Email sucks. I fucking hate it. It is, in my opinion, one of the worst technologies out there. If possible, it really needs to be replaced; I don't know how you'd go about doing that, because it'd have to be an all-or nothing change over (and likely gov't mandated), but I see this as one of the most crucial things for the whole of IT infrastructure.
The service provider doesn't matter; whether you're doing it yourself on a home server, paying $gazooks per year for a hosted service, or with a free service, they all have major and unpreventable downsides. It's always one (or more of) paying a lot, having to deal with spam and the appropriate filtering, lost mail, and downtime. This is unlike, say, web hosting, which while still facing problems, has significantly fewer and less severe issues.
Also unlike web services/hosting, email has become essential. It's the primary way that many people communicate professionally, and your email address is largely equivalent to many peoples' identity.
Is there a solution? Maybe. But I can't see it being workable without global oversight - "Internet 2" type governmental agency oversight, where anonymity is lost.
They did come up with a 'theory' as to why sensor drift doesn't matter: because they've got trendable data with that sensory drift model, which is not available elsewhere. Not exactly a theory, but it is an excuse.
Again, let me clarify that I'm not talking about DirectX. I'm talking about compiz in Linux, which is, as far as I know, utilizing OpenGL. AFAIK, there should not be the discrepancy due to software as there is with DirectX as the software is using the same mechanisms for rendering on both cards.
I fully realize (and expect!) that kind of performance hit with DirectX version increases; I've never heard of it with OGL (mainly because it's a pretty unmoving target).
IMO, they chose poorly. Yes, picking ATI allows them to hit the 'vertical' market better. However, provided they wanted to compete against Intel's GPU/chipset/processor combination in the business market segment, they picked the wrong choice, I think. Either that or they're marketing poorly, because the results are not doing well for the business market.
Yes, it does make it a better card. If a hardware "upgrade" results in a performance hit, in the same software, then the hardware (or hardware + software driver) is at fault. (Also, I seem to recall that the FX5x00 cards were considered crap compared to the GF4 cards, but that's just my recollection and I might have the model number/genealogy wrong.)
Recollect that I am primarily talking about 2D performance. You know, the visual performance factor that was a big priority back in the mid 1990s for almost everyone. We're not talking about DirectX here, in comparison of the Celeron to the Atom. And even if we were talking about 3D, we're talking about OGL - not DirectX - rendering. Performance decrease like that from generation to generation would be unacceptable, and it would not sell.
And how is an Atom not in the same market segment that the Celeron was in? If I recall correctly, the Celeron was marketed and sold in the "cheap and low-powered laptop" segment. That is the same market segment the Atom now holds.
Not only that, but the Celeron is half a dozen generations behind the Atom; are you saying that, clock for clock, the Atom is less than 1/2 as powerful as the Celeron? That's what I hear you saying. I mentioned the Celeron/Atom difference to say "no, the CPU was not compensating for the Radeon's age". If anything, relative performance should skew towards the Atom.
He already covered that - see "idiots". When has an intelligent person ever taken the overwhelming opinion of the masses as anything but reactionary bullshit?
Absolutely. I have a personal stance: if the results match my hypothesis, there's something wrong with either my hypothesis or the results, and they both need to be highly scrutinized. No estimation method is so accurate as to be dead-on every single time. Thorough backtracing is just a part of good science. Without such approaches, we'd never have discovered penicillin (at least when we did).
Likewise, if there's a drastic divergence from the status quo, all components of the system need to be analyzed, individually, to see if something unexpected is changing the results. In this specific scenario, looking outside the planet to find a possible cause of temperature changes might be appropriate on account of, oh, 99%+ of the heat on this planet coming from outside the atmosphere.
It really is frustrating how intensely climate science is doubted and denied. Economics - a far softer science with a (so far) vastly greater impact on human society - gets a staggering amount of leeway by comparison
It's somewhat funny, then, that prevailing economic theory has proven itself to be accurate to a much greater extent than climate theory has. Yes, economic theories change. But they change to match the data, and are usually not all that far from the mark. Contrast that with the "we're going to freeze/cook/freeze/cook/freeze/change/cook to death" approach of climatology 'science'.
