"self-similarity" and "fractal" are often used interchangeably in the literature. Do a literature search on either term -- it applies generally to a lot of things: NFS traffic, web site traffic, cars passing under an overpass, number of people in a bathroom stall at any moment, number of disks that fail in a massive storage array, etc...
One thing to think about when building such a system from a large number of hard disks is that disks will fail, all the time. The argument is fairly convincing:
Suppose each disk has a MTBF (mean time before failure) of 500,000 hours. That means that the average disk is expected to have a failure about every 57 years. Sounds good, right? Now, suppose you have 1000 disks. How long before the first one fails? Chances, are, not 57 years. If you assume that the failures are spread out evenly across time, a 1000-disk system will have a failure every 500 hours, or about every 3 weeks!
Now, of course the failures won't be spread out evenly, which makes this even trickier. A lot of your disks will be dead on arrival, or fail within the first few hundred hours. A lot will go for a long time without failure. The failure rates, in fact, will likely be fractal -- you'll have long periods without failures, or with few failures, and then a bunch of failures will occur in a short period of time, seemingly all at once.
You absolutely must plan on using some redundancy or erasure coding to store data on such a system. Some of the filesystems you mentioned do this. This allows the system to keep working under X number of failures. Redundancy/coding allows you to plan on scheduled maintanence, where you simply go in and swap out drives that have gone bad after the fact, rather than running around like a chicken with its head cut off every time a drive goes belly up.
My question is: how in the world does Amber LIT survive? Sure, they plead with users on their website to only use the conversion tool on content that they have legally purchased, but according to DMCA et al, users don't have the right to convert content to other formats. And creating tools to circumvent DRM is explicitly criminal according to the DMCA. So how long before Amber gets a cease-and-desist? And then what will our friend Mr. Gartenberg do?
You are supposed to buy a new copy of the DVD every couple of weeks. That's really the only moral thing to do. Making your own copies because the original might get scratched is no better than using BitTorrent to download last week's episode of your favorite show because you forgot to Tivo/Tape it -- THAT'S STEALING!
What will you do when you 2yr old turns into a 12yr old who robs little old ladies at gunpoint for heroine money, all because they saw Daddy do it to the movie studies when they were little? hmm?
Kademlia isn't a protocol per-se. If you read the Kademlia paper, you'll see that they do outline the system very well, but as for specifying the way bits/bytes are packaged up into messages, etc., there's no information. So if two different authors implement a kademlia-like DHT, you can be pretty confident that they won't talk to each other unless they've both agreed on the low-level details of the messages.
Why not use Shamir secret sharing to hold onto the private key? You can choose N people to hold pieces of the private key and choose K = N such that [any] K individuals can reconstruct the secret (but not any less than K).
The FSF describes, generally, how assignment of copyright to the FSF works for GNU projects, but I am having a devil of a time finding a copy of the assignment questionnaire that they mention.
Does anyone know what this thing looks like? Surely it involves more than emailing the maintainer and saying "I assign the copyright of my contribution to the FSF?"
Suppose I am Redhat. Why would I bother with LSB? In fact, wouldn't this be detrimental to my business? Right now, there is a huge community creating rpms for Redhat, and most commercial entities, if they offer rpms at all, offer them for Redhat. This preponderance of easy-to-install software encourages people to use Redhat, and encourages those already using Redhat to stick with it.
In that light, what does LSB buy me? An easy escape route for my customers to switch distros.
(used Redhat in this example, but I think it pretty much applies across the board).
Think about it: doesn't the fact that these Caltech guys were willing to shell out for airfare to travel 3000 miles in order to pull this stuff speak for itself? Doesn't this act itself turn MIT's campus into a sort of 'Mecca' for these people? Did they bow down and pray to The Great Dome before offering their sacrifices on the alter of prankdom?
Would anyone at MIT have entertained, even just for a moment, traveling to California to do something similar to the California Institute of Technology? I doubt it.
Not to say that Caltech isn't one of the top engineering schools in the country: of course it is. But it doesn't enjoy MIT's prestige, and these pranks just go to enlarge that prestige.
The issue that has become apparent to almost everyone involved in wireless networking is that the layered OSI-style network stack -- with its clean seperation of layers and no information exchanged except with the layer directly above or below -- just doesn't work very well in a non-wired environment. Why? Because of the time-scales involved in making and tearing down links. Compare with a wired network, where the frequency of plugging/unplugging a cable literally appears to happen eons apart, from a bit-time perspective.
For example, consider the case of directional antennas (which 802.11s doesn't concern itself with). A directional antenna must be pointed at another antenna. If there is more than one available place to point it, the decision of which antenna to point to should take many factors into consideration, including: 1) probability of being able to make a connection, 2) signal strength, which corresponds to the bandwidth that will be available, 3) the destination of traffic traveling from either end, 4) the priority of traffic destined for a particular address, etc.
Note that the act of pointing the antenna is a link layer responsibility. Also notice that while #1 and #2 could be done entirely at the physical/data link layers, #3 and #4 live way up the stack. Finally, note that #3 and #4 probably should have more to say than #1 or #2 -- it does us no good to point at B if we only have data destined for A (unless B can route it for us). In a traditional network stack, there is now way layers 1 or 2 could get the information they need to make this decision. In a mesh network there is.
Similar issues exist in omnidirectional networks. The RoofNet group at MIT, for instance, has a nice paper explaining that shortest path is about the worst thing you can do in a dense wireless network (you lose signal strength/power by making a direct link across the network, and thus have much less bandwidth than if you instead transmitted to the closest antenna between the two endpoints and had it relay to the next closest, etc)
What Linux distro put you through that in 2004? What you describe sounds more like Linux circa 1997. Seriously. USB mouse problems in XFree86? I have to admit that the first time I hooked up a USB mouse I didn't know if it would work, but you want to know how easy it was? -- 1) unplug ps2 mouse, 2) plug in usb mouse. No reconfigure. No restarting X. No nothing. This was in 2001.
