There's more to x86 than clone boxes
on
Phoenix's BIOS Roadmap
·
· Score: 2, Informative
I suppose you've never tried running Linux on any non-clone x86 box right? The BIOSes on the HP/Compaq blade and DL-series enterprise servers are pretty advanced. While the iLO (integrated lights out) feature on the Compaq BIOS is not perfect (it's too damn slow for one, especially when your console goes to graphics mode), it almost gives you a fully functional console over a 100 MBps Ethernet link. In fact, this is the only way to access the console on a BL20p or similar blade server. Basically the only thing you can't do with iLO that you can do on the physical console is insert and eject removable media. Yeah, with this feature you can network boot into single user mode when your disks and filesystems go bad and do hardware diagnostics too.
No, this isn't meant to be anything even remotely resembling these remote management features. Phoenix is seriously in bed with Microsoft, and well, this their monster offspring is meant to be the first step toward Palladium or NGSCB or whatever the hell Microsoft calls it now.
True, but with the BSA breathing down your neck, that's not such an attractive option. And besides, if you bothered to read the article, it says that one of the Israeli government's main concerns had to do with editing documents in Hebrew text, which is difficult to do with MS Office and is not something particularly high on Microsoft's priorities. They couldn't give a rat's ass about all of the other "features" that new versions of Word and Office had. The key feature they were interested in is not there. If they can't easily write documents written in their own national language, then what good is it? The version of OpenOffice they'll be using has this type of support.
As I recall, the same thing could have happened around 1996-7 with Iceland, had a viable alternative existed at the time. Microsoft was slow to add Icelandic to Windows and Office 95, despite repeated requests from the Icelandic government. The language eventually made it into Windows 98. Sadly, no viable alternatives to a Windows desktop existed at the time. (Before anyone shouts, I hope everyone remembers what Linux looked like at the time, and whether anyone would let barely computer-literate government workers use it in the state it was back in 1996.).
Internationalization and localization is really something that Free Software does very quickly and effectively, and something that Microsoft is particularly weak at by comparison. Perhaps the use of Linux and Free Software will begin to grow more rapidly in places where i18n and l10n matter a lot.
Jeez, I imagine that we here could ask a lot of really pointed questions, that could shed some light on this mess from someone who built the much-maligned company and until it became so hated was at its helm.
I misread the article, it would seem. I thought they were referring to feature size when it read 'the 300 mm fab will begin operation...' Sorry about that.
Damn. 300 mm = 0.3 m, or in imperial units you folks in the US are more used to, that's a minimum feature size of 10.8 inches. I'd hate to see one of the GIANT chips they make in that fab!;)
Even if it's a misprint for 300 um (common thing) still is too big. That's 0.3 millimiters, still much larger than all but the most primitive integrated circuits.
Reminds me of a fortune cookie
on
Mafia Tech Support
·
· Score: 5, Funny
From one of the BSD Games fortune cookies:
A gangster assembled an engineer, a chemist, and a physicist. He explained that he was entering a horse in a race the following week and the three
assembled guys had the job of assuring that the
gangster's horse would win. They were to reconvene
the day before the race to tell the gangster how
they each propose to ensure a win. When they
reconvened the gangster started with the engineer:
Gangster: OK, Mr. engineer, what have you got?
Engineer: Well, I've invented a way to weave metallic threads into the saddle blanket so that
they will act as the plates of a battery and provide electrical shock to the horse.
G: That's very good! But let's hear from the chemist.
Chemist: I've synthesized a powerful stimulant that dissolves into simple blood sugars after ten minutes and therefore cannot be detected in post-race tests.
G: Excellent, excellent! But I want to hear from the physicist before I decide what to do. Physicist?
Physicist: Well, first consider a spherical horse in simple harmonic motion...
