Allow me to introduce you to two very fun programs, "which" and "find". Which prints out where a program in your path is. This is useful when you install a new version and cant seem to run it, or for Paul to tell his coworker where he installed something. Find will search a specified tree for the properties you want, like perhaps the name of the file and executable permission.
Consider how your approach differs from what's usually considered ideal (or at least very close): one time pads. The key is used once and there's no way to show that any decryption is more valid than another. This is of course unsuitable for the internet and other places where communication are never trustworthy. HTTPS' approach to this doesn't solve the problem, it delays it. It's computationally expensive yes, but possible. Suitable at the moment for consumer technologies like banking, but long term, quantum computing could undo the mathematical challenges involved.
So these secure channels are intended to help solve this key exchange problem. With a secure line, you can distribute keys long distances without the fear of eavesdroppers making a copy. Or you can send the message if the link is fast enough. There are two kinds of "quantum" commonly associated with crypto. There's one that relies on photon uncertainty to create these sorts of secure channels, and another technology used to factor numbers. Potentially you could factor numbers fast enough to make cracking RSA feasible. At which point the whole HTTPS thing basically fails.
Basically, this field of research is intended for places and futures where HTTPS isn't enough. But these things will require long periods of scrutiny before we can safely declare them secure. Unfortunately, "ten years with no published holes" is rarely newsworthy. And of course, this is only a link level system. Information can still leak from other parts of the system in these technologies.
Ironically, most of Google's code has been rewritten to MapReduce style. Interestingly, the core functionality is written in C++. I guess it helps make the function oriented system a bit simpler to implement. The end result is that most of their programmers don't focus on savoring every cycle. Today's computers can run something like a million instructions in the time it takes to serve a single disk read; making code that can be run on 1800 nodes is far more important than how many fewer cycles your uniprocessor code can accommodate. Fewer instructions is better, but there's far more performance gains to be had from better locality -- process the data on the node that holds it.
The rest of the things you've mentioned all outdate C++, let alone newer languages. And most don't use assembly at at all (apache doesn't, I doubt BIND or sendmail does). The UNIX kernels don't use assembly because it's faster (though it helps), but because there are things that must be done that are platform specific, and C's provided way of handling those cases is assembly. But generally, x86 computers are changing too rapidly to make any optimizations worthwhile. Just make clear code and let the compiler handle the dirty work. If you don't think it's good enough, improve the compiler. But usually the people who claim to "write their inner loops in ASM" have troubles making solid contributions to compiler technology.
It's not Microsoft putting those on your XP image. It's your OEM. The copies you buy at the store do not have these "fun" add ons. I imagine this is because of Microsoft's monopoly conviction. You are correct that Dell isn't unique in putting crap on the drive, though.
People were going to think that the two projects were coincidentally similar? Or was he worried that people would think that Nintendo didn't rip him off? Unless he showed Okami to them sometime before Nintendo decided to put it in TP, I don't see where he'd have any convincing argument to say "Nintendo ripped me off".
Actually that does work for me on occasion. But only because they're bastards about IPs. Your cable modem gives out one IP, and binds it to a MAC address. Pulling the power on the modem drops that association, making your new computer work with the network. Maybe cable modems these days have NAT etc built into them, but back in 1997 no such luck.
I agree it's a far more complex reality. But you'll never find practical analysis if you never go looking for it, and our biologists in general are NOT equipped to even start.
Well, I've been able to blue screen my laptop a couple times. Bad hardware (in most cases) shouldn't be able to blue screen the computer. Sure, you can blame things on bad drivers, but why should windows be exempted from driver quality problems? The WHQL was created for a reason!
This isn't exactly a novice computer user approaching Ubuntu. She knows she has a 64 bit processor, and made a choice. Generally one I regard as a poor one -- 64bit binaries aren't all there yet, and you never know what code doesn't compile when pointers aren't 32 bits long. 64bit linux is fine if you're willing to accept the consequences of running a non-standard architecture, so there's no less reason to publish 64bit Ubuntu than PPC in that regard. But don't expect closed source crap to run on your system. Windows has similar problems.