There's a reason why climatic science has little impact on human society: people don't believe it. If people believed it (and all the sky-is-falling predictions which go with it), they'd invest in it: financially, socially, and emotionally. Yes, there are a few Chicken Littles, but nobody with any substance.
IANAS (Scientist) but I believe it has something to do with the prevalent carbon dating method relying on a form of circular logic: the age of a soil strata is based off of how old the object in it is, and the object is dated by which strata type it is in.
There's also the fact that it relies on the level of radioactive carbon in a piece of organic material. This is flawed, because carbon/CO2 in our atmosphere has varied in amount through time and by location.
Actually, science is an exact science. That's why we call it science, and not religion, and why things like economics are considered "soft sciences": IE, they're not based on concrete facts, just models. So, either this is exact science and bad science - having more in common with Scientology than the scientific method - or it's not actually a science at all, but soft science.
It's one thing for people to make mistakes; it's another thing entirely when an entire (government funded!) organization makes wild - officious! - claims that the sky is falling. Furthermore, it's simply preposterous when, upon being confronted with the incongruity of their claims and reality, they throw up a strawman about the "untested" nature of the other modeling method. Hello? If the modeling method you're using is fundamentally flawed to the point that it is, it doesn't matter how untested the other method is; the method you're using is fucked and anything else is going to be at least comparable, and likely better.
While reading your post, one thing came to mind: NVIDIA's Tegra.
Not only does it boast a multicore 1GHz (iirc) ARM9-based SoC architecture, but it's very low power for what it can do and can do more than current Intel shit can do by quite a bit. And, if it's keeping with other ARM based devices, manufacturing and licensing costs will be significantly less than x86 based counterparts.
With a device which does not require legacy support, it won't matter what the CPU is provided it performs. And when you've got a SoC which can do hardware (as opposed to software) decoding of video and a myriad of other nifty things, its going to be interesting to see whether such a device is selected with a BSD or Linux kernel over something else. I personally suspect, very highly, that we'll see a Tegra (or similar) based gaming system come from either Sony or Nintendo (or maybe even a start up), next time around. There are certainly a lot of options in the ARM9 SoC crowd right now (and NVIDIAs seems to be the best).
For that matter, NVIDIA might just try and get a corner on the "netbook" market with the Tegra. It would be very well suited to the task, if I do say so myself. It would be fitting if that market segment returned to its roots (ARM) with an upgrade (UNIX/Linux vs. WinCE/EPOC).
There's a slight problem with that article (and your reasoning). We're still a long way off to making most applications (and even operating systems) effectively use multicore CPUs through parallelism. Or, for that matter, even use multiple threads. We are a long, long way off to actually having software which would work in such parallelism.
Even if the hardware is there, it'll take a while for it to be adopted in any real sense, I think.
They most certainly wanted more than just the 2D chipset. If they'd wanted the 2D chipset, they'd have been very well off to have just picked up the Matrox company for a fraction of the price; Matrox always had very fast, well-made 2D cards (and drivers), and their early 3D stuff wasn't half bad. They were also the best cards for Linux for a long time - and may even still be,for that matter (but I haven't seen any of their stuff in so long I wouldn't know).
AMD could probably have cornered the "hobbyist Linux user" (granted, for only a couple percentage points of the market, at best) if they'd played it right. And with a lower overhead in providing that "integrated solution" they may have had a better time of cornering the business market. Unlike ATI, Matrox actually is (or was) known in the business world as having good hardware - and for that matter, having hardware which is well-suited to business applications.
No joke: aside from the more advanced support for DirectX and the like, an old Fujitsu laptop I've got with an ATI Radeon chipset has better performance with Compiz than an MSI Wind PC with i9x0 graphics (Atom). The old Fujitsu clocks at 500MHz; it's a Celeron.
Intel graphics are crap at basic 2D desktop manipulation, as well. 7-year-old Nvidia cards outperform the latest, greatest from Intel without breaking a sweat. Same for integrated ATI stuff from the same era (or earlier - ATI Radeon smokes i9x0 in a significantly older laptop, both 2d and with compiz). It's like the difference between a Matrox G200 and anything else from that era.
