And you'll identify these e-mail servers how? By hostname? (Domain stealing, DNS poisoning, DNS injection) By IP address? (Fake IP headers + source routing, Router table poisoning, Zombies on legit servers, Zombies on any machine between legit server and target) By mail headers? (Zombies anywhere)
And you guarantee inclusion of legit traffic from mobile sources, how? You don't know what IP address or ISP will be used. What about legit mailing lists, where the originator is indeterminate?
X.400 provides much better authentication, and offers an API for repudiation, but if that's what people really wanted, we'd be using it. Or maybe everyone would use SMTP-over-SSL where client-side and server-side certificates were validated. We don't use them because people need the privacy, anonymity and flexibility of the existing system, although I'd argue almost anything is technically superior to the existing system.
In the end, although a totally secure option should exist, an insecure option should also exist that is controlled by policy rather than technology, and that ultimately means laws.
An alternative would be to declare corporations as privately-owned governments. This would eliminate the need to handle the complexities of imposing the legal constraints imposed on individuals by imposing the legal constraints of the Constitution.
Another option is to abolish all rights for corporations and require that either a declared individual in the corporation represents the company for all rights and penalties, or that corporations are merely collections of individuals and that the individuals have wholly independent rights and responsibilities.
All of these approaches have drawbacks, as well as benefits. Some day, society might even figure out if any are any good.
You write a driver for the device that you want, then you write a shim or an abstraction layer for the kernel API. Or you write the driver for a hypervisor and write dummy drivers for the kernel. Or you write user-mode drivers that work through the kernel. Or, in the case of mice, keyboards and graphics cards, you code for GGI. Or you use a driver-development package to insert the kernel-specific code. Or a million other options.
I won't believe it until Comcast runs a 100 Gb/s link to my apartment for me to try out. For free. After that, I'll be happy to recommend them to Slashdot users and anyone else they want me to promote to. Hey, everyone has their price!:) (Now to see if the Comcast execs will take my bait...)
You could move to asychronous processors and eliminate the need to worry about timings issues. Or you could have RAM that worked faster than the CPU cores and interleave the cores. Or you could map 1000 virtual cores onto Nx1000 physical processing elements, and have the hypervisor schedule the vcores such that the I/O bandwidth was always optimal. Or each core could be given so much local L1/L2 that main memory accesses were infrequent enough for conflicts to not arise. Or you could provide each core with local "main memory", a-la the Transputer. There are probably hundreds more solutions to this problem. When the number of ways to implement a system outnumber the number of systems likely to be sold (at least within the next 20 years, anyway), I don't think there'll be a desperate craving for one specific technology.
I'm much more concerned with the fact that chip companies go for what can be made quickly and sold fast, whether or not it's any good. (Hence Intel rushing Itanium into production before it actually worked, Sun adding floating-point very late into the *Sparcs, why multicores never got beyond 4x4s even though single-core processors can go 16-way, why GPUs were so late in the game even though offloading had been mastered decades earlier, why multithreading and deep-pipelining processors were abandoned by Intel, and so on.)
If companies could afford the delay to get things technologically right and then to price them low enough for the market to handle, we'd have seen Moore's Law abandoned by now... for being too conservative. Of course, this isn't realistic. Companies do not have infinite money for research, development, testing and high discounts. And because it's not realistic, it is inevitable that short-cuts will be taken that produce flawed, sub-optimal products, whether or not it's obvious that such products are unnecessary and a distraction.
You see, they're also developing a parallel technology for detecting Microsoft bias in an article by counting the number of paperclips linking to it. Linux bias by herring-bone count is expected to be announced shortly.
The Micromouse tournaments have had systems that can navigate unfamiliar and changeable environments for something like 30+ years. Robot Table-Tennis has been going on for well over 15 years now and competitors can aim at (and hit) very dynamic targets. Robot Soccer is progressing to the point where multiple machines can target an individual object in an environment of moving targets. Open Source code available from NASA simplifies the development of mission-oriented robot devices. Open Source code available from various other groups simplifies the development of autonomous vehicles.
Given where student robotics already is, and given the software availability, what more can you possibly need for a perfectly viable autonomous robot league for Battlebots?
If you look at scientific open-source software for netcdf and other parallel data management systems, a very substantial portion is for climate and weather modelling. High-energy physics is another major area for open source. Computational fluid dynamics is also popular, but most of the code requires a posted and faxed agreement that the source not be opened to hostile countries. I would expect Linux and the *BSDs to be in widespread use in all three areas, simply because anyone who needs that level of control over the system -or- who can't afford the tens/hundreds of thousands of dollars to buy commercial packages to do the same thing is going to use Linux or a *BSD, with Linux dominating on clusters.