Note that the windows installer would probably fail her own requirements and testing. But unlike Vista, Ubuntu will detect most Windows installs and make a grub entry for you. It's probably not perfect, but I've never had to restore the MBR because Ubuntu overwrote it with something that didn't share. The good news is that the Dell scenario potentially avoids this whole mess (and highlights how crappy their MediaDirect system is).
I was chatting with an local entomologist student and during the conversation it became clear that biology is sort of stuck -- it's perceived as a mathematically weak science, which in turn leads to weaker math skills among scientists, and as a result never push mathematical approaches within the science. It might be that biology simply isn't suitable for the formulaic approach that physics and chemistry seem to be, but a side effect is that their students are also really bad at statistics, which is still very useful when studying populations.
In my experience, the reason nobody saw LOGO as an introduction to programming is because the teachers were not trained. This meant you went to the computer lab, a strangely dark place (I guess CRTs werent bright in the 80s), and everyone made a square. That was about the extent my peers received of the dark arts.
My problem with BASIC is that line numbers suck, and GOSUBs aren't very useful compared to full fledged procedures with parameters. QBASIC seems to have fixed that, but that was after my time. But similar to how I don't use a horse and buggy, I don't feel the need to try out older bad systems in order to use or appreciate new ones.
Actually, you're welcome to make your own linux port. Just download Squeak VM (Ubuntu has packages for it, I'm sure others do too) and the windows zip file. The Scratch.image is the part you're after. Just run "squeak Scratch.image". After that it worked perfectly for me, complete with walking meowing cat.
I just put up system monitor gnome applet. It's easy to tell because I have i/o wait set to yellow. Some day I'll remember to file a bug report on gnome's bugzilla about the default colors. It seems like such a silly thing to nit pick at, but I guess it helps determine such things.
Wind resistance is the primary reason you have peak efficiencies in geared motors, like you mentioned. The solar cars my roommate built had some pretty good aerodynamics, but they were seeing huge drops in power usage at 55 vs 50 which makes a hell of a lot of sense. As I understand it, ICE has a torque curve complete with a peak, while electrics are pretty much constant.
It's very difficult for me to imagine how going faster would increase engine efficiency past say 40 miles an hour. Peak efficiency at 80 mph just seems like a lousy excuse to tell your mom.
I've never seen a Hummer for to civilians driven in that manner. I have however taken a National Guard one out for a spin offroad, and it's a bit painful (maybe sand and brush nicer than unused farmland...). I can only imagine the one you can buy is less equipped to handle that mission. They were a bit amusing when you couldn't buy one off the dealership lot, but when I see stretch Hummers I cant help but think of Eisenhower's warning.
I call shenanigans. We've got solar cars that show severe drops going from 55 to 50, with arguably far better aerodynamics than the 240SX (at considerable expense). Throwing out numbers is not a substitute for automotive engineering expertise.
And Time splitters. And Turok. But Goldeneye was fucking huge. The game was popular enough that up until three years ago a 200 player tournament was held yearly on campus. Then the guy graduated and Halo's been the order of the day. The way I figure, Halo took off in part because Rare couldn't hold the Bond license.
I'm sure there are plenty of dorm room nerds who could make something valuable for 20 thousand but can't convince the local bank to make a business loan for what amounts to a website. And a few might even take the deal. But you're right. This is the kind of cash I could come up with as a grad student. There's no point in taking the money if I believe the idea has wheels. If that's the kind of bad businesmen they're after, well then maybe Morris hasn't quite fully recovered from prison.
Graham selling tactic in the past was that it was closer in nature to an internship in entrepreneurship. It'd have to be, if the net value is four hundred thousand. But the idea of using it as an IQ test has some (perhaps unintentional) merit: don't turn away a guy with money interested in funding your ideas, negiotiate.
...and essentially zero false positives. I do not check my junk mail folder, simply can't afford to.
Best paper at Graham's first spam conference was by an MIT undergrad from course 6 who completely debunked the whole notion with some rudimentary statistical analysis.
I'm pretty sure the first thing to do when measuring false positives is to count them. Maybe that paper has some insight on how to do that without looking at them, but I'm still waiting for the.iso they distribute the proceedings by to finish downloading. But you're welcome to continue the faith that your software is clearly working correctly simply because you paid for it.