I wonder what a computer scientist would be up to?;)
Interesting but that sounds kind of scary. Once you attempt to find optimizations that actually try to find other equivalent algorithms, you start treading into the dangerous realm of undecidability and Turing completeness. Consider what optimizing Turing-complete systems like partial recursive functions or their equivalent Turing machines entails. The first, and most important thing you need to do is decide whether or not some Turing machine runs faster (or more generally has a better performance metric) than some other Turing machine. You run it, and do your measurements. But then, how do you know whether there might exist some input for which your TM's will eventually enter an infinite loop? You can't know that, it's undecidable. It's the halting problem, and I imagine you'll run into this all the time when you try to do this level of optimization.
Any system that tried to optimize recursive functions or any other equivalent Turing-complete formalism according to some optimization metric (e.g. space or time complexity) is at most attempting to apply heuristics to decide the undecidable. I seriously doubt that such an approach would ever be useful enough to clean up after sloppy "programmers" who never bothered to study proper algorithms and data structures, and recode a quicksort where a bubble sort was originally written. At best, I imagine it would produce marginal improvement while increasing compile time dramatically, and would be totally impractical for any sort of useful work. I seriously doubt such techniques were ever used in anything but research compilers for toy languages.
Well, you're absolutely correct, but you've veered a bit from the mark. It seems that the advocates of biometric identification are not interested in using biometrics to augment existing security procedures, but to replace these procedures, and they seem to be gushing that biometric "authentication" is a silver bullet, or something very close to it. Few banks, and no military or intelligence agency in their right mind would be so foolish as to believe that. If you've taken the time to even read the article I linked to, you'd see that Schneier isn't advocating that we not use biometrics at all, but that we not treat them as keys. They have their uses, especially when combined with real keys and other authentication schemes, but to use them alone for authentication isn't generally a good idea.
Granted, sole biometric identification is better than some present identification methods, and could replace them in those areas, where the risk is not high enough to justify the use of more expensive and complex procedures, but does it give sufficient security for many of the domains to which it is being applied? I think not. Biometrics raise the bar a bit, but not high enough to be used on their own for many of the applications to which people want to put them to use.
By the way, you're highly out of date about Schneier's present views on security. If you've taken the time to read his many writings over the years, you'll see how much his attitude towards security has changed since the days of Applied Cryptography, where he naively talks about "protecting ourselves with mathematics." His most famous maxim is now "Security is a process, not a product," and he keeps constantly talking about how security is all about risk management, not risk avoidance. Exactly what you're saying, isn't it? Have a look at Secrets and Lies and the Crypto-Gram archive sometime.
I don't know why all of these so-called "security experts" keep on advocating biometrics with little or no understanding of their real properties, much less how they should be properly used. Biometrics can be used as unique identifiers, but biometrics are not secrets. They can provide a unique identifier in an already trusted environment, but alone they cannot be used for authentication, which is what so many of these "experts" are ready to do. If I steal your fingerprint using any of the simple yet effective techniques (none of which require me to cut off your finger) described by Ton van der Putte, it can't be un-stolen, and nobody will be able to give you a "replacement" fingerprint.
A quote that iluustrates this naivete from the USA Today article: "Biometrics is one way to really identify the customer you're dealing with," he [Steve Vallance] says. What a foolish, naive statement. Alone, biometrics cannot really identify anybody.
I really can't do any better than point people out to an article in yet another issue of Crypto-Gram, which first came out five years ago: Biometrics: Truths and Fictions.
Aye, shame, true shame. Fortunately I was able to find a copy of the original journal in our university library. The only other link I found was from the Turing Digital Archive, but it's all scanned images, scarcely better than the IE version. Anyhow, it's here.
"On computable numbers, with an application to the Entscheidungsproblem"" is unarguably the paper that began the field of computer science as we understand it today. Here we have the first descriptions of universal computing devices, Turing machines, which eventually led to the idea of universal stored-program digital computers. The paper even seems to describe, in what is unarguably the first ever conceptual programming language, a form of continuation passing style in the form of the "skeleton tables" Turing used to abbreviate his Turing machine designs. It's also relatively easy reading compared to many other scientific papers I've seen.
