Not to get into the old debate, but it's worth noting that Stallman considers licenses restricting military use (or use by pretty much anyone) to be immoral. Part of his argument is that you never know which (if any) side(s) is going to end up having a 'legitimate' need for military code in advance, so it's better to make it available to all. Restrictions on military use would also violate the literal text of the GPL.
If you wanted to model this kind of thing, you'd need something like prior probabilities for each of the generating mechanisms. The 'probability' of a stupid human is 0.5; the probability of a uniform rng being used is 0.1; etc., and each mechanism has its associated 'probability' of generating each number (or number sequence if you prefer, it's really the same). The probability of getting 0000000 given a stupid human will be higher than getting 0000000 given a uniform rng. Meh.
What you're talking about is more like Kolmogorov complexity, which is cool, but has a few problems.
The only "flaw" is that it reduces entropy. This reduction can be trivially computed given a well-defined discard scheme; you may then increase the length of the key slightly to compensate, if necessary.
In exchange, you rule out an entire class of script kiddies who would quickly find one of the discarded keys, but will need to use exhaustive search to find the remainders.
In short: the entropy is, strictly speaking, only valid if the attacker is using the same generative function as the key generator. Reality is slightly different. Losing a few bits of entropy to rule out the chance of presenting "low-hanging fruit" is often a smart move.
If by 'scale,' you mean compute power, then it obviously depends on whether the processing power reaches some kind of saturation level. By analogy with graphics, we may well reach a point where centralized is 'good enough' to simulate decentralized, but may be preferable for some other reason. Or, once financial trading is 99%+ algorithmic, how much does it really matter who is running the algorithm?
The usual free market argument is that the actual information involved in a free market cannot be elicited in any centralized way, and I think this is what parent poster was referring to. The government can run a central mandatory trading house and, as long as it keeps itself from causing distortions for its own ends(*), this would broker most, if not all, of the information that a decentralized market would, with the concomitant benefits of centralization.
*: This is where it usually all falls down, of course.
As i understand it, there was some concern about something like this happening to anti-malware organizations. So, call it "pups" instead. Everyone knows, or will soon know, what you really mean, but it's technically hard to argue that it's slander.
Interesting. IBM has been moving to Linux internally for a while, which only makes sense for business reasons. Do you have a cite on banning W8, and running W7 only on VMs?
fwiw, I feel the same way as metrix007. The application is statistical computing, and the environment is typically R with some C for non-vectorizable loops and such. I write in emacs.
The thing is, not everyone who codes is a coder. If your job is to write maintainable software, generally to other folks' specs or ideas, over a period of months then, yeah, you'd better play nice. That job sounds like hell to me, and I'm glad I don't have it.
95% of the code I write never gets re-used, because the idea it was implementing turns out to suck, and it's hard to know in advance. It makes a lot more sense to just rewrite from scratch the 5% that ends up useful or interesting to other people.
Well, in this case, they did agree in advance, quite clearly, when the app asked for privileges. The analogy would be hiring a ghostwriter with an agreement to not exercise editorial control, which is just fine (barring cases of libel). In fact, that's almost exactly what they did.
Yes, significant harm to reputation is actionable. This is insignificant.
As an aside, wouldn't making an agreement to allow an unknown third party to post anything on one's behalf make one, ipso facto, not a sensible and judicious person?
oh christ, it wouldn't help. people would just clamor for granularity on the level of "allow the app to post for me, but only if it makes me look cool, according to undefined standards decided retroactively by the twitter mob," and then be shocked (shocked!) when this convention is breached.
People like you cheapen the meaning of everything.
There's also an established principle in the law where you have to be able to show some kind of harm or imminent harm in order to sue. You would be laughed out of court.
