I didn't say "trying to compress", although perhaps you should have. It would be beyond smart, actually, if it *did* compress data that is incompressible; however, by definition... Oh, if you haven't worked out by now that I was being ironic, there's no hope for you.
Actually, that wild guess of $250/client storage cost target can be raised if you can be smart about which clients you keep backups for. Is there some feel for which clients are more likely to require their data to be restored later? If you can make a confident guess that a particular (identifiable) half of your clients are 1% likely to need you to recover the data, 30% are at 3%, and the remaining 20% are 40% likely, you can tune your backups to this. Make two backup sets for the 40% clients, maybe backup the 3% clients, and toss the data for the 1%ers. Of course, if you can't categorize the data sets in this way, you're stuck at finding an across-the-board solution at $250 per data set (or whatever your number is), or tossing all the data out.
You might also be able to categorize by time: if the likelihood of having to ever re-produce/restore data drops off over time, you can store data until the chances of ever needing it again fall below a certain level, and then re-use the media for new jobs.
Actually, it's not whether storage costs are more than that. It's whether storage costs multiplied by the chance of having to restore the set cost more than that. IOW, if it costs $5k to generate the data, but you only ever need to refer to old data in 5% of cases, your actual storage cost target is $250. If it cost $300 per client, you would be paying $6000 to store 20 clients' data, of which you only need to refer to one. So, $6000 per restored data set. Don't forget to include man-hours creating, restoring, and managing the backups.
This brings up a good point which I haven't seen addressed before now. How badly do you really need to keep the archives, and how reliable do you need them to be? If a backup fails, it will cost you $5-6k of CPU time. However, if you're archiving your (models/equations/whatever) together with your output, and you lose this as well, it may also cost you additional human time, which may be more expensive again.
That aside, most large organisations I've worked for (financial and scientific spaces) use tapes (mostly LTO-family, from memory), due to their high reliability over time and low cost/mb. Given the above, I would look at storing a one-off archive (3-4 tapes?) of your modeled data, and risk the outside chance that you happen to have to restore from a dead tape (my guess of 5% of data-sets need to be restored x chance of tape failure: estimate 2% = a very conservative 0.1% chance of having to regenerate data). This is very much a back-of-a-beermat figure, so run your own numbers through that.
More importantly, store your modeling configurations in something more reliable. It's much less costly to reliably store a few mb of modeling configuration, and if you do run into the 0.1% scenario, you can at least restore the modeling configuration and avoid having to re-do your research/analysis/whatever. Perhaps keep it online on a mirrored drive, and have multiple off-site tapes. If where you're at is anything like where I'm at, your scientists sometimes tweak the model code, so you may want to consider backing up their source code (or at least the compiled modeling code) with each configuration. Nothing like restoring the old model config only to realise that they've tweaked the models a dozen times since then, and have no idea which version was current at the time.
Perhaps if it was some kind of super-medical-record with the potential to save the human race from almost certain extinction. For anything else, that's WAY overkill. For instance, credit card providers (who are subject to some pretty stringent requirements) are only required to maintain one off-site backup and one off-site on-line redundant data set.
If you were storing my medical records that way, I'd be incredibly confident that they won't ever become unavailable, and incredibly nervous that, with my medical records being transferred about and stored in so many places, my privacy would be compromised. Even with the best data-protection and encryption systems in place, you're only compounding the risks of data leaking, while not appreciably decreasing the risk of data loss (it was vanishingly low long before you got to 5 backups and 5 backups of the backups.)
Either way, I'm sure they are happy with their site count today.
Except possibly for the fact that the intersection of "People who purchase women's sporting gear" and "Slashdot readers" is going to be pretty close to a null set.
"People who purchase women's sporting gear for wearing in public and "Slashdot readers" is going to be pretty close to a null set."
This was established in Australia decades ago too - this is not news. I don't care enough to check my reference, but I was fairly involved in it at the time, and from memory it was "Infopac International V RP Data, (late 80s or early 90s)". The judge ruled that the creative presentation of the data purchased at great expense by RP Data (the database and presentation software) was copyrightable, but that the data itself, being factual (real estate sales data), was not copyrightable. If Infopac International could access it legally (it could), there was nothing RP Data could do to restrict its use of the data.