...that a few years back, it was widely reported that someone was able to get Verisign to provide them with Microsoft's server-side certificates. I don't think it was ever really said how long the rogue copies were in the wild before action was taken. I can't see banks ever admitting to their security being compromised in such a way.
Having said that, phishing is still described as highly successful, social engineering is usually described as highly effective, and there are probably a few places where wardiallers can access unsecured lines to financial computer systems. Besides which, most people are lousy at securing their home computers, so I suspect that most attackers aiming at individuals would use keystroke loggers embedded in viruses or placed on a machine after using some public script. Most homes are not secure, either, so you could do just about as well with a powerful receiver and decoder for the output of computer monitors. People are often very bad about destroying sensitive documents, which is why dumpster diving is also said to be very effective.
In short, why would anyone hang around on the edge and wait for a single victim, when that would be slower and riskier than any of the alternatives, and produce just one possibility versus hundreds or thousands for the others?
For years, the aliens that land in Area 51 have had to make do with cheap, low-grade bobble antennae sticking out of their heads, which is very disconcerting when they run for Congress. The job of a Government antenna engineer is to design antenna that better blend in with the Congressmen's hairstyles, pointed ears, etc. This is why you don't see them any more.
If you want to do things the hard way, you could try: OpenWRT. If you prefer the easy way, use a tunnel broker. Then only your machine needs to support IPv6, your router doesn't.
I believe IPv5 is formally defined as TUBA, although another poster mentions realtime connections (which wouldn't seem to be an IP version, per se, but the layer running over IP) and POTS (which I'm damn sure is a layer 1 to layer 2 concept). There's also an IPv7. As far as I know, no TUBA drivers exist for Linux (damn shame) and I'm very certain no services (eg: DNS) exist for it.
(When it comes to Linux support for protocols, it's a popular platform for early developers, but maintenance can be an issue. enSKIP and SGI's STP code are abandonware, the real-time network driver for RTAI is infrequently updated, and the GAMMA Active Messages driver is seriously stalled in a number of areas. Many updates to Web100 have just been kernel increment updates, not bugfixes or added features. I don't recall seeing any support for VIA - which is fair enough, given it's dead - or iWarp. Linux' QoS supports RED, but neglected BLUE, GREEN, BLACK, WHITE and PURPLE the last time I looked.)
It ends up in the most extraordinary, improbable circumstances. It's hunted by fudd. It survives anything thrown at it. Nyahhhhhhhhhh. What's up,*.doc?
More of a refundable deposit, then, rather than a cost. Hmmm. Might work. You could also have a sliding scale on refunds depending on actual errors noted for correction during peer review, with some percentage of that retained going to the reviewer finding the errors, to give incentive for more thorough review.
The problem I see with this is that incentive schemes don't work as advertised. From the FSF's article on reward systems to the OLPC founder's comments on such methods, we find that people divert more and more of their attention towards the reward itself, away from whatever the reward was intended for in the first place.
In the case of articles accepted for publication, for example, you'd likely see a move away from a normal distribution of quality over the entire possible spectrum to a much narrower focus on articles just good enough to be accepted - or, in the case of my suggested modification, just good enough to make submitting articles cheap. You'd almost certainly lose the higher-end, high-quality stuff, as the rewards don't cover that and the emphasis has moved to the rewards.
Eliminating the cost of submission would produce more submitted papers, yes, requiring more extensive review practices and possibly generating either more publications or substantially larger publications, which would cost more but generate relatively little extra income. It could also encourage people to get peer reviewers to do more of the mundane work (spellchecking, grammar checking, etc) that authors generally hate to do. On the other hand, improving the circulation of ideas is part of the point in having journals in the first place. There's a decline on returns, as always, so improving circulation beyond a certain point offers no real benefits. But how do you define that point? By immediate cost? By the long-term cost/benefit to academia in general? And if the latter, how would you even measure that?
One of the more "orthodox" ways of measuring benefit is by citation. The more cited a paper is, the greater the impact. However, that's not necessarily true, as one paper refuting another must cite the paper being refuted and any paper based on that refutation is very likely to cite the original. On the other hand, one could argue that refuting one paper with another means the first paper has generated interest and research in a subject, thus still having a benefit. If we work on this strategy, then instead of refunding the cost of submission, one could imagine a royalty-like payment system where an author is payed per citation. There's a risk, though, of people gaming the system by colleagues citing each other not for quality reasons but to generate revenue.