Your principle is absurd, and would make ghostwriting a form of fraud.
they both "trend" toward average; the smaller ones just have more variation about the trend line.
also, just because the null hypothesis works (i doubt it would explain everything here, but even if it did), that doesn't mean that it's correct, especially if you have good prior reason to believe another explanation.
i'm sorry, but without definitions we're not going to get anywhere. it's been a while since my formal training in complexity theory, but i don't think i've committed any fundamental errors; i'm trying my best to both recall and summarize for a public forum, while you do nothing but bat around baseless assertions and try to insult me. there's nothing wrong with baseless assertions, but please don't assign them labels which you don't understand. consider refining your ideas into precise statements, rather than disparaging this as some kind of triviality you've transcended.
particularly, if you'd just follow the links from the wikipedia page you cited, you'd see the importance of probability in the formulation of Shannon information.
information theory tells you the limits of what you can know, without regard for how hard it might be to get there. (actually it mostly proves upper bounds on these limits, but that's the style of it.)
computation theory speaks of how hard it is to get there.
if your opponents have only a classical computer and polynomial time, then you don't actually need a one-time pad (which is an information theoretic limitation); you can fake it, and they won't have enough time to unweave your trick. in the other direction, if i know a message X, and give you a one-way invertible function applied to X, i haven't really told you anything, except that I've given you a way to check if someone else (under appropriate assumed restrictions) probably has exactly the same message; you can request they apply the same function and give the result.
the thing is, roughly speaking, Shannon theory depends on a probabilistic model (or something very close to it), while computation theory does not. computation theory can analyze probabilistic computation, but it's optional. i'd like to see a derivation of an equivalent of (or refinement of!) P and NP and NP-completeness from information theory. i don't think it's possible, and those are pretty important ideas in computational complexity.
We'd like to give your reverse engineers a complete run-down on the video card, the HDMI adapter, and the lock-out chip of the card... Yes! I mean... if we're unable to support third-party adapters, then... I'd say that, ah... well, ah... we're just gonna have to help you bypass it, Dmitri...
Slashdot Mirror
Not to get into the old debate, but it's worth noting that Stallman considers licenses restricting military use (or use by pretty much anyone) to be immoral. Part of his argument is that you never know which (if any) side(s) is going to end up having a 'legitimate' need for military code in advance, so it's better to make it available to all. Restrictions on military use would also violate the literal text of the GPL.
At least he's consistent.
Best damn car on the lot.
Using the standard modern definition, randomness is a property of an algebra of subsets of a sample space, but who really cares?
If you wanted to model this kind of thing, you'd need something like prior probabilities for each of the generating mechanisms. The 'probability' of a stupid human is 0.5; the probability of a uniform rng being used is 0.1; etc., and each mechanism has its associated 'probability' of generating each number (or number sequence if you prefer, it's really the same). The probability of getting 0000000 given a stupid human will be higher than getting 0000000 given a uniform rng. Meh.
What you're talking about is more like Kolmogorov complexity, which is cool, but has a few problems.
Okay, maybe not "often," but it is quite possible that it would be a good idea.
The only "flaw" is that it reduces entropy. This reduction can be trivially computed given a well-defined discard scheme; you may then increase the length of the key slightly to compensate, if necessary.
In exchange, you rule out an entire class of script kiddies who would quickly find one of the discarded keys, but will need to use exhaustive search to find the remainders.
In short: the entropy is, strictly speaking, only valid if the attacker is using the same generative function as the key generator. Reality is slightly different. Losing a few bits of entropy to rule out the chance of presenting "low-hanging fruit" is often a smart move.
If by 'scale,' you mean compute power, then it obviously depends on whether the processing power reaches some kind of saturation level. By analogy with graphics, we may well reach a point where centralized is 'good enough' to simulate decentralized, but may be preferable for some other reason. Or, once financial trading is 99%+ algorithmic, how much does it really matter who is running the algorithm?
The usual free market argument is that the actual information involved in a free market cannot be elicited in any centralized way, and I think this is what parent poster was referring to. The government can run a central mandatory trading house and, as long as it keeps itself from causing distortions for its own ends(*), this would broker most, if not all, of the information that a decentralized market would, with the concomitant benefits of centralization.
*: This is where it usually all falls down, of course.
apropos of nothing, but the "creatures" are probably "reptilians".
As i understand it, there was some concern about something like this happening to anti-malware organizations. So, call it "pups" instead. Everyone knows, or will soon know, what you really mean, but it's technically hard to argue that it's slander.
rubycodez is right.
mod up, or live with your hypocrisy.