Along with this we might also include Alonzo Church's 1941 paper "The Calculi of Lambda Abstraction" (which sadly does not appear to be anywhere online), where the lambda calculus, the basis for all functional programming languages, is first described.
The original 16-bit color mode of the EGA cards and VGA cards must have been designed by somebody who was high on crack. You can't get at the pixel memory without setting up a bewildering array of registers that control mandatory and mostly non-useful logic operations on your bits. The memory is accessed as 4 independent planes, so you have to unnaturally slice every pixel up into individual bits and have a PhD in boolean logic to get them on the screen as you intended. It easily could take a newbie a whole day of reading manuals and hacking before they could get a single white dot on the screen.
Ah, I remember those days, and yes, I remember the reasons, which you seem to have forgotten. I assume you're talking about the old VGA mode 12h 640x480 16-color modes and the similar EGA 640x350 16-color mode. The reason for the somewhat absurd (in this day anyhow) system for getting pixels onto the screen is the 640k limit that existed in those days, and to be more precise, all of video memory was fixed to remain in two 64k segments beginning at physical address A0000h. Think about it. (640*480)/2 (because 16 colors == 4-bit color) is 153,600 bytes. How the heck are you going to squeeze that much video memory into two segments of 128K total? They could have used paging, like EMS in those days, but that would mean that scrolling and a lot of useful features would be excruciatingly slow as your blitter would need to switch banks every frame of scroll. Instead, we got bitplanes and masking and all that, which allowed many of these operations to happen very quickly. Granted, it's totally unintuitive to a newcomer, and reeks totally of a kluge, but what PC hardware in those days was a simple thing to program?
And these kluges also inadvertently served to create the undocumented Mode X 320x240 256-color mode which was nirvana indeed for PC animation enthusiasts (like Mike Abrash) in the early nineties...:)
Oh how I think we would all wish it was just crack! If it were, then McBride and Boies and the whole lot would have OD'ed by now and we'd just drag their corpses out of their office in Utah and bury them at the crossroads with a stake through the heart.
Unfortunately, they seem to be high as a kite on something that you can't OD on that easily.:(
AFAIK, the classical channel can be eavesdropped upon, and the eavesdropper will gain no information about the key being exchanged, so it doesn't need to be absolutely secure. It just needs to be safe from jamming, because if you jam the classical channel the two parties have no way of discussing the level of interference on the quantum channel, and the protocol breaks down. As has been mentioned repeatedly, it is impossible to passively eavesdrop on the quantum channel and gain useful information without being detected by the two communicants.
To be more exact, I think a more valid description of what's up with this device is that it's too far ahead of its time and is unripe. The ideas have been proved in the lab, but to make one for consumption by the general public is premature.
Well, some modern protocols (e.g. Kerberos) still require you to exchange symmetric crypto keys beforehand. I suppose these could be useful in situations similar to those. You could in theory build a quantum crypto-based Kerberos variant using these protocols.
Right. But well, some people have attempted to develop quantum authentication protocolslike this one, this one, and this one. Dunno if the device in question does any of them, or even if any of them are actually practical to use with today's technology. If the device in question doesn't use quantum authentication of some kind, well, they're selling snake oil, but I wouldn't dismiss the whole concept of quantum cryptography out of hand totally the way you seem so ready to.
By the way, I've found a link to the paper I mention above. It's the paper by Charles Bennett, Francois Bessette, Gilles Brassard, Louis Salvail, and John Smolin, "Experimental Quantum Cryptography" (Citeseer link).
I imagine they're using the protocol, or a close variation thereof, of the one described by Charles H. Bennett, Gilles Brassard, Louis Salvail, and John Smolin in their paper "Experimental Quantum Cryptography" (Citeseer link). They use the quantum channel to allow Alice and Bob to negotiate a random key known only to them, which cannot be eavesdropped upon, and also allows them to discover any eavesdropper attempting to obtain their key. You can't use a quantum channel to transmit real information, but you can use it to negotiate a perfectly random key between two communicants without anyone but the two parties knowing it. In other words, the quantum cryptographic protocols are just like Diffie-Hellman and other key exchange protocols we all know well, only they provide a physical proof that according to the laws of physics as we understand it today, eavesdropping is impossible.