Some highlights...
"No, it's not composing. It's still music." (What is music?)
"This is both music and composition." (What is music?)
"It's still music; it's not composing." (What is music?)
I was just trying to bring it back into context with the original post. That's what we're discussing, right? On the comments section for the post?
If I pick up my violin and play Mozart, then that's not music because I didn't write the notes?...
No, it's not composing. It's still music. If you play to a room full of people and claim you wrote it yourself, it's plagiarism.
What about a compositum, where I've taken a cool backbeat from another song, and improvise over that? According to your definition, what I'd be playing would be art, but the backbeat that (perhaps) provides the inspiration was originally music when played by the original creator, but now that I'm re-using it, it isn't music? Is the whole of it music, or just the part?
This is both music and composition. If you try to pass it off as your own work without mentioning the poor starving artist (or vastly rich dripping-with-gold-chains artist, or whatever) who wrote the backbeat and inspired your creation, it's plagiarism, unless you apply your creativity to the back-beat as well, rather than just lifting it verbatim.
What about a boom-box? If I have a recording of Itzhak Perlman playing that same Mozart piece, is it music if I play it out loud? Or was it only music the first time he played it?
This isn't about music; it's about creation. It's still music; it's not composing. The piece is about an author who 'performed' someone else's writing in text, and didn't tell people it was only a performance, not an original creation.
I'm terribly sorry, but you've done nothing but muddy the waters. Perhaps you could make a more precise definition that I could work with? Also, you seem very angry -- that's a strange self defeating choice as a method to appreciate art or music.
It's a shame you decided to stir the mud in with your own shit.
I'm not sure what your point is; for or against the author in question?
It's a little more modern than your examples, but imagine if Lizst's "Fantasies on a theme by Paganini" had just been excerpts of a violin playing the orininal piece, verbatim, bridged by sections of blending piano solo? That's not culture. What made it culture was when Lizst took, not the notation verbatim, but the idea, and built something entirely new but recognizably derived. Classical (and modern) music didn't excerpt: it used themes as a basis, not as a substitute for talent.
Music doesn't just 'sample and include' to create something new. It takes themes and creates on top of them. Imagine if any of the big 'derived' music styles had just put in whole excerpts from other styles, and added some bridging sections? They wouldn't have been the new sound they were. While I think it's okay to use elements of something someone else created in something new you create, there are three things you need to do:
Create. You don't just 'cut and paste'; you take the old and add your own touch to it, lending a creative element to the copying.
Attribute. Van Halen was only exposed to new audiences from being sampled because the audience could easily find out who was being sampled. If not for that, the audience would just think they were listening to something new.
Compensate. If you make a whole pile of money, some of it needs to go to people who's effort made your profit possible. For an idea barely lifted and heavily changed, maybe you just tell your fans to check out the other creator. If you took substantial chunks from the other work, prepare to hand over substantial chunks of profit (or prepare for a lawsuit!)
Also, while a mutt can be an excellent dog, you're not allowed to sell it claiming it to be a pure-bred. And the analogy is stupid anyway.
This is exactly right. Steal my mag-stripe card, and you can copy it, but can't use it. Steal my EMV card, and you can't copy it, but you can take all my money.
The/. Effect: Thousands of users simultaneously accessing a site to not read its content.
How does that work? Link a page on/. and it's crippled by the extra load, but virtually none of the observable population of/. shows any signs of having even managed to work their browser well enough to click the link. It feels like there should be some analogy with Quantum Uncertainty.
Ah, it was meant to be much more than that. It was meant to make it easier for the banks to blame the consumer for fraud, and avoid liability themselves. It has achieved this. You're right as regards actual security.
It's a cost-benefit analysis, and "pushing $x worth of liability onto someone else" works out to be not too much worse than "preventing $y worth of fraud", if $x = $y. Now if $x > $y, suddenly they're better off moving the liability than preventing the fraud. Similarly, if the first option costs more than the second, even if $x = $y, they're better off just moving liability.
This is why Britain has a historical problem of card fraud, and the US has a much better record. The US never let the banks push liability onto the consumer, regardless of the technology. This removes a whole segment of the CBA, meaning it's purely "cost of security vs. cost of fraud it would prevent".
