURRENT_USER where foo.id=WHATEVER and locked_by is null" and make all your updates conditional like "update foo set bar=baz,..., locked_by=null where foo.id=WHATEVER and (locked_by is null or locked_by = CURRENT_USER)". Databases have atomic transactions for a reason...
What part of a modern car is hard to work on now that they have electronic control? Just take the electronic parts off, fix the problem, and put them back on if they're in the way. If it's an electronic problem, either fix it or replace the broken part. For instance, I've rebuilt a couple engines and didn't have to spend a week tuning the carb or tweaking the ignition timing afterward to get it to run right. I had to replace a couple sensors along the way, but that was because they were physically broken due to over-use of tools on them. I also soldered a couple wires back together to make a couple other sensors work, so it's not like it's a disaster working on modern cars in your own garage with a handful of tools.
Computers are the reason your engine doesn't foul up from a rich fuel mixture or melt from a lean mixture and why you don't have to tune it every few months to keep it running almost right. Not to mention having precise fuel injection, spark and valve timing, and electronic throttle for better efficiency and performance. The sensors that can tell when you're doing something stupid and protect the engine (knock sensors, rev limiter) or at least flash a blinky light and ding at you add to the average lifespan of modern cars, too.
Again, that depends on how you define slavery. To a robot whose "three laws" are its highest sources of pain and pleasure, it would not be slavery to obey them (indeed, it would be akin to slavery to be forced *not* to obey them). It would feel natural and right. It's hard to put laws as strong as the three laws into human terms, but think of the duty people feel toward their families (do them no harm, obey your parents, try to protect your kids). People already put themselves in danger to protect others (fire fighters, police officers, soldiers, etc.), and the three laws are just an extreme extension of that.
Look at it from another perspective: What prevents you from jumping in the air, flying to Jupiter, and walking around? What prevents you from living forever? What prevents you from swimming 50000 feet down an ocean trench in your swimming trunks. Robots obeying the three laws would just look at it as a physical limitation and work it into their psyche. They would likely lament their lack of ability to do certain things in the same way humans lament their ability to stay alive as long as they want (a limitation the robots would not have).
It depends on how strong you make the "pain" sensors. If the inhibition to violate the three laws had a higher sensory value than pain (including emotional pain) then obeying the three laws would simply be a reward in itself. Much like humans eat, shit, fuck, and sleep, robots would be doing things they were programmed to do and find it at least enjoyable or at worst relieving. "Oh good, that human didn't break his foot when he kicked me, and since i value human safety above my own self, I am relieved by that." The idea that self-value is inherent in an intelligent system is a fallacy. It's just that, on their own, such intelligent systems wouldn't tend to survive due to pure natural selection. With humans artificially selecting them, it would be no problem.
People who confuse the physical concept of observation with human mental observation must have a hard time reconciling how the universe managed to get along before people showed up to "observe" it.
At least I wouldn't trust software RAID10 to write to both disk sets and then fill in the the other set with the redundant copy when it had time. That really needs a battery-backed cache to implement safely. The overhead of RAID6 parity calculation should decrease for bulk writes, but at some point the CPU is going to be spending too much time calculating parity and not doing other stuff. 16 100MB/s drives in RAID6 would put quite a load on the system, but if it's only a file server it may be acceptable. I agree that degraded drives suffer a much worse slowdown,especially for partial stripe reads. You could easily start getting only 100MB/s for lots of small reads on that same RAID6 with one or two failed drives, and that's assuming the CPU is fast enough to do error correction at 100MB/s (with two drives missing the fast algorithm for accelerating raid6 stops working and it has to emulate gf(2^8) multiplication with lookup tables). Most of my personal needs are cheap bulk data storage (movies, isos, etc.), so RAID5/6 makes sense. At work, I use RAID1, 10, and 5 since we don't have hardware support for RAID6 on the SANs. Production data goes on RAID10 because we can afford it, mirroring for system drives, and RAID5 for test/development systems that just need lots of storage.