And that doesn't even consider the fact that Vista, by default, needs 3D acceleration AND a fast CPU to work well. This isn't such an issue now that we've got multicore laptops and desktops as standard fare, but shit: what sane person is going to want to sacrifice that much CPU overhead to run stupid 3D windowing bling? Same goes for compiz in Linux.
And, despite all the rutting and fanfare Intel has provoked over its strong Linux support (it's very good), it doesn't even begin to make up for Intel's graphic hardware shortcomings. Their hardware is least-common-denominator in terms of speed and performance. It's as bad as, if not worse than, the horrid SiS-based graphics leech chipsets from ~1996 or so when all is said and done. At least those worked decently in Linux once drivers were written.
Yep. He has, at this point, essentially been ruined whether he gets off or not. This is not only due to this specific line of work - security and IT - but because he lives in a modern society where you can not simply change identities at whim.
One, two hundred years ago, it was possible to simply leave town and start over. Not anymore. You can travel across the world and you will still have to identify yourself as yourself, unless you want to fly under the radar. Today, simply being tried - and sometimes even accused - of a crime is enough to ruin someone's life, irrevocably.
What is Childs going to do, if he were to get out? Even if he were to change his name and fib work history for the past year+, the background check would show otherwise. If he were honest, there'd still be the stigma of wrongdoing associated with him, and he'd be fucked by HR.
Hell, we don't even rely on vetting people through association or demonstration of competence anymore. It's all a matter of whether or not you can pass a background check, and whether your credit and
Good luck to Childs if he's innocent. He's going to need it: he'll never work in his field of choice again (or any field which requires responsibility), and he'll need every bit of help he can get in suing his state's government to get a damages settlement. A reasonable settlement would be, what, $2.5 million for working in his field for the next 25 years?
Biometrics falls flat in that regard because people are misusing biometrics. Biometrics is not authentication; it's verification of who you are. You still need the "what you know" part of a security mechanism to make it one.
If biometrics ever become endemic, with a proper auth/verification pair, we might, however, run into a big problem in the event of biometric information databases becoming hacked. It would provide a gateway for anyone who has hacked the database to change your authentication method. This is true with conventional methods, as well, but this seems like it'd be somewhat more tricky to clean up than simply changing a username.
Biometrics are, however, very good for authentication - much better than a usename, at least.
Security authentication and verification is a two step process, naturally. The authentication is usually something that you have, and the verification (password) is something that you know. Unlike in the movies where your eyes are scanned and your "voice is your password" or however that goes, two biometric scans do not count as verification - that's simply two-fold authentication.
In the real world, regardless of how good the biometrics are, you still need the "authentication" stage. (OT: thank you, Hollywood, for making even computer people think that biometrics qualify as verification and authentication.) For the verification, biometrics are substantially better than a username or access badge: in order to break through the initial verification stage, you need to undertake a fairly convoluted, obvious approach to breach them (at least for photo verification). This holds true even with the relatively 'weak' crackability of facial recognition.
But you still need an authentication method - something you know - to back up the verification, or it's closer to having a "root" account with a null password than having a secure 10-digit alphanumeric one.
It would seem to me that the biggest cause of the facial recognition failing would be its inability to effectively measure depth. Does it use a standard web cam for this facial recognition?
If so - and it does not do this already - why not simply have it do facial recognition at multiple (2, 3, 4+) depths? IE, compare the image of the ears at full focus, the nose, the temples, jaw, and cheeks. Additionally, using two cameras (say, one on either side of the screen) could fix the problem - but I suspect that taking multiple shots at different depths of the face (provided the camera could change focus finitely and quickly enough) would be a sufficient software solution.
This doesn't seem like an insurmountable flaw to me; it seems like an incomplete solution.
I'm not going to disagree with your initial premise, but: isn't this specific problem being caused by a binary blob module, where the perl module was compiled against a different perl binary? It's a packaging and version control issue, not a scripting language/perl problem. Though CPAN could, arguably, be partially to blame for this snafu, from the sound of it.
Slashdot Mirror
The truly funny thing is that they only just barely beat out this QZHTTP, which is likely just a rebranded Apache or another open source web server.