This makes things worse. Before, the evidence trail was clear and understandable, and the consumer was well-protected. Signatures don't match? Consumer not liable.
Now, the system is still very breakable, but the evidence is complicated and often makes it look like the consumer was at fault. Consumer liable for transactions they never authorized.
The PIN storage or retry delay is not the issue. Cards ALREADY block access to the PIN after several failed attempts. The problem is that the bank is not able to detect that the card was never presented the PIN in the first place. The terminal thinks the PIN was verified, but it never gets passed on to the card.
The problem is that the MitM can spoof the "correct PIN was entered" response, and this isn't a part of the transaction MAC (basically an encrypted summary of what happened). If the card used the PIN verification result in the MAC the card issuer could detect this attack trivially. Unfortunately, this is an implementation detail which each bank would have to manage separately. New cards need to be rolled out.
And the consumer still has no idea if his card is safe or not.
Currently, people in the market for materials like child porn go to sites which carry it. It's not really that hard to find (so they say, to justify the blacklist.) Police get warrants to monitor sites as they're found (judicial oversight!), then pounce and lock up both the producers and the people who create the market.
Now, create a fence around the web, making sure people can't get to these places at all. Nary an immoral thought to be found on the web! (Who decides? No oversight.) What's a child-pornographer to do? They won't just stop: they'll look for alternate means to distribute. So, it goes further underground (already, some of these bad industries are using VPNs to keep traffic better-hidden and encrypted.) No more warrants (because it's harder to intercept, and harder to prove it SHOULD be intercepted), and it's not just the technically-capable and paranoid who avoid getting caught: suddenly, you're creating a strong drive for ALL of these bad guys to use anti-detection technologies like this. No more low-hanging fruit, and more kids who never get rescued. Sure, it's harder to find: but it's still out there.
It's kind of like the gun debate: you only keep the guns out of the hands of the honest ones. In the gun debate, that's a fallacy: honest people still decide to carry guns, and they still sometimes end up shooting people. Honest people don't decide to casual-carry child porn: the only market is the baddies anyway. And they'll get it anyway. And now, we won't catch them.
Slashdot Mirror
I didn't say "trying to compress", although perhaps you should have. It would be beyond smart, actually, if it *did* compress data that is incompressible; however, by definition... Oh, if you haven't worked out by now that I was being ironic, there's no hope for you.
Actually, that wild guess of $250/client storage cost target can be raised if you can be smart about which clients you keep backups for. Is there some feel for which clients are more likely to require their data to be restored later? If you can make a confident guess that a particular (identifiable) half of your clients are 1% likely to need you to recover the data, 30% are at 3%, and the remaining 20% are 40% likely, you can tune your backups to this. Make two backup sets for the 40% clients, maybe backup the 3% clients, and toss the data for the 1%ers. Of course, if you can't categorize the data sets in this way, you're stuck at finding an across-the-board solution at $250 per data set (or whatever your number is), or tossing all the data out.
You might also be able to categorize by time: if the likelihood of having to ever re-produce/restore data drops off over time, you can store data until the chances of ever needing it again fall below a certain level, and then re-use the media for new jobs.
Actually, it's not whether storage costs are more than that. It's whether storage costs multiplied by the chance of having to restore the set cost more than that. IOW, if it costs $5k to generate the data, but you only ever need to refer to old data in 5% of cases, your actual storage cost target is $250. If it cost $300 per client, you would be paying $6000 to store 20 clients' data, of which you only need to refer to one. So, $6000 per restored data set. Don't forget to include man-hours creating, restoring, and managing the backups.
This brings up a good point which I haven't seen addressed before now. How badly do you really need to keep the archives, and how reliable do you need them to be? If a backup fails, it will cost you $5-6k of CPU time. However, if you're archiving your (models/equations/whatever) together with your output, and you lose this as well, it may also cost you additional human time, which may be more expensive again.
That aside, most large organisations I've worked for (financial and scientific spaces) use tapes (mostly LTO-family, from memory), due to their high reliability over time and low cost/mb. Given the above, I would look at storing a one-off archive (3-4 tapes?) of your modeled data, and risk the outside chance that you happen to have to restore from a dead tape (my guess of 5% of data-sets need to be restored x chance of tape failure: estimate 2% = a very conservative 0.1% chance of having to regenerate data). This is very much a back-of-a-beermat figure, so run your own numbers through that.