And, you do realize that Graham's paper specifically cited headers as a reason Bayesian techniques in the past had failed, right? I'm not sure how timing etc plays into it; I get valid emails at all times of the day -- some students are up as late as I am, some are early risers, and sometimes I get mail from my boss during the day. Really, it all depends on the person. If your use doesn't engage the public much, you could probably get away with white lists. If you're part of a large company, you'll have access to a lot of email for known good and labelled junk mail. By the amount of mail I get with no obvious purpose, I surmise that some statistical analysis tools are working.
Actually, thats one of the things that a VP could probably promise without severe problems for engineering. I guess you'd probably have to filter out the curse words in comments though...
So many comments posted thus far, and nothing about Affero. You can't discuss what free software in a "WebOS" (what a horrible term) means without bringing this up. The fundamental point is that the GPL allows anyone modification without no restrictions or obligations as long as you don't distribute the software to other people. Distribution is generally interpreted as different than running it. Basically, if you GPL a PHP driven website, anyone can take it, modify it, give other users access to "running" it without distributing any changes you've made.
As best I can tell, the Affero license addresses this, by a clause partnered with specific functionality in the program. 'If this software came with functionality to give source code to users, you may not remove it.' or something to that effect. I've already seen one or two sites that have decided that it didn't apply to them, and the development community behind it sounded like "so what?" Which is fine I suppose, but it feels strange to me that they were able to make a site for a client based largely on existing OAGPL'd code, integrate the code with existing technology(that is also widely used and would be interesting to many people), fix bugs, and then turn around and declare yourself unable and unrequired to fulfill the obligations imposed on you by the agreement. This is compounded by the fact that generally its much harder to tell when a site is appropriating OAGPL'd code. With traditional software, strings will usually catch hidden strings, debugging statements etc. Web software only has the output to look at.
I am somewhat comforted by the fact their site is already out of date, so either they or their client will be facing increasing costs in maintaining the site.
Slashdot Mirror
Allow me to introduce you to two very fun programs, "which" and "find". Which prints out where a program in your path is. This is useful when you install a new version and cant seem to run it, or for Paul to tell his coworker where he installed something. Find will search a specified tree for the properties you want, like perhaps the name of the file and executable permission.
Consider how your approach differs from what's usually considered ideal (or at least very close): one time pads. The key is used once and there's no way to show that any decryption is more valid than another. This is of course unsuitable for the internet and other places where communication are never trustworthy. HTTPS' approach to this doesn't solve the problem, it delays it. It's computationally expensive yes, but possible. Suitable at the moment for consumer technologies like banking, but long term, quantum computing could undo the mathematical challenges involved.
So these secure channels are intended to help solve this key exchange problem. With a secure line, you can distribute keys long distances without the fear of eavesdroppers making a copy. Or you can send the message if the link is fast enough. There are two kinds of "quantum" commonly associated with crypto. There's one that relies on photon uncertainty to create these sorts of secure channels, and another technology used to factor numbers. Potentially you could factor numbers fast enough to make cracking RSA feasible. At which point the whole HTTPS thing basically fails.
Basically, this field of research is intended for places and futures where HTTPS isn't enough. But these things will require long periods of scrutiny before we can safely declare them secure. Unfortunately, "ten years with no published holes" is rarely newsworthy. And of course, this is only a link level system. Information can still leak from other parts of the system in these technologies.
Ironically, most of Google's code has been rewritten to MapReduce style. Interestingly, the core functionality is written in C++. I guess it helps make the function oriented system a bit simpler to implement. The end result is that most of their programmers don't focus on savoring every cycle. Today's computers can run something like a million instructions in the time it takes to serve a single disk read; making code that can be run on 1800 nodes is far more important than how many fewer cycles your uniprocessor code can accommodate. Fewer instructions is better, but there's far more performance gains to be had from better locality -- process the data on the node that holds it.
