Actually, when I was TAing data structures, we called that 52 card pick up sort. You take the deck of cards and throw it in the air. Pick up the cards and if they end up sorted, stop. If not, throw them in the air again. We used it as an example of "just because it works, doesn't mean you should do it" and as an example of algorithms with big os in the 'bad' column.
Actually, if you would read the treaty, instead of the/. write up, you would realize that it doesn't at all. It is an agreement to share evidence. That's pretty much it. What I meant with 'in these two countries' is that if you commit a crime in country A and another one in country B and some evidence is of interest to both cases, A and B share evidence. And depending on the way the german law is written, it is quite possible that the american _did_ break a law in the example you gave-- but would probably never face charges. In a more obvious example, if you hired someone to kill a german in germany, you would probably be arrested if you stepped foot in germany. That isn't a dual criminality issue. That's right, imagine the outrage-- you can commit a crime in a country you've never even been to! But yes, it is possible.
They difference between your example and mine is the location of the sale: in yours the sale occurs in the US, and it is not your liability where the goods go after that. However, it _is_ illegal for you to sell those goods in germany, and that includes via some form of mail order (i.e. the internet). Under whose laws? German laws. Could the germans bring you up on charges? Yes. Could the americans? No. Could they even investigate you? No.
However, say you were under investigation for cracking into an american network and also by german authorities for cracking into a computer system in germany. The US, under the terms of this treaty, would share evidence it had gathered (for example, log files from your computer and your ISP) with the german government.
As for the NYT example-- a) not relavent, because it isn't a computer crime b) even if it were a general treaty, the NYT would already have to be under investigation in the US for offending the dignity of government officials, which would not happen, because it isn't illegal here. Many people have said that this treaty essentially creates the strictest of all of the availible laws, however, what it really does is say "Hey, you broke the law, on these two occasions, in these two countries. They are both computer crimes, and so the evidence helpful to one investigation might help another. Let's share the evidence that we already had legally, so there won't be duplicate effort between our agencies."
a FPGA is just a block of logic gates that can be connected after the original manufacture. Typically, they are used to implement simple logic cheaply and easily. This is more of an entire processor designed on a similar principle. I would guess that it includes registers, a clock, bus connection facilities, etc. If anything, this is closer to a CPLD which combines i/o blocks, function cells and interconnection blocks to create somewhat more complicated (and often times sequential, as opposed to combinational) logic.
IANALIAOTCIQ (... in any of the countries in question), but I would think that would be an issue for the laws of country B, especially those regarding illegal gathering of evidence. For example, I think that it would be illegal to use any evidence in the united states that was gathered by torture, regardless of where it was gathered.
The problem with your argument is that you have to commit a german crime in Germany or against germans for it to fall under german law, this treaty or not. I.E. you have to sell the nazi stuff to a german, which would be illegal anyway. That's right, it is already illegal. All this treaty says is that if you break german laws (by attacking a german computer system) then the americans (if they ratify the treaty) are under obligation to turn over any evidence they have on you for computer crimes.
For example, say you hack into the computer system of a large american university and get investigated for it. While under investigation, you also crack into a large german corporation's server farm. Under the terms of the treaty, the american and german governments should exchange their information about you.
Actually, they aren't under requirement of the treaty to provide information they don't have. If they can't have the keys legally, then they are under no requirement to give it. What the treaty prevents is law enforcement in one country stonewalling law enforcement in another country to prevent cybercrimes evidence from one country to be used in another.
Thank you, I was about the post that:) It looks like this is another one of those cases of the/. community getting irrationally exuberant about a document before they have even read it and/or applied basically reading comprehension skills.
I've looked through this treaty, and it appears that the only explicit mention of encryption is that each participating country must ensure that if they have encryption keys needed to help another participating country, they should hand them over (i.e. Country A got Mr. Baddy's RSA key during an investigation and he is being tried in Country B for another offense. Country A should give the key to Country B to help them). Presumably, the key must be obtained by legal means in country A before it can be given to country B. They also mention that encryption should be used, if necessary, to ensure secure communications between the governments... I would hope this is the case anyway.
This treaty doesn't expand the definition of computer crime really. All it is is a promise between countries that if someone commits a crime in another participating country, the other countries will turn over the criminal. To me, this makes perfect sense-- think about it. If someone from a european nation stole your credit card information, for example, you would want them to be accountable for their damages, even if you were an american, right?