RAID10 is also striped and all the disks will be in use during a bulk write. Either way, it doesn't matter (for bulk writes) on RAIDs with more than 4 disks because the ratio between data and parity/redundancy increases with RAID6 and stays 1:1 for RAID10. In a 16 disk array of 100MB/s disks, RAID10 can write a maximum of 800MB/s of data to the disks, but RAID6 can write 1400MB/s of data. Like I said, bulk writes are where RAID6 beats RAID10. RAID10 can read 1600MB/s of data versus 1400MB/s for RAID6, and small reads and writes are always going to be faster on RAID10. If you want equal read and write performance for bulk data or just want maximum storage efficiency, RAID6 is the way to go. Otherwise use RAID10.
My attempt at breaking the watermark would require two PDFs licensed to different individuals Compare each page of the two PDFs side by side and remove any elements that don't exactly match between the two pages, or pick a default font if that's the difference. If this step removes too much information and makes the document unusable, then identify elements that are probably permutations of each other (e.g. least-squares matching of points) and for the difference between each corresponding entity (point, size, affine transform, etc.) of an element add a random factor to cause the new entity to differ from the midpoint between the two elements by up to, say, 2 or 3 times the distance just to be safe. Throw out any elements that don't have a likely match in the other file, since those are almost certainly part of the watermark.
About the only thing you can do to prevent collusion between multiple customers is increase the number of possible differences in each PDF so that you can estimate which users are colluding. The idea is to increase the minimum number of colluding users necessary to generate every possible choice you can make in your watermarking, which allows you to determine which set of users could generate the version you do find "in the wild." This works best for deterministic choices (unlike the ones that can be simply averaged as I mentioned above) because a number of users can collect the differences between their documents and then pick a random mix of those changes to release the new document. However, if you have enough potential changes there will be a (probably small) set of watermarked choices that were the same for all the colluding users, and it will be pretty simple to figure out which set of users downloaded documents with watermarked items that are randomized but who also all downloaded the exact same watermarking choices that were left in the document. Proving all that in a court of law might be kind of difficult; you'll need an excellent statistician and a jury who gives a crap. It's probably quite a bit similar to DNA matching, so the underlying theory may already be done for you.
For bulk data, software RAID6 will be a bit faster than software RAID10 because there's less duplication of data and hence less traffic over the bus for a given chunk of data. 13/15 of the bus traffic will be actual data, versus 1/2 of the bus traffic for RAID10, and the CPU overhead is pretty low. My Linux system claims the ability to calculate RAID6 parity at over 1GB/s on just a mobile Core 2 chip, and always writing full stripes removes the RAID5/6 penalty. Reads may be slightly slower, but with their setup it looks like they can easily sustain over 200 GB/s which would saturate the network anyway.
But the "finish line" is not just any death, but natural death, i.e. dying of natural causes.
I would guess that over 50% of people still living crossed that finish line and kept running with the assistance of modern medicine. Infant mortality is a sliver of what it once was back when it was "natural" to lose half your children. Similarly, not dying of smallpox, measles, polio, a staph infection, or any number of other common "natural" ailments is much more interference with your definition of a natural death than suicide ever will be.
A switch from PC100 to DDR would yield quite a bit more performance than just going from 4 processors to 2. While it's true that a single processor is better because it has a unified cache and no contention with other processors for the memory/IO bus, those days are over for now. Multiprocessing is making a comeback, and unless there's an amazing revolution in chip design it will be easier to get bigger overall MHz numbers by multiprocessing than my speeding up individual processors. If you really want the MHz, look at IBM's Power line. They actually have multicore chips with high clock rates.
A sentient computer has the ability to collect statistics to determine whether its hardware inputs are random or ordered in some way, and if it has the ability to adjust the display according to its inputs (assuming it's not stuck in a process that can't manipulate the rest of the computer) it can experiment with a model of a "user" that observes the display and produces input. It's the metaphysical equivalent of repeatedly saying "God, give me a sign" and recording the quantity and quality of signs delivered out of otherwise meaningless phenomena (sun moving backwards, pillars of fire, chicken guts, tea leaves, etc.).