I don't think amplification is going to do you good. If this implements the first quantum key exchange protocol, designed by Gilles Brassard and others (notably this is the easiest one to implement using present-day technology so far), they actually work by transmitting single photons one at a time, with the transmitter whom we shall call Alice (the one generating the key to be used for your subsequent symmetric cryptography) controlling the polarization. The recipient of the photons (Bob) then measures the polarization, then depending on whether the answer was correct or not decide what the value of the key bits should be. Doesn't matter if Eve is able to eavesdrop on the classical channel where they're communicating about polarization methods, Brassard and his colleagues showed that she will not be able to gain any information about the key Bob is receiving from there. If Eve tries to tap into the photon stream she'll be noticed by Alice and Bob because of the way her measurements are affecting Bob's measurements... Anyone feel free to correct me on this, I don't have Brassard's paper with me on hand at the moment and can't find it on the web, so I've been just going by it from memory.
(in case you haven't yet noticed, this quantum cryptographic protocol, like most the others I've come across, is a key exchange protocol that works as a replacement for Diffie-Hellman or something similar).
The upshot is that the fiber cable here will be from the point of view of conventional fiber optic equipment a dark fiber. Light intensity is so low that only specialized equipment can detect it. Anything else inside the fiber, be it an amplifier or a DWDM switch, or whatever else, won't recognize the transmission and may do something totally unexpected that breaks the protocol. Bob could always be reading the wrong polarization, and hence Alice would have to keep discarding key bits. However, apparently there are excellent fiber cables that can go for several kilometers and still maintain the kind of integrity required for the protocol (IIRC, Brassard mentions somewhere that as early as 1996 someone actually built a real setup employing his protocol, over a 15km fiber optic cable).
Other quantum key exchange protocols I've come across are noticeably similar, and suffer from the many of the same limitations.
The only use that I can see for this is if you're someone with the resources to lay your own fiber around, say if you have a large complex covering several acres, and are more paranoid than the NSA. There is a market, I think, but that market looks more like the military and intelligence agencies of large industrialized nations. Building such an infrastructure could cost billions, even if you all you wanted to do was securely interconnect several dispersed branch offices in one large city...
If you can already transmit random bits by means of quantum cryptography, why not go the whole nine yards and exchange random keys that are as long as your entire message? The papers I've read on quantum cryptography all employ it as a method for the secure exchange of one-time pads. The only reasons I can think of are that 1. the effective bandwidth of the quantum channel is too low and/or 2. they're having problems setting up a reliable and fast source of real random numbers on their device (this is actually a lot harder to do than most people think).
Do the words "Cyber-Armageddon" mean anything to you? I imagine what some of the things they'll try to simulate are the really monstrous distributed denial of service attacks that would cripple the entire Internet for days and cause untold amounts of damage and inconvenience for millions all over the world. Things that would make a major slashdotting look like a mild hiccup. If they did that on the real Internet, the damage it would cause worldwide would be worth a helluvalot more than a piddly US$5.5 million.
I imagine they'll be dealing more with network structural problems than cracking problems. They'd also be doing experiments on some conjectures like the Warhol worms and flash worms we've heard about that can infect 60% or more of all the hosts on the Internet in under 15 minutes. If they did such an experiment on the real Internet, that would again cause a lot of major disruption, especially if they made a mistake in the worm's programming, and that's not exactly an unknown occurrence... The Great Worm itself had a bug that rendered it far less benign than Robert T. Morris planned. If that happened on one of the world's largest WAN's however, no big deal, it'll just inconvenience the grad students in charge of it, and they'd need to reinstall, which is probably part of what they'll be getting paid to do anyway.