The rest of the things you've mentioned all outdate C++, let alone newer languages. And most don't use assembly at at all (apache doesn't, I doubt BIND or sendmail does). The UNIX kernels don't use assembly because it's faster (though it helps), but because there are things that must be done that are platform specific, and C's provided way of handling those cases is assembly. But generally, x86 computers are changing too rapidly to make any optimizations worthwhile. Just make clear code and let the compiler handle the dirty work. If you don't think it's good enough, improve the compiler. But usually the people who claim to "write their inner loops in ASM" have troubles making solid contributions to compiler technology.
I'm pretty sure the Google Toolbar installs in Linux.
It's not Microsoft putting those on your XP image. It's your OEM. The copies you buy at the store do not have these "fun" add ons. I imagine this is because of Microsoft's monopoly conviction. You are correct that Dell isn't unique in putting crap on the drive, though.
But it's a good question: how much do you trust the CVS authors? 16 thousand might be chump change, but how bout a couple million?
People were going to think that the two projects were coincidentally similar? Or was he worried that people would think that Nintendo didn't rip him off? Unless he showed Okami to them sometime before Nintendo decided to put it in TP, I don't see where he'd have any convincing argument to say "Nintendo ripped me off".
Umm, I imagine it would look a lot like Four Swords.
Actually that does work for me on occasion. But only because they're bastards about IPs. Your cable modem gives out one IP, and binds it to a MAC address. Pulling the power on the modem drops that association, making your new computer work with the network. Maybe cable modems these days have NAT etc built into them, but back in 1997 no such luck.
I agree it's a far more complex reality. But you'll never find practical analysis if you never go looking for it, and our biologists in general are NOT equipped to even start.
Well, I've been able to blue screen my laptop a couple times. Bad hardware (in most cases) shouldn't be able to blue screen the computer. Sure, you can blame things on bad drivers, but why should windows be exempted from driver quality problems? The WHQL was created for a reason!
This isn't exactly a novice computer user approaching Ubuntu. She knows she has a 64 bit processor, and made a choice. Generally one I regard as a poor one -- 64bit binaries aren't all there yet, and you never know what code doesn't compile when pointers aren't 32 bits long. 64bit linux is fine if you're willing to accept the consequences of running a non-standard architecture, so there's no less reason to publish 64bit Ubuntu than PPC in that regard. But don't expect closed source crap to run on your system. Windows has similar problems.
Note that the windows installer would probably fail her own requirements and testing. But unlike Vista, Ubuntu will detect most Windows installs and make a grub entry for you. It's probably not perfect, but I've never had to restore the MBR because Ubuntu overwrote it with something that didn't share. The good news is that the Dell scenario potentially avoids this whole mess (and highlights how crappy their MediaDirect system is).
I was chatting with an local entomologist student and during the conversation it became clear that biology is sort of stuck -- it's perceived as a mathematically weak science, which in turn leads to weaker math skills among scientists, and as a result never push mathematical approaches within the science. It might be that biology simply isn't suitable for the formulaic approach that physics and chemistry seem to be, but a side effect is that their students are also really bad at statistics, which is still very useful when studying populations.
In my experience, the reason nobody saw LOGO as an introduction to programming is because the teachers were not trained. This meant you went to the computer lab, a strangely dark place (I guess CRTs werent bright in the 80s), and everyone made a square. That was about the extent my peers received of the dark arts.
My problem with BASIC is that line numbers suck, and GOSUBs aren't very useful compared to full fledged procedures with parameters. QBASIC seems to have fixed that, but that was after my time. But similar to how I don't use a horse and buggy, I don't feel the need to try out older bad systems in order to use or appreciate new ones.
Actually, you're welcome to make your own linux port. Just download Squeak VM (Ubuntu has packages for it, I'm sure others do too) and the windows zip file. The Scratch.image is the part you're after. Just run "squeak Scratch.image". After that it worked perfectly for me, complete with walking meowing cat.
I just put up system monitor gnome applet. It's easy to tell because I have i/o wait set to yellow. Some day I'll remember to file a bug report on gnome's bugzilla about the default colors. It seems like such a silly thing to nit pick at, but I guess it helps determine such things.