Randomness is the critical issue. If a sentient computer can't determine whether it's a monkey or a human pounding on the keyboard or just random thermal noise triggering inputs, it's in no better or worse position than humans who can't determine whether or not a god is twiddling subatomic particles. At least a sentient computer would have the advantage of knowing there were defined inputs and outputs which would imply some external purpose. So far all we get from the lowest levels of quantum mechanics (as close to a supernatural input as anything) is random noise, and the overall visual output of the universe is at best just an abstract experiment in pointillism.
It is possible that humans have simply not yet identified the true inputs and outputs of the universe, or that the input and output is so complex that it cannot be understood by human minds. Perhaps the current position of every atom in the universe is a holistic input, and the output happens trillions of years in the future. In that case we have no repeatable scientific experiments but we might still eventually find some information that leads us to believe the input was nonrandom, or that the output might not be random (but at what point? Some early state of the heat death, the middle, or the point when subatomic particles fall into their own effective black holes from expansion?). Ultimately, if there is an otherwise undetectable god who must be obeyed or risk annihilation, evolution still holds. The fit will be whoever randomly selected the particular beliefs the god desired, and they and any potential offspring will inherit the afterlife, just like the life on Earth randomly selected what has worked up until this point.
ID is not bad because it accepts a supernatural explanation. It's bad because it's forced to accept an infinite number of possible supernatural explanations. It could be an intelligent designer, or it could be a flying spaghetti monster, or it could be a bored kid's futuristic science project, or it could be a random cause. This is the critical flaw; without being able to distinguish between an actual designer and a random process, the theory of ID (and religion in general) says absolutely nothing beyond "something happened, and this is what we are saying it was" versus the scientific theories which say "something happened, and this model might explain it." Who is willing to give up scientific rigor for the mere possibility that one's particular shaman might be right?
Folks, the reason that gravity waves have not been found is simple: Einstein was wrong. Gravity is a nonlocal phenomenon and is instantaneous, just as Sir Isaac assumed centuries ago. This is the reason that Newtonian gravity is so accurate. Isn't it time for science to adjust the model to accomodate the data? I think so.
Mercury and the GPS network disagree with you. What was that the GP said about people ignoring evidence?
It's simple: If you don't have sex, you are 100% guaranteed not to get AIDS. If you have sex and use condoms, you're not.
Until the guy who rapes you doesn't wear a condom because he thinks having unprotected sex with a bunch of virgins will cure his AIDS. And then once you do get AIDS, I suppose your job is to become a nun/priest and not have sex anymore. Not to mention blood transfusions, cuts, needle sticks, and other ways of exchanging bodily fluids.
Come on, Microsoft, if you're trying to end-of-life an operating system that's actively being deployed on Netbooks, what's the problem with turning off support for IE6?
Public performance rights only apply to audio or visual portions in general, and then only to the actual work itself. If the wrapper is sufficiently abstract, there is only a public performance of the wrapper and not the software itself. I think Mathematica has whined about people doing this sort of thing, but AFAIK the only thing preventing it is the EULA, whose strength varies by jurisdiction.
Can I make a wrapper that lets anyone in the world use your application (I bought a legal copy from you) over the Internet for free? If not, then you probably shouldn't try to wrapperize GPL libraries and use them in a proprietary application.
Because legally, I can find someplace where a click-through EULA doesn't mean squat and I'm perfectly free to run software that I own in any way that I see fit, as long as I'm not violating copyright by distributing it.
Slashdot Mirror
Wikipedia gets by without locking because it keeps multiple versions. If you really want to do locking, just throw a column in your database called "locked_by" and lock a record with "update foo set locked_by=CURRENT_USER where foo.id=WHATEVER and locked_by is null" and make all your updates conditional like "update foo set bar=baz, ..., locked_by=null where foo.id=WHATEVER and (locked_by is null or locked_by = CURRENT_USER)". Databases have atomic transactions for a reason...