What you're proposing by the way, would be tantamount to condoning acts that the Department of Homeland Security (which appears to be one of the agencies funding this project) has labeled terrorism, and which have been illegal since long before 9/11.
Slashdot Mirror
I suppose you've never tried running Linux on any non-clone x86 box right? The BIOSes on the HP/Compaq blade and DL-series enterprise servers are pretty advanced. While the iLO (integrated lights out) feature on the Compaq BIOS is not perfect (it's too damn slow for one, especially when your console goes to graphics mode), it almost gives you a fully functional console over a 100 MBps Ethernet link. In fact, this is the only way to access the console on a BL20p or similar blade server. Basically the only thing you can't do with iLO that you can do on the physical console is insert and eject removable media. Yeah, with this feature you can network boot into single user mode when your disks and filesystems go bad and do hardware diagnostics too.
No, this isn't meant to be anything even remotely resembling these remote management features. Phoenix is seriously in bed with Microsoft, and well, this their monster offspring is meant to be the first step toward Palladium or NGSCB or whatever the hell Microsoft calls it now.
True, but with the BSA breathing down your neck, that's not such an attractive option. And besides, if you bothered to read the article, it says that one of the Israeli government's main concerns had to do with editing documents in Hebrew text, which is difficult to do with MS Office and is not something particularly high on Microsoft's priorities. They couldn't give a rat's ass about all of the other "features" that new versions of Word and Office had. The key feature they were interested in is not there. If they can't easily write documents written in their own national language, then what good is it? The version of OpenOffice they'll be using has this type of support.
As I recall, the same thing could have happened around 1996-7 with Iceland, had a viable alternative existed at the time. Microsoft was slow to add Icelandic to Windows and Office 95, despite repeated requests from the Icelandic government. The language eventually made it into Windows 98. Sadly, no viable alternatives to a Windows desktop existed at the time. (Before anyone shouts, I hope everyone remembers what Linux looked like at the time, and whether anyone would let barely computer-literate government workers use it in the state it was back in 1996.).
Internationalization and localization is really something that Free Software does very quickly and effectively, and something that Microsoft is particularly weak at by comparison. Perhaps the use of Linux and Free Software will begin to grow more rapidly in places where i18n and l10n matter a lot.
Jeez, I imagine that we here could ask a lot of really pointed questions, that could shed some light on this mess from someone who built the much-maligned company and until it became so hated was at its helm.
I misread the article, it would seem. I thought they were referring to feature size when it read 'the 300 mm fab will begin operation...' Sorry about that.
The article seems to be misleading. Apparently it's wafer size, not feature size, on reading a bit more carefully. ;)
Damn. 300 mm = 0.3 m, or in imperial units you folks in the US are more used to, that's a minimum feature size of 10.8 inches. I'd hate to see one of the GIANT chips they make in that fab! ;)
Even if it's a misprint for 300 um (common thing) still is too big. That's 0.3 millimiters, still much larger than all but the most primitive integrated circuits.
From one of the BSD Games fortune cookies:
A gangster assembled an engineer, a chemist, and a physicist. He explained that he was entering a horse in a race the following week and the three assembled guys had the job of assuring that the gangster's horse would win. They were to reconvene the day before the race to tell the gangster how they each propose to ensure a win. When they reconvened the gangster started with the engineer:
Gangster: OK, Mr. engineer, what have you got?
Engineer: Well, I've invented a way to weave metallic threads into the saddle blanket so that they will act as the plates of a battery and provide electrical shock to the horse.
G: That's very good! But let's hear from the chemist.
Chemist: I've synthesized a powerful stimulant that dissolves into simple blood sugars after ten minutes and therefore cannot be detected in post-race tests.
G: Excellent, excellent! But I want to hear from the physicist before I decide what to do. Physicist?
Physicist: Well, first consider a spherical horse in simple harmonic motion...