I'm not familiar with libsafe specifically, but I have spent some time looking into obsd's w^x and PaX linux, which have similar goals. My question is, if the buffer overflow is in your own code, would libsafe stop it, or is it just for libraries? Also, does it only protect the stack, or also heap memory and the data regions of an executing program?
I think he was being clever, but I wasn't there to personally witness the quote. I think what he was saying, in a roundabout kind of way was "OS X is kind of like linux, but I like it even more."
Translation and protection need to be done for every instruction fetch _in a new segment and/or page_. In x86-64, this means that in a best case (no long jumps, large page size) this means that the number of checks can be significantly reduced. A similar system is already in place for the segmentation model on x86: when the segment registers are loaded, the processor loads a ghost register which contains the protection information, base address, etc for the segment. This register is checked against (not the segment descriptor in main memory) when checking permission issues and when calculating addresses. It isn't entirely an issue of caching.
Also, the 'weird' low-perfomance processors you are talking about are _everywhere_. Just because x86 is the most popular in the desktop market doesn't mean that the x86 way is the only way to do it-- or the best. There is a lot to be said for reducing the power consumed by the core and reducing core cost by moving some logic off of the primary chip.
Care to explain why? As the other child explained the MMU doesn't have to be on-die (I'd also like to point out that the memory controller doesn't have to be off-die, either, which gives are variety of options: mmu and mc external, mmu internal and mc external, both on the chip). In the x86 architecture, just about any hardware can create an interrupt. A certain interrupt from an offboard memory management unit could signal a NX fault. Also, I'll have you note that I said nothing about the NX circuitry being located outside the MMU (I guess, theoretically, it could be, but that would probably be a bad design decision). The way you assert things as absolute is really what bothers me-- very little in higher level computer engineering is an absolute rule.
But what he is saying is that if the hash is md5, you find one string that corresponds to each possible hash. Then it doesn't matter if it is salted or not, because the salted version would be in the dictionary as well. This isn't really feasible with current technology for md5, but it is part of the reason the old unix crypt has a limited shelflife.
They do, the problem is that they don't want to get it. They have too much invested in the old way of doing business. So, they are going to sabotage the new way by jacking up prices. That way, they can say "Look, we tried but people wouldn't buy our (overpriced, horrible) albums online. Guess we have to go back to the stores." They are covered in case they ever get accused of stomping on competition or price fixing, but they don't actually have to invest the money in retooling their strategy.
A lot of this kind of discussion came up in one of the recent e-voting stories. The problem is, if I use an open source compiler (assuming the code is audited by a non-malicious party, in the worse case, me), then you would have to resort to a hardware based attack (i.e. the device essentially 'patches' in the exploit). Of course, this is a lot harder than people seem to think. What if I, for example, compiled it with a compiler that used stack and memory map randomization? The hardware would patch over the wrong functionality. Suddenly, the router would not function as intended and it would be obvious that shenanigans were occuring. So then, the hardware would have to have a totally seperate control system outside of what was availible in the software. It might, for example, interpret certain seemingly normal sequence of packets as a command to send back some naughty data. BUT all the logic would have to be implemented in hardware, which would lead to a noticable increase in development and production costs. In fact, I think that from a practical standpoint, a complex backdoor system would far overshadow the actual engineering work for the router itself.
Had you read the original, you would have known that I mentioned checksumming. I also mentioned a) a specific compiler (as in, a specific version) and b) a standard procedure (as in, a certain set of build flags). If they won't let you run the code, and you know it is good, you get to raise a stink, and you have more evidence than in the 'I pressed the button, but it counted my vote for someone else completely' case. If they just plain don't want you running it, then they aren't following the 'rules.' That isn't a flaw with the system. You can't say "they wouldn't let me vote because I'm ____" is a valid complaint against a ballot box system, would you?
Unless, of course, you don't use their gcc. I'm saying they should pick a compiler, say, gcc 3.1. Then I can get (I wouldn't be required to do any of this to vote, see my clarification attached to the other reply) the gcc (from gnu) and the code (from them) and compile mine. Like I said, it doesn't solve the problem, it makes it makes the problem more difficult to exploit. If you combined this with hardware audits (i.e. somebody buys a voting machine and the company is unaware whether it will go to an election or to someone checking up on them-- then you see if it gives the 'right' answer).
Not everyone would have to do it-- trust their stuff if you want. I am not advocating the ability to use a compiler as a prerequiste to be able to vote. I am saying, the option should be availible to those who care. It would make it possible to randomly check these machines for honesty.