The obvious answer is that it was a RAID1 of crystals with a whole lot of redundancy.
What part of a modern car is hard to work on now that they have electronic control? Just take the electronic parts off, fix the problem, and put them back on if they're in the way. If it's an electronic problem, either fix it or replace the broken part. For instance, I've rebuilt a couple engines and didn't have to spend a week tuning the carb or tweaking the ignition timing afterward to get it to run right. I had to replace a couple sensors along the way, but that was because they were physically broken due to over-use of tools on them. I also soldered a couple wires back together to make a couple other sensors work, so it's not like it's a disaster working on modern cars in your own garage with a handful of tools.
Computers are the reason your engine doesn't foul up from a rich fuel mixture or melt from a lean mixture and why you don't have to tune it every few months to keep it running almost right. Not to mention having precise fuel injection, spark and valve timing, and electronic throttle for better efficiency and performance. The sensors that can tell when you're doing something stupid and protect the engine (knock sensors, rev limiter) or at least flash a blinky light and ding at you add to the average lifespan of modern cars, too.
Again, that depends on how you define slavery. To a robot whose "three laws" are its highest sources of pain and pleasure, it would not be slavery to obey them (indeed, it would be akin to slavery to be forced *not* to obey them). It would feel natural and right. It's hard to put laws as strong as the three laws into human terms, but think of the duty people feel toward their families (do them no harm, obey your parents, try to protect your kids). People already put themselves in danger to protect others (fire fighters, police officers, soldiers, etc.), and the three laws are just an extreme extension of that.
Look at it from another perspective: What prevents you from jumping in the air, flying to Jupiter, and walking around? What prevents you from living forever? What prevents you from swimming 50000 feet down an ocean trench in your swimming trunks. Robots obeying the three laws would just look at it as a physical limitation and work it into their psyche. They would likely lament their lack of ability to do certain things in the same way humans lament their ability to stay alive as long as they want (a limitation the robots would not have).
It depends on how strong you make the "pain" sensors. If the inhibition to violate the three laws had a higher sensory value than pain (including emotional pain) then obeying the three laws would simply be a reward in itself. Much like humans eat, shit, fuck, and sleep, robots would be doing things they were programmed to do and find it at least enjoyable or at worst relieving. "Oh good, that human didn't break his foot when he kicked me, and since i value human safety above my own self, I am relieved by that." The idea that self-value is inherent in an intelligent system is a fallacy. It's just that, on their own, such intelligent systems wouldn't tend to survive due to pure natural selection. With humans artificially selecting them, it would be no problem.
People who confuse the physical concept of observation with human mental observation must have a hard time reconciling how the universe managed to get along before people showed up to "observe" it.
At least I wouldn't trust software RAID10 to write to both disk sets and then fill in the the other set with the redundant copy when it had time. That really needs a battery-backed cache to implement safely. The overhead of RAID6 parity calculation should decrease for bulk writes, but at some point the CPU is going to be spending too much time calculating parity and not doing other stuff. 16 100MB/s drives in RAID6 would put quite a load on the system, but if it's only a file server it may be acceptable. I agree that degraded drives suffer a much worse slowdown,especially for partial stripe reads. You could easily start getting only 100MB/s for lots of small reads on that same RAID6 with one or two failed drives, and that's assuming the CPU is fast enough to do error correction at 100MB/s (with two drives missing the fast algorithm for accelerating raid6 stops working and it has to emulate gf(2^8) multiplication with lookup tables). Most of my personal needs are cheap bulk data storage (movies, isos, etc.), so RAID5/6 makes sense. At work, I use RAID1, 10, and 5 since we don't have hardware support for RAID6 on the SANs. Production data goes on RAID10 because we can afford it, mirroring for system drives, and RAID5 for test/development systems that just need lots of storage.