I wonder what a computer scientist would be up to? ;)
Interesting but that sounds kind of scary. Once you attempt to find optimizations that actually try to find other equivalent algorithms, you start treading into the dangerous realm of undecidability and Turing completeness. Consider what optimizing Turing-complete systems like partial recursive functions or their equivalent Turing machines entails. The first, and most important thing you need to do is decide whether or not some Turing machine runs faster (or more generally has a better performance metric) than some other Turing machine. You run it, and do your measurements. But then, how do you know whether there might exist some input for which your TM's will eventually enter an infinite loop? You can't know that, it's undecidable. It's the halting problem, and I imagine you'll run into this all the time when you try to do this level of optimization.
Any system that tried to optimize recursive functions or any other equivalent Turing-complete formalism according to some optimization metric (e.g. space or time complexity) is at most attempting to apply heuristics to decide the undecidable. I seriously doubt that such an approach would ever be useful enough to clean up after sloppy "programmers" who never bothered to study proper algorithms and data structures, and recode a quicksort where a bubble sort was originally written. At best, I imagine it would produce marginal improvement while increasing compile time dramatically, and would be totally impractical for any sort of useful work. I seriously doubt such techniques were ever used in anything but research compilers for toy languages.
Well, you're absolutely correct, but you've veered a bit from the mark. It seems that the advocates of biometric identification are not interested in using biometrics to augment existing security procedures, but to replace these procedures, and they seem to be gushing that biometric "authentication" is a silver bullet, or something very close to it. Few banks, and no military or intelligence agency in their right mind would be so foolish as to believe that. If you've taken the time to even read the article I linked to, you'd see that Schneier isn't advocating that we not use biometrics at all, but that we not treat them as keys. They have their uses, especially when combined with real keys and other authentication schemes, but to use them alone for authentication isn't generally a good idea.
Granted, sole biometric identification is better than some present identification methods, and could replace them in those areas, where the risk is not high enough to justify the use of more expensive and complex procedures, but does it give sufficient security for many of the domains to which it is being applied? I think not. Biometrics raise the bar a bit, but not high enough to be used on their own for many of the applications to which people want to put them to use.
By the way, you're highly out of date about Schneier's present views on security. If you've taken the time to read his many writings over the years, you'll see how much his attitude towards security has changed since the days of Applied Cryptography, where he naively talks about "protecting ourselves with mathematics." His most famous maxim is now "Security is a process, not a product," and he keeps constantly talking about how security is all about risk management, not risk avoidance. Exactly what you're saying, isn't it? Have a look at Secrets and Lies and the Crypto-Gram archive sometime.
I don't know why all of these so-called "security experts" keep on advocating biometrics with little or no understanding of their real properties, much less how they should be properly used. Biometrics can be used as unique identifiers, but biometrics are not secrets. They can provide a unique identifier in an already trusted environment, but alone they cannot be used for authentication, which is what so many of these "experts" are ready to do. If I steal your fingerprint using any of the simple yet effective techniques (none of which require me to cut off your finger) described by Ton van der Putte, it can't be un-stolen, and nobody will be able to give you a "replacement" fingerprint.
A quote that iluustrates this naivete from the USA Today article: "Biometrics is one way to really identify the customer you're dealing with," he [Steve Vallance] says. What a foolish, naive statement. Alone, biometrics cannot really identify anybody.
I really can't do any better than point people out to an article in yet another issue of Crypto-Gram, which first came out five years ago: Biometrics: Truths and Fictions.
Ah yes, you're right. Thank you for the correction.
Aye, shame, true shame. Fortunately I was able to find a copy of the original journal in our university library. The only other link I found was from the Turing Digital Archive, but it's all scanned images, scarcely better than the IE version. Anyhow, it's here.
Shannon's 1948 paper, "A Mathematical Theory of Communication", the seminal work on information theory and coding.