The problem is that this code might not be the code that is in the machines. I think we should give people to compile the released source themselves and somehow load that into the machine before voting. There would be a standard build procedure, so everyone would use the same (open) compiler. The code would be checked to make sure it is bit for bit identical to a version compiled by a trusted third party. That way, having produced the executable yourself, you can be fairly sure that the software can be trusted. Unless, of course: a) you don't trust gcc. In this case, you probably have at least 3 layers of tin foil in your hat, and there is _nothing_ I can do to make you happy. b) The company/spooky gov't agents/other private interests set up the hardware to handle the backdoor/vote rigging itself. In this case, the voting machine company would have to at the least partner with a hardware manufacturer. There would have to be somekind of audit process for the firmware.
You're right. The advantage to the code being publically availible is that if there is an unintentional backdoor (security flaw) it might be spotted sooner. It wouldn't protect from intentionally mallicious actions by the company.
Ok, you might know this already, and you aren't really arguing to the contrary, but I thought I would point out the flaw in SCO's argument here.
Quoth the GPL:
2. You may modify your copy or copies of the Program or any portion of it, thus forming a work based on the Program, and copy and distribute such modifications or work under the terms of Section 1 above, provided that you also meet all of these conditions:
a) You must cause the modified files to carry prominent notices
stating that you changed the files and the date of any change.
b) You must cause any work that you distribute or publish, that in
whole or in part contains or is derived from the Program or any
part thereof, to be licensed as a whole at no charge to all third
parties under the terms of this License.
End quoth.
Therefore, by modifying a piece of GPL software (linux), they agreed to distribute the modification under the GPL. If they didn't use the appropriate notification of the change, they are violating the terms of the GPL, which they agreed to by contributing to a GPL program. They aren't released from the terms of the GPL. If you refuse to make a payment on your apartment, does that entitle you (because you broke the contract) to choose any new contract terms you want (rent is now $-10,000 a month, k thx!)? No, obviously. So why does a violation of the terms of the GPL entitle them to exemption from their legal requirements?
Slashdot Mirror
You see? And the drama goes on... dum dum dum!!
Actually, when I was TAing data structures, we called that 52 card pick up sort. You take the deck of cards and throw it in the air. Pick up the cards and if they end up sorted, stop. If not, throw them in the air again. We used it as an example of "just because it works, doesn't mean you should do it" and as an example of algorithms with big os in the 'bad' column.
Actually, if you would read the treaty, instead of the /. write up, you would realize that it doesn't at all. It is an agreement to share evidence. That's pretty much it. What I meant with 'in these two countries' is that if you commit a crime in country A and another one in country B and some evidence is of interest to both cases, A and B share evidence. And depending on the way the german law is written, it is quite possible that the american _did_ break a law in the example you gave-- but would probably never face charges. In a more obvious example, if you hired someone to kill a german in germany, you would probably be arrested if you stepped foot in germany. That isn't a dual criminality issue. That's right, imagine the outrage-- you can commit a crime in a country you've never even been to! But yes, it is possible.
and that is about all I have to say on the subject
They difference between your example and mine is the location of the sale: in yours the sale occurs in the US, and it is not your liability where the goods go after that. However, it _is_ illegal for you to sell those goods in germany, and that includes via some form of mail order (i.e. the internet). Under whose laws? German laws. Could the germans bring you up on charges? Yes. Could the americans? No. Could they even investigate you? No.
However, say you were under investigation for cracking into an american network and also by german authorities for cracking into a computer system in germany. The US, under the terms of this treaty, would share evidence it had gathered (for example, log files from your computer and your ISP) with the german government.
As for the NYT example-- a) not relavent, because it isn't a computer crime b) even if it were a general treaty, the NYT would already have to be under investigation in the US for offending the dignity of government officials, which would not happen, because it isn't illegal here. Many people have said that this treaty essentially creates the strictest of all of the availible laws, however, what it really does is say "Hey, you broke the law, on these two occasions, in these two countries. They are both computer crimes, and so the evidence helpful to one investigation might help another. Let's share the evidence that we already had legally, so there won't be duplicate effort between our agencies."
a FPGA is just a block of logic gates that can be connected after the original manufacture. Typically, they are used to implement simple logic cheaply and easily. This is more of an entire processor designed on a similar principle. I would guess that it includes registers, a clock, bus connection facilities, etc. If anything, this is closer to a CPLD which combines i/o blocks, function cells and interconnection blocks to create somewhat more complicated (and often times sequential, as opposed to combinational) logic.