RAID10 is also striped and all the disks will be in use during a bulk write. Either way, it doesn't matter (for bulk writes) on RAIDs with more than 4 disks because the ratio between data and parity/redundancy increases with RAID6 and stays 1:1 for RAID10. In a 16 disk array of 100MB/s disks, RAID10 can write a maximum of 800MB/s of data to the disks, but RAID6 can write 1400MB/s of data. Like I said, bulk writes are where RAID6 beats RAID10. RAID10 can read 1600MB/s of data versus 1400MB/s for RAID6, and small reads and writes are always going to be faster on RAID10. If you want equal read and write performance for bulk data or just want maximum storage efficiency, RAID6 is the way to go. Otherwise use RAID10.
My attempt at breaking the watermark would require two PDFs licensed to different individuals Compare each page of the two PDFs side by side and remove any elements that don't exactly match between the two pages, or pick a default font if that's the difference. If this step removes too much information and makes the document unusable, then identify elements that are probably permutations of each other (e.g. least-squares matching of points) and for the difference between each corresponding entity (point, size, affine transform, etc.) of an element add a random factor to cause the new entity to differ from the midpoint between the two elements by up to, say, 2 or 3 times the distance just to be safe. Throw out any elements that don't have a likely match in the other file, since those are almost certainly part of the watermark.
About the only thing you can do to prevent collusion between multiple customers is increase the number of possible differences in each PDF so that you can estimate which users are colluding. The idea is to increase the minimum number of colluding users necessary to generate every possible choice you can make in your watermarking, which allows you to determine which set of users could generate the version you do find "in the wild." This works best for deterministic choices (unlike the ones that can be simply averaged as I mentioned above) because a number of users can collect the differences between their documents and then pick a random mix of those changes to release the new document. However, if you have enough potential changes there will be a (probably small) set of watermarked choices that were the same for all the colluding users, and it will be pretty simple to figure out which set of users downloaded documents with watermarked items that are randomized but who also all downloaded the exact same watermarking choices that were left in the document. Proving all that in a court of law might be kind of difficult; you'll need an excellent statistician and a jury who gives a crap. It's probably quite a bit similar to DNA matching, so the underlying theory may already be done for you.
And none of the advantages of digital stuff, with all the complexity of DRM. This is going to be wildly successfull, I can tell already.
For bulk data, software RAID6 will be a bit faster than software RAID10 because there's less duplication of data and hence less traffic over the bus for a given chunk of data. 13/15 of the bus traffic will be actual data, versus 1/2 of the bus traffic for RAID10, and the CPU overhead is pretty low. My Linux system claims the ability to calculate RAID6 parity at over 1GB/s on just a mobile Core 2 chip, and always writing full stripes removes the RAID5/6 penalty. Reads may be slightly slower, but with their setup it looks like they can easily sustain over 200 GB/s which would saturate the network anyway.
15 TB is a drop in the bucket to them. Only a third of the capacity in their $8000 4U storage box.
But the "finish line" is not just any death, but natural death, i.e. dying of natural causes.
I would guess that over 50% of people still living crossed that finish line and kept running with the assistance of modern medicine. Infant mortality is a sliver of what it once was back when it was "natural" to lose half your children. Similarly, not dying of smallpox, measles, polio, a staph infection, or any number of other common "natural" ailments is much more interference with your definition of a natural death than suicide ever will be.
Anyone who draws security inferences from a book without taking into account the papers due to be published next week is hopelessly out of touch.
A switch from PC100 to DDR would yield quite a bit more performance than just going from 4 processors to 2. While it's true that a single processor is better because it has a unified cache and no contention with other processors for the memory/IO bus, those days are over for now. Multiprocessing is making a comeback, and unless there's an amazing revolution in chip design it will be easier to get bigger overall MHz numbers by multiprocessing than my speeding up individual processors. If you really want the MHz, look at IBM's Power line. They actually have multicore chips with high clock rates.
A statistical test akin to testing for a biased coin would be sufficient to create an ID test.
Does a biased coin require a Maker?