"On computable numbers, with an application to the Entscheidungsproblem"" is unarguably the paper that began the field of computer science as we understand it today. Here we have the first descriptions of universal computing devices, Turing machines, which eventually led to the idea of universal stored-program digital computers. The paper even seems to describe, in what is unarguably the first ever conceptual programming language, a form of continuation passing style in the form of the "skeleton tables" Turing used to abbreviate his Turing machine designs. It's also relatively easy reading compared to many other scientific papers I've seen.
Along with this we might also include Alonzo Church's 1941 paper "The Calculi of Lambda Abstraction" (which sadly does not appear to be anywhere online), where the lambda calculus, the basis for all functional programming languages, is first described.
And remember that Oceania has ALWAYS been at war with Eastasia.
Ah, I remember those days, and yes, I remember the reasons, which you seem to have forgotten. I assume you're talking about the old VGA mode 12h 640x480 16-color modes and the similar EGA 640x350 16-color mode. The reason for the somewhat absurd (in this day anyhow) system for getting pixels onto the screen is the 640k limit that existed in those days, and to be more precise, all of video memory was fixed to remain in two 64k segments beginning at physical address A0000h. Think about it. (640*480)/2 (because 16 colors == 4-bit color) is 153,600 bytes. How the heck are you going to squeeze that much video memory into two segments of 128K total? They could have used paging, like EMS in those days, but that would mean that scrolling and a lot of useful features would be excruciatingly slow as your blitter would need to switch banks every frame of scroll. Instead, we got bitplanes and masking and all that, which allowed many of these operations to happen very quickly. Granted, it's totally unintuitive to a newcomer, and reeks totally of a kluge, but what PC hardware in those days was a simple thing to program?
And these kluges also inadvertently served to create the undocumented Mode X 320x240 256-color mode which was nirvana indeed for PC animation enthusiasts (like Mike Abrash) in the early nineties... :)
Oh how I think we would all wish it was just crack! If it were, then McBride and Boies and the whole lot would have OD'ed by now and we'd just drag their corpses out of their office in Utah and bury them at the crossroads with a stake through the heart.
Unfortunately, they seem to be high as a kite on something that you can't OD on that easily. :(
AFAIK, the classical channel can be eavesdropped upon, and the eavesdropper will gain no information about the key being exchanged, so it doesn't need to be absolutely secure. It just needs to be safe from jamming, because if you jam the classical channel the two parties have no way of discussing the level of interference on the quantum channel, and the protocol breaks down. As has been mentioned repeatedly, it is impossible to passively eavesdrop on the quantum channel and gain useful information without being detected by the two communicants.
To be more exact, I think a more valid description of what's up with this device is that it's too far ahead of its time and is unripe. The ideas have been proved in the lab, but to make one for consumption by the general public is premature.
Well, some modern protocols (e.g. Kerberos) still require you to exchange symmetric crypto keys beforehand. I suppose these could be useful in situations similar to those. You could in theory build a quantum crypto-based Kerberos variant using these protocols.
Right. But well, some people have attempted to develop quantum authentication protocols like this one, this one, and this one. Dunno if the device in question does any of them, or even if any of them are actually practical to use with today's technology. If the device in question doesn't use quantum authentication of some kind, well, they're selling snake oil, but I wouldn't dismiss the whole concept of quantum cryptography out of hand totally the way you seem so ready to.
By the way, I've found a link to the paper I mention above. It's the paper by Charles Bennett, Francois Bessette, Gilles Brassard, Louis Salvail, and John Smolin, "Experimental Quantum Cryptography" (Citeseer link).
I imagine they're using the protocol, or a close variation thereof, of the one described by Charles H. Bennett, Gilles Brassard, Louis Salvail, and John Smolin in their paper "Experimental Quantum Cryptography" (Citeseer link). They use the quantum channel to allow Alice and Bob to negotiate a random key known only to them, which cannot be eavesdropped upon, and also allows them to discover any eavesdropper attempting to obtain their key. You can't use a quantum channel to transmit real information, but you can use it to negotiate a perfectly random key between two communicants without anyone but the two parties knowing it. In other words, the quantum cryptographic protocols are just like Diffie-Hellman and other key exchange protocols we all know well, only they provide a physical proof that according to the laws of physics as we understand it today, eavesdropping is impossible.