Email sucks. I fucking hate it. It is, in my opinion, one of the worst technologies out there. If possible, it really needs to be replaced; I don't know how you'd go about doing that, because it'd have to be an all-or nothing change over (and likely gov't mandated), but I see this as one of the most crucial things for the whole of IT infrastructure.
The service provider doesn't matter; whether you're doing it yourself on a home server, paying $gazooks per year for a hosted service, or with a free service, they all have major and unpreventable downsides. It's always one (or more of) paying a lot, having to deal with spam and the appropriate filtering, lost mail, and downtime. This is unlike, say, web hosting, which while still facing problems, has significantly fewer and less severe issues.
Also unlike web services/hosting, email has become essential. It's the primary way that many people communicate professionally, and your email address is largely equivalent to many peoples' identity.
Is there a solution? Maybe. But I can't see it being workable without global oversight - "Internet 2" type governmental agency oversight, where anonymity is lost.
C'mon man. Where's your critical thinking? The answer is obvious. The Atlantians were obviously GIANTS.
They did come up with a 'theory' as to why sensor drift doesn't matter: because they've got trendable data with that sensory drift model, which is not available elsewhere. Not exactly a theory, but it is an excuse.
Again, let me clarify that I'm not talking about DirectX. I'm talking about compiz in Linux, which is, as far as I know, utilizing OpenGL. AFAIK, there should not be the discrepancy due to software as there is with DirectX as the software is using the same mechanisms for rendering on both cards.
I fully realize (and expect!) that kind of performance hit with DirectX version increases; I've never heard of it with OGL (mainly because it's a pretty unmoving target).
IMO, they chose poorly. Yes, picking ATI allows them to hit the 'vertical' market better. However, provided they wanted to compete against Intel's GPU/chipset/processor combination in the business market segment, they picked the wrong choice, I think. Either that or they're marketing poorly, because the results are not doing well for the business market.
Yes, it does make it a better card. If a hardware "upgrade" results in a performance hit, in the same software, then the hardware (or hardware + software driver) is at fault. (Also, I seem to recall that the FX5x00 cards were considered crap compared to the GF4 cards, but that's just my recollection and I might have the model number/genealogy wrong.)
Recollect that I am primarily talking about 2D performance. You know, the visual performance factor that was a big priority back in the mid 1990s for almost everyone. We're not talking about DirectX here, in comparison of the Celeron to the Atom. And even if we were talking about 3D, we're talking about OGL - not DirectX - rendering. Performance decrease like that from generation to generation would be unacceptable, and it would not sell.
And how is an Atom not in the same market segment that the Celeron was in? If I recall correctly, the Celeron was marketed and sold in the "cheap and low-powered laptop" segment. That is the same market segment the Atom now holds.
Not only that, but the Celeron is half a dozen generations behind the Atom; are you saying that, clock for clock, the Atom is less than 1/2 as powerful as the Celeron? That's what I hear you saying. I mentioned the Celeron/Atom difference to say "no, the CPU was not compensating for the Radeon's age". If anything, relative performance should skew towards the Atom.
He already covered that - see "idiots". When has an intelligent person ever taken the overwhelming opinion of the masses as anything but reactionary bullshit?
Absolutely. I have a personal stance: if the results match my hypothesis, there's something wrong with either my hypothesis or the results, and they both need to be highly scrutinized. No estimation method is so accurate as to be dead-on every single time. Thorough backtracing is just a part of good science. Without such approaches, we'd never have discovered penicillin (at least when we did).
Likewise, if there's a drastic divergence from the status quo, all components of the system need to be analyzed, individually, to see if something unexpected is changing the results. In this specific scenario, looking outside the planet to find a possible cause of temperature changes might be appropriate on account of, oh, 99%+ of the heat on this planet coming from outside the atmosphere.
It really is frustrating how intensely climate science is doubted and denied. Economics - a far softer science with a (so far) vastly greater impact on human society - gets a staggering amount of leeway by comparison
It's somewhat funny, then, that prevailing economic theory has proven itself to be accurate to a much greater extent than climate theory has. Yes, economic theories change. But they change to match the data, and are usually not all that far from the mark. Contrast that with the "we're going to freeze/cook/freeze/cook/freeze/change/cook to death" approach of climatology 'science'.