People are so good at finding loopholes that I don't think any system is going to achieve the intended results. There's going to be some unintended negative consequence. The question is, what negative consequences can we live with most easily?
You've got to be careful on the specious. Whales are descended from land animals and you can get huge physical changes with very small genetic differences. It is also possible to GM genetically distant mammals to grow human organs - although I'm suspicious about potential genetic contamination and abnormal stresses. Hypothetically, provided whales have not become too distant, it should be possible to get relatively close land animals to at least grow "whale meat", and provided the changes are (at the genetic level) superficial, cross-species fertilization would not be impossible.
Herein lies the problem. The first, which would have eliminated the need for a whaling fleet, was not attempted, as far as I can tell. No great surprise, given who financed the research. The second is extremely unlikely - whales are ancient, and the genetic differences with their closest land relatives are significant - and don't apply to cows. If they'd worked with hippos, I might be inclined to believe that they took their own research seriously. They're still way too distant for it to be remotely credible, but given it's the closest land relative going, it would at least make some sense. Sure, cows are easier to obtain, but you need to be in a truly Dilbertesque situation, incredibly stupid or believe everyone else to be incredibly stupid, to go in that direction.
(Sadly, many people are incredibly stupid when it comes to bad science, which is why there's so much out there any why it's so profitable. I suggest reviewing the animated Dilbert episode on Chronic Cubicle Syndrome for further information on credulity. It's not restricted to any group of people - plenty of people regarded as geniuses believed incredibly stupid things. Intelligence provides an extra tool to filter out nonsense, but it must be applied for it to work and it is easily negated by flawed assumptions and preconcieved notions. Problem is, as the cartoon notes, it's impossible to investigate everything, which means everyone works from flawed assumptions and preconcieved notions.)
In the original series, the computer was needed for any telepathy over extreme distances. Additionally, the characters needed to link up to the computer directly "to boost the signal". Routing wireless-to-wireless through the computer wasn't possible, making it very similar to the way wireless access points actually operate. Over insterstellar distances, both telepathy and jaunting not only required the computer but also a specialized communications system that operated distinctly from the regular mechanisms. Again, that could be likened to the difference between a terrestrial FM broadcast and a deep-space mission-control-to-probe communications hookup.
Oh, I agree it's "unrealistic" and out-there. Most sci-fi is. However, there are a respectable number of real-world parallels - enough, I think, to say that they got the complexities involved covered. The mechanisms, scale, sophistication, etc, are totally fictional and implausible - though mostly necessary for the plots and not as excessive as some sci-fi out there. The premise that there was no universal solution but rather lots of very specialized ones was probably the thing that made the series respectable.
As much as I dislike his oil-from-volcanos and continuous-creation ideas, he did come up with some interesting sci-fi, especially in the area you're talking about. One of his stories, "The Black Cloud", hypothesises beings with immense bandwidth between individuals and discusses at length the impact of bandwidth on individualism and communications. It also suggests the impact of very high-bandwidth communication from such an individual to the human mind (the human mind might initially be taken over but would rapidly fry).
Since the transceiver would be relatively small in comparison to most parts of the brain, there is no reason to assume only the speech centre would have such a system. Indeed, as the complexity of the brain is a function of the complexity of the interconnects, there would be an advantage in an organism where all parts of the brain were electrically isolated (less risk of seizures going non-local) but totally interconnected (you couldn't have that many interconnects physically), provided there was sufficient bandwidth. If you can also dispense with any neurons just there for switching and routing purposes, you would be able to have more processing neurons for the same space and heat output. This allows all senses, memories, etc, to be linked and would be closer to Isaac Asimov's "Gaia" than to conventional notions of telepathy in science-fiction.
(By the way, The Tomorrow People distinguishes between telepathy and telekinesis and is generally closer to the concepts described by ESP enthusiasts than, say, The X-Men comics. Sapphire And Steel does an even better job of it. If you don't mind slow-paced, plot-heavy sci-fi, then I would suggest trying either. It's not necessary to live on a monoculture diet. There's many flavours of sci-fi out there.)
Slashdot Mirror
And you guarantee inclusion of legit traffic from mobile sources, how? You don't know what IP address or ISP will be used. What about legit mailing lists, where the originator is indeterminate?
X.400 provides much better authentication, and offers an API for repudiation, but if that's what people really wanted, we'd be using it. Or maybe everyone would use SMTP-over-SSL where client-side and server-side certificates were validated. We don't use them because people need the privacy, anonymity and flexibility of the existing system, although I'd argue almost anything is technically superior to the existing system.