I don't think amplification is going to do you good. If this implements the first quantum key exchange protocol, designed by Gilles Brassard and others (notably this is the easiest one to implement using present-day technology so far), they actually work by transmitting single photons one at a time, with the transmitter whom we shall call Alice (the one generating the key to be used for your subsequent symmetric cryptography) controlling the polarization. The recipient of the photons (Bob) then measures the polarization, then depending on whether the answer was correct or not decide what the value of the key bits should be. Doesn't matter if Eve is able to eavesdrop on the classical channel where they're communicating about polarization methods, Brassard and his colleagues showed that she will not be able to gain any information about the key Bob is receiving from there. If Eve tries to tap into the photon stream she'll be noticed by Alice and Bob because of the way her measurements are affecting Bob's measurements... Anyone feel free to correct me on this, I don't have Brassard's paper with me on hand at the moment and can't find it on the web, so I've been just going by it from memory.
(in case you haven't yet noticed, this quantum cryptographic protocol, like most the others I've come across, is a key exchange protocol that works as a replacement for Diffie-Hellman or something similar).
The upshot is that the fiber cable here will be from the point of view of conventional fiber optic equipment a dark fiber. Light intensity is so low that only specialized equipment can detect it. Anything else inside the fiber, be it an amplifier or a DWDM switch, or whatever else, won't recognize the transmission and may do something totally unexpected that breaks the protocol. Bob could always be reading the wrong polarization, and hence Alice would have to keep discarding key bits. However, apparently there are excellent fiber cables that can go for several kilometers and still maintain the kind of integrity required for the protocol (IIRC, Brassard mentions somewhere that as early as 1996 someone actually built a real setup employing his protocol, over a 15km fiber optic cable).
Other quantum key exchange protocols I've come across are noticeably similar, and suffer from the many of the same limitations.
The only use that I can see for this is if you're someone with the resources to lay your own fiber around, say if you have a large complex covering several acres, and are more paranoid than the NSA. There is a market, I think, but that market looks more like the military and intelligence agencies of large industrialized nations. Building such an infrastructure could cost billions, even if you all you wanted to do was securely interconnect several dispersed branch offices in one large city...
If you can already transmit random bits by means of quantum cryptography, why not go the whole nine yards and exchange random keys that are as long as your entire message? The papers I've read on quantum cryptography all employ it as a method for the secure exchange of one-time pads. The only reasons I can think of are that 1. the effective bandwidth of the quantum channel is too low and/or 2. they're having problems setting up a reliable and fast source of real random numbers on their device (this is actually a lot harder to do than most people think).
Do the words "Cyber-Armageddon" mean anything to you? I imagine what some of the things they'll try to simulate are the really monstrous distributed denial of service attacks that would cripple the entire Internet for days and cause untold amounts of damage and inconvenience for millions all over the world. Things that would make a major slashdotting look like a mild hiccup. If they did that on the real Internet, the damage it would cause worldwide would be worth a helluvalot more than a piddly US$5.5 million.
I imagine they'll be dealing more with network structural problems than cracking problems. They'd also be doing experiments on some conjectures like the Warhol worms and flash worms we've heard about that can infect 60% or more of all the hosts on the Internet in under 15 minutes. If they did such an experiment on the real Internet, that would again cause a lot of major disruption, especially if they made a mistake in the worm's programming, and that's not exactly an unknown occurrence... The Great Worm itself had a bug that rendered it far less benign than Robert T. Morris planned. If that happened on one of the world's largest WAN's however, no big deal, it'll just inconvenience the grad students in charge of it, and they'd need to reinstall, which is probably part of what they'll be getting paid to do anyway.
What you're proposing by the way, would be tantamount to condoning acts that the Department of Homeland Security (which appears to be one of the agencies funding this project) has labeled terrorism, and which have been illegal since long before 9/11.