More importantly, store your modeling configurations in something more reliable. It's much less costly to reliably store a few mb of modeling configuration, and if you do run into the 0.1% scenario, you can at least restore the modeling configuration and avoid having to re-do your research/analysis/whatever. Perhaps keep it online on a mirrored drive, and have multiple off-site tapes. If where you're at is anything like where I'm at, your scientists sometimes tweak the model code, so you may want to consider backing up their source code (or at least the compiled modeling code) with each configuration. Nothing like restoring the old model config only to realise that they've tweaked the models a dozen times since then, and have no idea which version was current at the time.
I hope that's useful.
and is smart enough not to compress data that is incompressible.
Smart would be compressing data that is incompressible.
Perhaps if it was some kind of super-medical-record with the potential to save the human race from almost certain extinction. For anything else, that's WAY overkill. For instance, credit card providers (who are subject to some pretty stringent requirements) are only required to maintain one off-site backup and one off-site on-line redundant data set.
If you were storing my medical records that way, I'd be incredibly confident that they won't ever become unavailable, and incredibly nervous that, with my medical records being transferred about and stored in so many places, my privacy would be compromised. Even with the best data-protection and encryption systems in place, you're only compounding the risks of data leaking, while not appreciably decreasing the risk of data loss (it was vanishingly low long before you got to 5 backups and 5 backups of the backups.)
Except possibly for the fact that the intersection of "People who purchase women's sporting gear" and "Slashdot readers" is going to be pretty close to a null set.
"People who purchase women's sporting gear for wearing in public and "Slashdot readers" is going to be pretty close to a null set."
FTFY
This was established in Australia decades ago too - this is not news. I don't care enough to check my reference, but I was fairly involved in it at the time, and from memory it was "Infopac International V RP Data, (late 80s or early 90s)". The judge ruled that the creative presentation of the data purchased at great expense by RP Data (the database and presentation software) was copyrightable, but that the data itself, being factual (real estate sales data), was not copyrightable. If Infopac International could access it legally (it could), there was nothing RP Data could do to restrict its use of the data.
Guess you didn't read my post.
Some highlights...
"No, it's not composing. It's still music." (What is music?)
"This is both music and composition." (What is music?)
"It's still music; it's not composing." (What is music?)
I was just trying to bring it back into context with the original post. That's what we're discussing, right? On the comments section for the post?
Hey, I was just attacking the content.
If I pick up my violin and play Mozart, then that's not music because I didn't write the notes?...
No, it's not composing. It's still music. If you play to a room full of people and claim you wrote it yourself, it's plagiarism.
What about a compositum, where I've taken a cool backbeat from another song, and improvise over that? According to your definition, what I'd be playing would be art, but the backbeat that (perhaps) provides the inspiration was originally music when played by the original creator, but now that I'm re-using it, it isn't music? Is the whole of it music, or just the part?
This is both music and composition. If you try to pass it off as your own work without mentioning the poor starving artist (or vastly rich dripping-with-gold-chains artist, or whatever) who wrote the backbeat and inspired your creation, it's plagiarism, unless you apply your creativity to the back-beat as well, rather than just lifting it verbatim.
What about a boom-box? If I have a recording of Itzhak Perlman playing that same Mozart piece, is it music if I play it out loud? Or was it only music the first time he played it?
This isn't about music; it's about creation. It's still music; it's not composing. The piece is about an author who 'performed' someone else's writing in text, and didn't tell people it was only a performance, not an original creation.
I'm terribly sorry, but you've done nothing but muddy the waters. Perhaps you could make a more precise definition that I could work with? Also, you seem very angry -- that's a strange self defeating choice as a method to appreciate art or music.
It's a shame you decided to stir the mud in with your own shit.
I'm not sure what your point is; for or against the author in question?
It's a little more modern than your examples, but imagine if Lizst's "Fantasies on a theme by Paganini" had just been excerpts of a violin playing the orininal piece, verbatim, bridged by sections of blending piano solo? That's not culture. What made it culture was when Lizst took, not the notation verbatim, but the idea, and built something entirely new but recognizably derived. Classical (and modern) music didn't excerpt: it used themes as a basis, not as a substitute for talent.