In the end, although a totally secure option should exist, an insecure option should also exist that is controlled by policy rather than technology, and that ultimately means laws.
Another option is to abolish all rights for corporations and require that either a declared individual in the corporation represents the company for all rights and penalties, or that corporations are merely collections of individuals and that the individuals have wholly independent rights and responsibilities.
All of these approaches have drawbacks, as well as benefits. Some day, society might even figure out if any are any good.
"...as if a billion Pinky jokes wooshed overhead and fell silent."
You write a driver for the device that you want, then you write a shim or an abstraction layer for the kernel API. Or you write the driver for a hypervisor and write dummy drivers for the kernel. Or you write user-mode drivers that work through the kernel. Or, in the case of mice, keyboards and graphics cards, you code for GGI. Or you use a driver-development package to insert the kernel-specific code. Or a million other options.
Floccinoccinilihilipilification.
"Are you thinking what I'm thinking?"
"I think so Brain, but what would the NSA do with ten billion smoked herrings?"
I won't believe it until Comcast runs a 100 Gb/s link to my apartment for me to try out. For free. After that, I'll be happy to recommend them to Slashdot users and anyone else they want me to promote to. Hey, everyone has their price! :) (Now to see if the Comcast execs will take my bait...)
I'm much more concerned with the fact that chip companies go for what can be made quickly and sold fast, whether or not it's any good. (Hence Intel rushing Itanium into production before it actually worked, Sun adding floating-point very late into the *Sparcs, why multicores never got beyond 4x4s even though single-core processors can go 16-way, why GPUs were so late in the game even though offloading had been mastered decades earlier, why multithreading and deep-pipelining processors were abandoned by Intel, and so on.)
If companies could afford the delay to get things technologically right and then to price them low enough for the market to handle, we'd have seen Moore's Law abandoned by now... for being too conservative. Of course, this isn't realistic. Companies do not have infinite money for research, development, testing and high discounts. And because it's not realistic, it is inevitable that short-cuts will be taken that produce flawed, sub-optimal products, whether or not it's obvious that such products are unnecessary and a distraction.
You see, they're also developing a parallel technology for detecting Microsoft bias in an article by counting the number of paperclips linking to it. Linux bias by herring-bone count is expected to be announced shortly.
You mean we should have Warhammer: 4000 volts?
Given where student robotics already is, and given the software availability, what more can you possibly need for a perfectly viable autonomous robot league for Battlebots?
If you look at scientific open-source software for netcdf and other parallel data management systems, a very substantial portion is for climate and weather modelling. High-energy physics is another major area for open source. Computational fluid dynamics is also popular, but most of the code requires a posted and faxed agreement that the source not be opened to hostile countries. I would expect Linux and the *BSDs to be in widespread use in all three areas, simply because anyone who needs that level of control over the system -or- who can't afford the tens/hundreds of thousands of dollars to buy commercial packages to do the same thing is going to use Linux or a *BSD, with Linux dominating on clusters.
Having said that, phishing is still described as highly successful, social engineering is usually described as highly effective, and there are probably a few places where wardiallers can access unsecured lines to financial computer systems. Besides which, most people are lousy at securing their home computers, so I suspect that most attackers aiming at individuals would use keystroke loggers embedded in viruses or placed on a machine after using some public script. Most homes are not secure, either, so you could do just about as well with a powerful receiver and decoder for the output of computer monitors. People are often very bad about destroying sensitive documents, which is why dumpster diving is also said to be very effective.
In short, why would anyone hang around on the edge and wait for a single victim, when that would be slower and riskier than any of the alternatives, and produce just one possibility versus hundreds or thousands for the others?
For years, the aliens that land in Area 51 have had to make do with cheap, low-grade bobble antennae sticking out of their heads, which is very disconcerting when they run for Congress. The job of a Government antenna engineer is to design antenna that better blend in with the Congressmen's hairstyles, pointed ears, etc. This is why you don't see them any more.
If you want to do things the hard way, you could try: OpenWRT. If you prefer the easy way, use a tunnel broker. Then only your machine needs to support IPv6, your router doesn't.
(When it comes to Linux support for protocols, it's a popular platform for early developers, but maintenance can be an issue. enSKIP and SGI's STP code are abandonware, the real-time network driver for RTAI is infrequently updated, and the GAMMA Active Messages driver is seriously stalled in a number of areas. Many updates to Web100 have just been kernel increment updates, not bugfixes or added features. I don't recall seeing any support for VIA - which is fair enough, given it's dead - or iWarp. Linux' QoS supports RED, but neglected BLUE, GREEN, BLACK, WHITE and PURPLE the last time I looked.)