Wind resistance is the primary reason you have peak efficiencies in geared motors, like you mentioned. The solar cars my roommate built had some pretty good aerodynamics, but they were seeing huge drops in power usage at 55 vs 50 which makes a hell of a lot of sense. As I understand it, ICE has a torque curve complete with a peak, while electrics are pretty much constant.
It's very difficult for me to imagine how going faster would increase engine efficiency past say 40 miles an hour. Peak efficiency at 80 mph just seems like a lousy excuse to tell your mom.
I've never seen a Hummer for to civilians driven in that manner. I have however taken a National Guard one out for a spin offroad, and it's a bit painful (maybe sand and brush nicer than unused farmland...). I can only imagine the one you can buy is less equipped to handle that mission. They were a bit amusing when you couldn't buy one off the dealership lot, but when I see stretch Hummers I cant help but think of Eisenhower's warning.
I call shenanigans. We've got solar cars that show severe drops going from 55 to 50, with arguably far better aerodynamics than the 240SX (at considerable expense). Throwing out numbers is not a substitute for automotive engineering expertise.
And Time splitters. And Turok. But Goldeneye was fucking huge. The game was popular enough that up until three years ago a 200 player tournament was held yearly on campus. Then the guy graduated and Halo's been the order of the day. The way I figure, Halo took off in part because Rare couldn't hold the Bond license.
I love FPSs. Counter-Strike, Battlefield, Day of Defeat, Unreal Tournament, they're all fine games.
I also like Metroid. Should I be worried about spontaneous explosion at any moment?
I'm sure there are plenty of dorm room nerds who could make something valuable for 20 thousand but can't convince the local bank to make a business loan for what amounts to a website. And a few might even take the deal. But you're right. This is the kind of cash I could come up with as a grad student. There's no point in taking the money if I believe the idea has wheels. If that's the kind of bad businesmen they're after, well then maybe Morris hasn't quite fully recovered from prison.
Graham selling tactic in the past was that it was closer in nature to an internship in entrepreneurship. It'd have to be, if the net value is four hundred thousand. But the idea of using it as an IQ test has some (perhaps unintentional) merit: don't turn away a guy with money interested in funding your ideas, negiotiate.
I'm pretty sure the first thing to do when measuring false positives is to count them. Maybe that paper has some insight on how to do that without looking at them, but I'm still waiting for the
And, you do realize that Graham's paper specifically cited headers as a reason Bayesian techniques in the past had failed, right? I'm not sure how timing etc plays into it; I get valid emails at all times of the day -- some students are up as late as I am, some are early risers, and sometimes I get mail from my boss during the day. Really, it all depends on the person. If your use doesn't engage the public much, you could probably get away with white lists. If you're part of a large company, you'll have access to a lot of email for known good and labelled junk mail. By the amount of mail I get with no obvious purpose, I surmise that some statistical analysis tools are working.
Actually, thats one of the things that a VP could probably promise without severe problems for engineering. I guess you'd probably have to filter out the curse words in comments though...
So many comments posted thus far, and nothing about Affero. You can't discuss what free software in a "WebOS" (what a horrible term) means without bringing this up. The fundamental point is that the GPL allows anyone modification without no restrictions or obligations as long as you don't distribute the software to other people. Distribution is generally interpreted as different than running it. Basically, if you GPL a PHP driven website, anyone can take it, modify it, give other users access to "running" it without distributing any changes you've made.
As best I can tell, the Affero license addresses this, by a clause partnered with specific functionality in the program. 'If this software came with functionality to give source code to users, you may not remove it.' or something to that effect. I've already seen one or two sites that have decided that it didn't apply to them, and the development community behind it sounded like "so what?" Which is fine I suppose, but it feels strange to me that they were able to make a site for a client based largely on existing OAGPL'd code, integrate the code with existing technology(that is also widely used and would be interesting to many people), fix bugs, and then turn around and declare yourself unable and unrequired to fulfill the obligations imposed on you by the agreement. This is compounded by the fact that generally its much harder to tell when a site is appropriating OAGPL'd code. With traditional software, strings will usually catch hidden strings, debugging statements etc. Web software only has the output to look at.
I am somewhat comforted by the fact their site is already out of date, so either they or their client will be facing increasing costs in maintaining the site.