There's a reason why climatic science has little impact on human society: people don't believe it. If people believed it (and all the sky-is-falling predictions which go with it), they'd invest in it: financially, socially, and emotionally. Yes, there are a few Chicken Littles, but nobody with any substance.
IANAS (Scientist) but I believe it has something to do with the prevalent carbon dating method relying on a form of circular logic: the age of a soil strata is based off of how old the object in it is, and the object is dated by which strata type it is in.
There's also the fact that it relies on the level of radioactive carbon in a piece of organic material. This is flawed, because carbon/CO2 in our atmosphere has varied in amount through time and by location.
A few degrees C? Don't you mean a few hundredths of a degree C?
Actually, science is an exact science. That's why we call it science, and not religion, and why things like economics are considered "soft sciences": IE, they're not based on concrete facts, just models. So, either this is exact science and bad science - having more in common with Scientology than the scientific method - or it's not actually a science at all, but soft science.
It's one thing for people to make mistakes; it's another thing entirely when an entire (government funded!) organization makes wild - officious! - claims that the sky is falling. Furthermore, it's simply preposterous when, upon being confronted with the incongruity of their claims and reality, they throw up a strawman about the "untested" nature of the other modeling method. Hello? If the modeling method you're using is fundamentally flawed to the point that it is, it doesn't matter how untested the other method is; the method you're using is fucked and anything else is going to be at least comparable, and likely better.
While reading your post, one thing came to mind: NVIDIA's Tegra.
Not only does it boast a multicore 1GHz (iirc) ARM9-based SoC architecture, but it's very low power for what it can do and can do more than current Intel shit can do by quite a bit. And, if it's keeping with other ARM based devices, manufacturing and licensing costs will be significantly less than x86 based counterparts.
With a device which does not require legacy support, it won't matter what the CPU is provided it performs. And when you've got a SoC which can do hardware (as opposed to software) decoding of video and a myriad of other nifty things, its going to be interesting to see whether such a device is selected with a BSD or Linux kernel over something else. I personally suspect, very highly, that we'll see a Tegra (or similar) based gaming system come from either Sony or Nintendo (or maybe even a start up), next time around. There are certainly a lot of options in the ARM9 SoC crowd right now (and NVIDIAs seems to be the best).
For that matter, NVIDIA might just try and get a corner on the "netbook" market with the Tegra. It would be very well suited to the task, if I do say so myself. It would be fitting if that market segment returned to its roots (ARM) with an upgrade (UNIX/Linux vs. WinCE/EPOC).
There's a slight problem with that article (and your reasoning). We're still a long way off to making most applications (and even operating systems) effectively use multicore CPUs through parallelism. Or, for that matter, even use multiple threads. We are a long, long way off to actually having software which would work in such parallelism.
Even if the hardware is there, it'll take a while for it to be adopted in any real sense, I think.
Definately! If you don't think Nvidia can produce low-power GPUs, look at their supposedly-upcoming Tegra ARM-based SoC: it's power use is a pittance.
They most certainly wanted more than just the 2D chipset. If they'd wanted the 2D chipset, they'd have been very well off to have just picked up the Matrox company for a fraction of the price; Matrox always had very fast, well-made 2D cards (and drivers), and their early 3D stuff wasn't half bad. They were also the best cards for Linux for a long time - and may even still be ,for that matter (but I haven't seen any of their stuff in so long I wouldn't know).
AMD could probably have cornered the "hobbyist Linux user" (granted, for only a couple percentage points of the market, at best) if they'd played it right. And with a lower overhead in providing that "integrated solution" they may have had a better time of cornering the business market. Unlike ATI, Matrox actually is (or was) known in the business world as having good hardware - and for that matter, having hardware which is well-suited to business applications.
No joke: aside from the more advanced support for DirectX and the like, an old Fujitsu laptop I've got with an ATI Radeon chipset has better performance with Compiz than an MSI Wind PC with i9x0 graphics (Atom). The old Fujitsu clocks at 500MHz; it's a Celeron.
Intel graphics are crap at basic 2D desktop manipulation, as well. 7-year-old Nvidia cards outperform the latest, greatest from Intel without breaking a sweat. Same for integrated ATI stuff from the same era (or earlier - ATI Radeon smokes i9x0 in a significantly older laptop, both 2d and with compiz). It's like the difference between a Matrox G200 and anything else from that era.