It ends up in the most extraordinary, improbable circumstances. It's hunted by fudd. It survives anything thrown at it. Nyahhhhhhhhhh. What's up,*.doc?
The problem I see with this is that incentive schemes don't work as advertised. From the FSF's article on reward systems to the OLPC founder's comments on such methods, we find that people divert more and more of their attention towards the reward itself, away from whatever the reward was intended for in the first place.
In the case of articles accepted for publication, for example, you'd likely see a move away from a normal distribution of quality over the entire possible spectrum to a much narrower focus on articles just good enough to be accepted - or, in the case of my suggested modification, just good enough to make submitting articles cheap. You'd almost certainly lose the higher-end, high-quality stuff, as the rewards don't cover that and the emphasis has moved to the rewards.
Eliminating the cost of submission would produce more submitted papers, yes, requiring more extensive review practices and possibly generating either more publications or substantially larger publications, which would cost more but generate relatively little extra income. It could also encourage people to get peer reviewers to do more of the mundane work (spellchecking, grammar checking, etc) that authors generally hate to do. On the other hand, improving the circulation of ideas is part of the point in having journals in the first place. There's a decline on returns, as always, so improving circulation beyond a certain point offers no real benefits. But how do you define that point? By immediate cost? By the long-term cost/benefit to academia in general? And if the latter, how would you even measure that?
One of the more "orthodox" ways of measuring benefit is by citation. The more cited a paper is, the greater the impact. However, that's not necessarily true, as one paper refuting another must cite the paper being refuted and any paper based on that refutation is very likely to cite the original. On the other hand, one could argue that refuting one paper with another means the first paper has generated interest and research in a subject, thus still having a benefit. If we work on this strategy, then instead of refunding the cost of submission, one could imagine a royalty-like payment system where an author is payed per citation. There's a risk, though, of people gaming the system by colleagues citing each other not for quality reasons but to generate revenue.
People are so good at finding loopholes that I don't think any system is going to achieve the intended results. There's going to be some unintended negative consequence. The question is, what negative consequences can we live with most easily?
It's i.
Wrong exponent. Turns out it's closer to rabbits.
Herein lies the problem. The first, which would have eliminated the need for a whaling fleet, was not attempted, as far as I can tell. No great surprise, given who financed the research. The second is extremely unlikely - whales are ancient, and the genetic differences with their closest land relatives are significant - and don't apply to cows. If they'd worked with hippos, I might be inclined to believe that they took their own research seriously. They're still way too distant for it to be remotely credible, but given it's the closest land relative going, it would at least make some sense. Sure, cows are easier to obtain, but you need to be in a truly Dilbertesque situation, incredibly stupid or believe everyone else to be incredibly stupid, to go in that direction.
(Sadly, many people are incredibly stupid when it comes to bad science, which is why there's so much out there any why it's so profitable. I suggest reviewing the animated Dilbert episode on Chronic Cubicle Syndrome for further information on credulity. It's not restricted to any group of people - plenty of people regarded as geniuses believed incredibly stupid things. Intelligence provides an extra tool to filter out nonsense, but it must be applied for it to work and it is easily negated by flawed assumptions and preconcieved notions. Problem is, as the cartoon notes, it's impossible to investigate everything, which means everyone works from flawed assumptions and preconcieved notions.)
Oh, I agree it's "unrealistic" and out-there. Most sci-fi is. However, there are a respectable number of real-world parallels - enough, I think, to say that they got the complexities involved covered. The mechanisms, scale, sophistication, etc, are totally fictional and implausible - though mostly necessary for the plots and not as excessive as some sci-fi out there. The premise that there was no universal solution but rather lots of very specialized ones was probably the thing that made the series respectable.
As much as I dislike his oil-from-volcanos and continuous-creation ideas, he did come up with some interesting sci-fi, especially in the area you're talking about. One of his stories, "The Black Cloud", hypothesises beings with immense bandwidth between individuals and discusses at length the impact of bandwidth on individualism and communications. It also suggests the impact of very high-bandwidth communication from such an individual to the human mind (the human mind might initially be taken over but would rapidly fry).
(By the way, The Tomorrow People distinguishes between telepathy and telekinesis and is generally closer to the concepts described by ESP enthusiasts than, say, The X-Men comics. Sapphire And Steel does an even better job of it. If you don't mind slow-paced, plot-heavy sci-fi, then I would suggest trying either. It's not necessary to live on a monoculture diet. There's many flavours of sci-fi out there.)
Keith and Orville are still touring?