IANALIAOTCIQ (... in any of the countries in question), but I would think that would be an issue for the laws of country B, especially those regarding illegal gathering of evidence. For example, I think that it would be illegal to use any evidence in the united states that was gathered by torture, regardless of where it was gathered.
The problem with your argument is that you have to commit a german crime in Germany or against germans for it to fall under german law, this treaty or not. I.E. you have to sell the nazi stuff to a german, which would be illegal anyway. That's right, it is already illegal. All this treaty says is that if you break german laws (by attacking a german computer system) then the americans (if they ratify the treaty) are under obligation to turn over any evidence they have on you for computer crimes.
For example, say you hack into the computer system of a large american university and get investigated for it. While under investigation, you also crack into a large german corporation's server farm. Under the terms of the treaty, the american and german governments should exchange their information about you.
Actually, they aren't under requirement of the treaty to provide information they don't have. If they can't have the keys legally, then they are under no requirement to give it. What the treaty prevents is law enforcement in one country stonewalling law enforcement in another country to prevent cybercrimes evidence from one country to be used in another.
Thank you, I was about the post that :) It looks like this is another one of those cases of the /. community getting irrationally exuberant about a document before they have even read it and/or applied basically reading comprehension skills.
I've looked through this treaty, and it appears that the only explicit mention of encryption is that each participating country must ensure that if they have encryption keys needed to help another participating country, they should hand them over (i.e. Country A got Mr. Baddy's RSA key during an investigation and he is being tried in Country B for another offense. Country A should give the key to Country B to help them). Presumably, the key must be obtained by legal means in country A before it can be given to country B. They also mention that encryption should be used, if necessary, to ensure secure communications between the governments... I would hope this is the case anyway.
This treaty doesn't expand the definition of computer crime really. All it is is a promise between countries that if someone commits a crime in another participating country, the other countries will turn over the criminal. To me, this makes perfect sense-- think about it. If someone from a european nation stole your credit card information, for example, you would want them to be accountable for their damages, even if you were an american, right?
I'm not familiar with libsafe specifically, but I have spent some time looking into obsd's w^x and PaX linux, which have similar goals. My question is, if the buffer overflow is in your own code, would libsafe stop it, or is it just for libraries? Also, does it only protect the stack, or also heap memory and the data regions of an executing program?
I think he was being clever, but I wasn't there to personally witness the quote. I think what he was saying, in a roundabout kind of way was "OS X is kind of like linux, but I like it even more."
Translation and protection need to be done for every instruction fetch _in a new segment and/or page_. In x86-64, this means that in a best case (no long jumps, large page size) this means that the number of checks can be significantly reduced. A similar system is already in place for the segmentation model on x86: when the segment registers are loaded, the processor loads a ghost register which contains the protection information, base address, etc for the segment. This register is checked against (not the segment descriptor in main memory) when checking permission issues and when calculating addresses. It isn't entirely an issue of caching.
Also, the 'weird' low-perfomance processors you are talking about are _everywhere_. Just because x86 is the most popular in the desktop market doesn't mean that the x86 way is the only way to do it-- or the best. There is a lot to be said for reducing the power consumed by the core and reducing core cost by moving some logic off of the primary chip.
Care to explain why? As the other child explained the MMU doesn't have to be on-die (I'd also like to point out that the memory controller doesn't have to be off-die, either, which gives are variety of options: mmu and mc external, mmu internal and mc external, both on the chip). In the x86 architecture, just about any hardware can create an interrupt. A certain interrupt from an offboard memory management unit could signal a NX fault. Also, I'll have you note that I said nothing about the NX circuitry being located outside the MMU (I guess, theoretically, it could be, but that would probably be a bad design decision). The way you assert things as absolute is really what bothers me-- very little in higher level computer engineering is an absolute rule.
But what he is saying is that if the hash is md5, you find one string that corresponds to each possible hash. Then it doesn't matter if it is salted or not, because the salted version would be in the dictionary as well. This isn't really feasible with current technology for md5, but it is part of the reason the old unix crypt has a limited shelflife.
In AMD's implementation it is. There are no assurances that intel will implement the feature a) at all or b) in the processor, like AMD did.
I wish that the RIAA would "get it".