Interesting. IBM has been moving to Linux internally for a while, which only makes sense for business reasons. Do you have a cite on banning W8, and running W7 only on VMs?
fwiw, I feel the same way as metrix007. The application is statistical computing, and the environment is typically R with some C for non-vectorizable loops and such. I write in emacs.
The thing is, not everyone who codes is a coder. If your job is to write maintainable software, generally to other folks' specs or ideas, over a period of months then, yeah, you'd better play nice. That job sounds like hell to me, and I'm glad I don't have it.
95% of the code I write never gets re-used, because the idea it was implementing turns out to suck, and it's hard to know in advance. It makes a lot more sense to just rewrite from scratch the 5% that ends up useful or interesting to other people.
There are several reputable third-party listening tests linked at the bottom of the page, below their own, admittedly unjustified, graphs.
Well, in this case, they did agree in advance, quite clearly, when the app asked for privileges. The analogy would be hiring a ghostwriter with an agreement to not exercise editorial control, which is just fine (barring cases of libel). In fact, that's almost exactly what they did.
Yes, significant harm to reputation is actionable. This is insignificant.
As an aside, wouldn't making an agreement to allow an unknown third party to post anything on one's behalf make one, ipso facto, not a sensible and judicious person?
oh christ, it wouldn't help. people would just clamor for granularity on the level of "allow the app to post for me, but only if it makes me look cool, according to undefined standards decided retroactively by the twitter mob," and then be shocked (shocked!) when this convention is breached.
People like you cheapen the meaning of everything.
There's also an established principle in the law where you have to be able to show some kind of harm or imminent harm in order to sue. You would be laughed out of court.
Your principle is absurd, and would make ghostwriting a form of fraud.
they both "trend" toward average; the smaller ones just have more variation about the trend line.
also, just because the null hypothesis works (i doubt it would explain everything here, but even if it did), that doesn't mean that it's correct, especially if you have good prior reason to believe another explanation.
That's a trick. They're not relevant anywhere.
a good supply of media-addled, uncritical, disposable human garbage is also economically beneficial.
i'm sorry, but without definitions we're not going to get anywhere. it's been a while since my formal training in complexity theory, but i don't think i've committed any fundamental errors; i'm trying my best to both recall and summarize for a public forum, while you do nothing but bat around baseless assertions and try to insult me. there's nothing wrong with baseless assertions, but please don't assign them labels which you don't understand. consider refining your ideas into precise statements, rather than disparaging this as some kind of triviality you've transcended.
particularly, if you'd just follow the links from the wikipedia page you cited, you'd see the importance of probability in the formulation of Shannon information.
this is starting to sound interesting. could you elucidate please?
information theory tells you the limits of what you can know, without regard for how hard it might be to get there. (actually it mostly proves upper bounds on these limits, but that's the style of it.)
computation theory speaks of how hard it is to get there.
if your opponents have only a classical computer and polynomial time, then you don't actually need a one-time pad (which is an information theoretic limitation); you can fake it, and they won't have enough time to unweave your trick. in the other direction, if i know a message X, and give you a one-way invertible function applied to X, i haven't really told you anything, except that I've given you a way to check if someone else (under appropriate assumed restrictions) probably has exactly the same message; you can request they apply the same function and give the result.
the thing is, roughly speaking, Shannon theory depends on a probabilistic model (or something very close to it), while computation theory does not. computation theory can analyze probabilistic computation, but it's optional. i'd like to see a derivation of an equivalent of (or refinement of!) P and NP and NP-completeness from information theory. i don't think it's possible, and those are pretty important ideas in computational complexity.
We'd like to give your reverse engineers a complete run-down on the video card, the HDMI adapter, and the lock-out chip of the card... Yes! I mean... if we're unable to support third-party adapters, then... I'd say that, ah... well, ah... we're just gonna have to help you bypass it, Dmitri...
only restricted to citizens in so far as it's talking about the rights of citizens
uh, yeah, that's how governments work. were you dropped in your head as a child?
Uh, it still only applies to citizens which, one notes, Chinese nationals are not.
lol. not giving away everything is now "protectionism."