Music doesn't just 'sample and include' to create something new. It takes themes and creates on top of them. Imagine if any of the big 'derived' music styles had just put in whole excerpts from other styles, and added some bridging sections? They wouldn't have been the new sound they were. While I think it's okay to use elements of something someone else created in something new you create, there are three things you need to do:
Also, while a mutt can be an excellent dog, you're not allowed to sell it claiming it to be a pure-bred. And the analogy is stupid anyway.
Don't people read the articles? It clearly states that it does not work against ATMs. :P
This is exactly right. Steal my mag-stripe card, and you can copy it, but can't use it. Steal my EMV card, and you can't copy it, but you can take all my money.
The /. Effect: Thousands of users simultaneously accessing a site to not read its content.
How does that work? Link a page on /. and it's crippled by the extra load, but virtually none of the observable population of /. shows any signs of having even managed to work their browser well enough to click the link. It feels like there should be some analogy with Quantum Uncertainty.
EMV does, though. The banks promise, if your money is ever lost or stolen, that they will accept Zero Liability.
Ah, it was meant to be much more than that. It was meant to make it easier for the banks to blame the consumer for fraud, and avoid liability themselves. It has achieved this. You're right as regards actual security.
It's a cost-benefit analysis, and "pushing $x worth of liability onto someone else" works out to be not too much worse than "preventing $y worth of fraud", if $x = $y. Now if $x > $y, suddenly they're better off moving the liability than preventing the fraud. Similarly, if the first option costs more than the second, even if $x = $y, they're better off just moving liability.
This is why Britain has a historical problem of card fraud, and the US has a much better record. The US never let the banks push liability onto the consumer, regardless of the technology. This removes a whole segment of the CBA, meaning it's purely "cost of security vs. cost of fraud it would prevent".
As I did work in banking security for years...
This makes things worse. Before, the evidence trail was clear and understandable, and the consumer was well-protected. Signatures don't match? Consumer not liable.
Now, the system is still very breakable, but the evidence is complicated and often makes it look like the consumer was at fault. Consumer liable for transactions they never authorized.
The PIN storage or retry delay is not the issue. Cards ALREADY block access to the PIN after several failed attempts. The problem is that the bank is not able to detect that the card was never presented the PIN in the first place. The terminal thinks the PIN was verified, but it never gets passed on to the card.
The problem is that the MitM can spoof the "correct PIN was entered" response, and this isn't a part of the transaction MAC (basically an encrypted summary of what happened). If the card used the PIN verification result in the MAC the card issuer could detect this attack trivially. Unfortunately, this is an implementation detail which each bank would have to manage separately. New cards need to be rolled out.
And the consumer still has no idea if his card is safe or not.
5, Insightful? For one line of unjustified speculation? Are people really that desperate to spend their mod points?
Ever watch Inspector Gadget?
Not all asians are A-cups, you insensitive clod!
Some of them are Bs.
Currently, people in the market for materials like child porn go to sites which carry it. It's not really that hard to find (so they say, to justify the blacklist.) Police get warrants to monitor sites as they're found (judicial oversight!), then pounce and lock up both the producers and the people who create the market.
Now, create a fence around the web, making sure people can't get to these places at all. Nary an immoral thought to be found on the web! (Who decides? No oversight.) What's a child-pornographer to do? They won't just stop: they'll look for alternate means to distribute. So, it goes further underground (already, some of these bad industries are using VPNs to keep traffic better-hidden and encrypted.) No more warrants (because it's harder to intercept, and harder to prove it SHOULD be intercepted), and it's not just the technically-capable and paranoid who avoid getting caught: suddenly, you're creating a strong drive for ALL of these bad guys to use anti-detection technologies like this. No more low-hanging fruit, and more kids who never get rescued. Sure, it's harder to find: but it's still out there.
It's kind of like the gun debate: you only keep the guns out of the hands of the honest ones. In the gun debate, that's a fallacy: honest people still decide to carry guns, and they still sometimes end up shooting people. Honest people don't decide to casual-carry child porn: the only market is the baddies anyway. And they'll get it anyway. And now, we won't catch them.