They do, the problem is that they don't want to get it. They have too much invested in the old way of doing business. So, they are going to sabotage the new way by jacking up prices. That way, they can say "Look, we tried but people wouldn't buy our (overpriced, horrible) albums online. Guess we have to go back to the stores." They are covered in case they ever get accused of stomping on competition or price fixing, but they don't actually have to invest the money in retooling their strategy.
A lot of this kind of discussion came up in one of the recent e-voting stories. The problem is, if I use an open source compiler (assuming the code is audited by a non-malicious party, in the worse case, me), then you would have to resort to a hardware based attack (i.e. the device essentially 'patches' in the exploit). Of course, this is a lot harder than people seem to think. What if I, for example, compiled it with a compiler that used stack and memory map randomization? The hardware would patch over the wrong functionality. Suddenly, the router would not function as intended and it would be obvious that shenanigans were occuring. So then, the hardware would have to have a totally seperate control system outside of what was availible in the software. It might, for example, interpret certain seemingly normal sequence of packets as a command to send back some naughty data. BUT all the logic would have to be implemented in hardware, which would lead to a noticable increase in development and production costs. In fact, I think that from a practical standpoint, a complex backdoor system would far overshadow the actual engineering work for the router itself.
Had you read the original, you would have known that I mentioned checksumming. I also mentioned a) a specific compiler (as in, a specific version) and b) a standard procedure (as in, a certain set of build flags). If they won't let you run the code, and you know it is good, you get to raise a stink, and you have more evidence than in the 'I pressed the button, but it counted my vote for someone else completely' case. If they just plain don't want you running it, then they aren't following the 'rules.' That isn't a flaw with the system. You can't say "they wouldn't let me vote because I'm ____" is a valid complaint against a ballot box system, would you?
Unless, of course, you don't use their gcc. I'm saying they should pick a compiler, say, gcc 3.1. Then I can get (I wouldn't be required to do any of this to vote, see my clarification attached to the other reply) the gcc (from gnu) and the code (from them) and compile mine. Like I said, it doesn't solve the problem, it makes it makes the problem more difficult to exploit. If you combined this with hardware audits (i.e. somebody buys a voting machine and the company is unaware whether it will go to an election or to someone checking up on them-- then you see if it gives the 'right' answer).
Not everyone would have to do it-- trust their stuff if you want. I am not advocating the ability to use a compiler as a prerequiste to be able to vote. I am saying, the option should be availible to those who care. It would make it possible to randomly check these machines for honesty.
The problem is that this code might not be the code that is in the machines. I think we should give people to compile the released source themselves and somehow load that into the machine before voting. There would be a standard build procedure, so everyone would use the same (open) compiler. The code would be checked to make sure it is bit for bit identical to a version compiled by a trusted third party. That way, having produced the executable yourself, you can be fairly sure that the software can be trusted. Unless, of course:
a) you don't trust gcc. In this case, you probably have at least 3 layers of tin foil in your hat, and there is _nothing_ I can do to make you happy.
b) The company/spooky gov't agents/other private interests set up the hardware to handle the backdoor/vote rigging itself. In this case, the voting machine company would have to at the least partner with a hardware manufacturer. There would have to be somekind of audit process for the firmware.
You're right. The advantage to the code being publically availible is that if there is an unintentional backdoor (security flaw) it might be spotted sooner. It wouldn't protect from intentionally mallicious actions by the company.
Ok, you might know this already, and you aren't really arguing to the contrary, but I thought I would point out the flaw in SCO's argument here.
Quoth the GPL:
2. You may modify your copy or copies of the Program or any portion
of it, thus forming a work based on the Program, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:
a) You must cause the modified files to carry prominent notices
stating that you changed the files and the date of any change.
b) You must cause any work that you distribute or publish, that in
whole or in part contains or is derived from the Program or any
part thereof, to be licensed as a whole at no charge to all third
parties under the terms of this License.
End quoth.
Therefore, by modifying a piece of GPL software (linux), they agreed to distribute the modification under the GPL. If they didn't use the appropriate notification of the change, they are violating the terms of the GPL, which they agreed to by contributing to a GPL program. They aren't released from the terms of the GPL. If you refuse to make a payment on your apartment, does that entitle you (because you broke the contract) to choose any new contract terms you want (rent is now $-10,000 a month, k thx!)? No, obviously. So why does a violation of the terms of the GPL entitle them to exemption from their legal requirements?