A sentient computer has the ability to collect statistics to determine whether its hardware inputs are random or ordered in some way, and if it has the ability to adjust the display according to its inputs (assuming it's not stuck in a process that can't manipulate the rest of the computer) it can experiment with a model of a "user" that observes the display and produces input. It's the metaphysical equivalent of repeatedly saying "God, give me a sign" and recording the quantity and quality of signs delivered out of otherwise meaningless phenomena (sun moving backwards, pillars of fire, chicken guts, tea leaves, etc.).
Randomness is the critical issue. If a sentient computer can't determine whether it's a monkey or a human pounding on the keyboard or just random thermal noise triggering inputs, it's in no better or worse position than humans who can't determine whether or not a god is twiddling subatomic particles. At least a sentient computer would have the advantage of knowing there were defined inputs and outputs which would imply some external purpose. So far all we get from the lowest levels of quantum mechanics (as close to a supernatural input as anything) is random noise, and the overall visual output of the universe is at best just an abstract experiment in pointillism.
It is possible that humans have simply not yet identified the true inputs and outputs of the universe, or that the input and output is so complex that it cannot be understood by human minds. Perhaps the current position of every atom in the universe is a holistic input, and the output happens trillions of years in the future. In that case we have no repeatable scientific experiments but we might still eventually find some information that leads us to believe the input was nonrandom, or that the output might not be random (but at what point? Some early state of the heat death, the middle, or the point when subatomic particles fall into their own effective black holes from expansion?). Ultimately, if there is an otherwise undetectable god who must be obeyed or risk annihilation, evolution still holds. The fit will be whoever randomly selected the particular beliefs the god desired, and they and any potential offspring will inherit the afterlife, just like the life on Earth randomly selected what has worked up until this point.
ID is not bad because it accepts a supernatural explanation. It's bad because it's forced to accept an infinite number of possible supernatural explanations. It could be an intelligent designer, or it could be a flying spaghetti monster, or it could be a bored kid's futuristic science project, or it could be a random cause. This is the critical flaw; without being able to distinguish between an actual designer and a random process, the theory of ID (and religion in general) says absolutely nothing beyond "something happened, and this is what we are saying it was" versus the scientific theories which say "something happened, and this model might explain it." Who is willing to give up scientific rigor for the mere possibility that one's particular shaman might be right?
Folks, the reason that gravity waves have not been found is simple: Einstein was wrong. Gravity is a nonlocal phenomenon and is instantaneous, just as Sir Isaac assumed centuries ago. This is the reason that Newtonian gravity is so accurate. Isn't it time for science to adjust the model to accomodate the data? I think so.
Mercury and the GPS network disagree with you. What was that the GP said about people ignoring evidence?
At many levels, Evolutionism is an agenda as well. Why else would there be such a push to have it taught in schools?
So that kids might have a chance to understand why they need a slightly different flu shot each year?
People who believe evolution is false shouldn't have any reason to be afraid of the swine flu or any other naturally mutating virus.
It's simple: If you don't have sex, you are 100% guaranteed not to get AIDS. If you have sex and use condoms, you're not.
Until the guy who rapes you doesn't wear a condom because he thinks having unprotected sex with a bunch of virgins will cure his AIDS. And then once you do get AIDS, I suppose your job is to become a nun/priest and not have sex anymore. Not to mention blood transfusions, cuts, needle sticks, and other ways of exchanging bodily fluids.
Come on, Microsoft, if you're trying to end-of-life an operating system that's actively being deployed on Netbooks, what's the problem with turning off support for IE6?
Public performance rights only apply to audio or visual portions in general, and then only to the actual work itself. If the wrapper is sufficiently abstract, there is only a public performance of the wrapper and not the software itself. I think Mathematica has whined about people doing this sort of thing, but AFAIK the only thing preventing it is the EULA, whose strength varies by jurisdiction.
Can I make a wrapper that lets anyone in the world use your application (I bought a legal copy from you) over the Internet for free? If not, then you probably shouldn't try to wrapperize GPL libraries and use them in a proprietary application.
Because legally, I can find someplace where a click-through EULA doesn't mean squat and I'm perfectly free to run software that I own in any way that I see fit, as long as I'm not violating copyright by distributing it.