And that doesn't even consider the fact that Vista, by default, needs 3D acceleration AND a fast CPU to work well. This isn't such an issue now that we've got multicore laptops and desktops as standard fare, but shit: what sane person is going to want to sacrifice that much CPU overhead to run stupid 3D windowing bling? Same goes for compiz in Linux.
And, despite all the rutting and fanfare Intel has provoked over its strong Linux support (it's very good), it doesn't even begin to make up for Intel's graphic hardware shortcomings. Their hardware is least-common-denominator in terms of speed and performance. It's as bad as, if not worse than, the horrid SiS-based graphics leech chipsets from ~1996 or so when all is said and done. At least those worked decently in Linux once drivers were written.
Yep. He has, at this point, essentially been ruined whether he gets off or not. This is not only due to this specific line of work - security and IT - but because he lives in a modern society where you can not simply change identities at whim.
One, two hundred years ago, it was possible to simply leave town and start over. Not anymore. You can travel across the world and you will still have to identify yourself as yourself, unless you want to fly under the radar. Today, simply being tried - and sometimes even accused - of a crime is enough to ruin someone's life, irrevocably.
What is Childs going to do, if he were to get out? Even if he were to change his name and fib work history for the past year+, the background check would show otherwise. If he were honest, there'd still be the stigma of wrongdoing associated with him, and he'd be fucked by HR.
Hell, we don't even rely on vetting people through association or demonstration of competence anymore. It's all a matter of whether or not you can pass a background check, and whether your credit and
Good luck to Childs if he's innocent. He's going to need it: he'll never work in his field of choice again (or any field which requires responsibility), and he'll need every bit of help he can get in suing his state's government to get a damages settlement. A reasonable settlement would be, what, $2.5 million for working in his field for the next 25 years?
Biometrics falls flat in that regard because people are misusing biometrics. Biometrics is not authentication; it's verification of who you are. You still need the "what you know" part of a security mechanism to make it one.
If biometrics ever become endemic, with a proper auth/verification pair, we might, however, run into a big problem in the event of biometric information databases becoming hacked. It would provide a gateway for anyone who has hacked the database to change your authentication method. This is true with conventional methods, as well, but this seems like it'd be somewhat more tricky to clean up than simply changing a username.
Biometrics are, however, very good for authentication - much better than a usename, at least.
Security authentication and verification is a two step process, naturally. The authentication is usually something that you have, and the verification (password) is something that you know. Unlike in the movies where your eyes are scanned and your "voice is your password" or however that goes, two biometric scans do not count as verification - that's simply two-fold authentication.
In the real world, regardless of how good the biometrics are, you still need the "authentication" stage. (OT: thank you, Hollywood, for making even computer people think that biometrics qualify as verification and authentication.) For the verification, biometrics are substantially better than a username or access badge: in order to break through the initial verification stage, you need to undertake a fairly convoluted, obvious approach to breach them (at least for photo verification). This holds true even with the relatively 'weak' crackability of facial recognition.
But you still need an authentication method - something you know - to back up the verification, or it's closer to having a "root" account with a null password than having a secure 10-digit alphanumeric one.
It would seem to me that the biggest cause of the facial recognition failing would be its inability to effectively measure depth. Does it use a standard web cam for this facial recognition?
If so - and it does not do this already - why not simply have it do facial recognition at multiple (2, 3, 4+) depths? IE, compare the image of the ears at full focus, the nose, the temples, jaw, and cheeks. Additionally, using two cameras (say, one on either side of the screen) could fix the problem - but I suspect that taking multiple shots at different depths of the face (provided the camera could change focus finitely and quickly enough) would be a sufficient software solution.
This doesn't seem like an insurmountable flaw to me; it seems like an incomplete solution.
That's metrics for ya!
I'm not going to disagree with your initial premise, but: isn't this specific problem being caused by a binary blob module, where the perl module was compiled against a different perl binary? It's a packaging and version control issue, not a scripting language/perl problem. Though CPAN could, arguably, be partially to blame for this snafu, from the sound of it.