Backspace didn't work out of the box? What did you install, "Larry's Linux Distro Extraordinaire"?
Sorry, I don't mean to trivialize your problems, but I am seriously shocked -- no redhat, SuSE, or Debian distro that I know of in the past 4 years would have any of the problems you describe.
On the other hand, I have had to press magic F8 buttons in the past 2 years while installing NT, just to get a standard 3com NIC driver installed (and the Dell instructions warned that if I installed the video driver first, the NIC driver wouldn't install properly!).
I'll admit though, that the last time I installed XP, it was fairly painless. And I do have to jump through a hoop or two to get my cheapo scanner working under Linux. But for the most part, it's stick in the CD, click on my locale and install class, and let 'er rip.
Dvorak makes a couple of assumptions that immediately betray his lack of knowledge:
1) Linux device drivers are a big problem
and
2) Putting Windows PnP in Linux would be an easy task
I have a problem with #1 because, well, I haven't had a problem with device drivers for years. The first thing I do with a new computer (and I've gone through 5, from Dell and HP, in the last few years) is reformat, install Windows, and then install Linux. Guess which one is easier to install? Guess which one requires special driver disks and arcane "press-F8-at-the-right-time-during-the-install" crazieness to get things working? That's right: windows. With Linux, stick the CD in, click a few buttons, and done.
The problem with #2 should be obvious to everyone: one of the main tasks of an OS is to manage devices. Look at the code in the kernel that does this. Sure, there's other important stuff (vfs, memory management, process management, etc), but if you count the lines, the heaviest piece of the OS is device driver management. Ripping this out and sticking in Redmonds garbage would be disastrous.
Now, user-mode linux is a different beast. Even virtualizing the hardware could get things to work correctly under Dvorak's scheme without so much effort. But what he suggests is not only ludicrous, its outright silly, and really illustrates how out of touch he is with how technology works.
Kim isn't, by many years, the first to try to make robots that exhibit emotions. Take a look, for example, at the groundbreaking work that was done with Kismet by Cynthia Breazeal and Rod Brooks (et al) at MIT in the late 90s.
I do like his vision of robots run amok trying to destroy humanity a la Will Smith, though. There's some good thinking.
I used to set up all my boxes so that I could plug my HP48G in to the serial port via a special cable that I made our of an old Microsoft mouse.
This was nice for when X locked up (which it did occasionally) or for when the network was down or if I didn't have another box to get network access from -- I could still telnet in through serial, kill X, and be good as new.
The only hitch was that most distros didn't come with the serial port enabled for this in the first place.
Godel just showed, in effect, that some forms of self-reference are meaningless
Actually, that isn't correct. Hofstadter makes a lot out of self reference, and certainly Zen makes a big deal out of it. But here's the kicker: there are an infinite number of incompleteness proofs, and many of them don't rely on self-reference. For instance, determining that a particular Diophantine equation doesn't have a solution for P(n)=0 is provably undecidable. Also, Chaitin's Number (which he calls Omega) also provides an interesting proof of Incompleteness using Algorithmic Information Theory instead of self-reference. Turing's Halting Problem is also a proof of incompleteness, though you could make a case that it uses some variant of self-reference (a computer program examining its own code).
The point is, mathmatical incompleteness is not a result of a paradoxical statement. In fact, the surprising result of all of this is that some mathematical facts are true by pure chance, and that the vast majority of facts in math (or in any rational system) occur randomly, and not because of derivation from some primary truth.
Math exists only in the mind. The symbols we use to do math may represent quantities in the real world, but we use symbolic manipulation to turn these symbols into more symbols, which we then project onto the world. But the world operates without these symbols. The stones or oranges or apples that you count to try to prove 1+1=2 are completely ignorant of the math involved, as is all substance that makes up those objects.
But I'm just rambling. It is true, however, that integer arithmetic is provably incomplete (which means that there are infinite "truths" in such a system which are not provable or derivable from any sort of rational process), or provably inconsistent (in which case the system is just broken, producing erroneous or conflicting results). Godel's Incompleteness proofs considered just arithmetic operations on the natural numbers and showed, irrevocably, that this was the case.
I once believed that math was intrinsically true, and that not even God could refute 1+1=2. Now I refute it.
I believe that, ultimately, reason is an invention of the mind used only to compress information into storable, manageable constructs. Outside of the mind, there is no reason, no rationality.
The universe is unreasonable, the world is irrational, and there is no "Theory of Everything" (Grand Unification Theory, etc) or simple explanation for how the whole thing really works.
And, I can prove it. Or, better stated, Godel already did. Others have also provided proofs (Turing, Chaitain) or at least alluded to the incompleteness of rational thought (Rusell, Epimedes, etc).
A related topic you may be interested in, TrustBuilder, has some nice techniques for automated trust negotiation and protocols to deal with such. It is hosted at UIUC and BYU:
"Passport, of course, doesn't uniquely identify you--you can easily get multiple passport accounts."
I'm convinced that this isn't really a problem for most uses of identification technology -- the problem is not that one guy can get 10 or 15 or 40 IDs, its that one guy and some software can get 1,000 or 100,000 or 100,000,000 IDs. An online unique identification system that solves the former and not the latter is "good enough" for most online collaborative applications that require any sort of integrity of "voting."
By calling this number, I fear you will be harassing an innocent man:
Can someone please confirm that this is the Darl McBride of SCO (what is the first letter of Darl's middle name, for starters)? For more than one reason, I believe that if it really were him, he wouldn't have his number listed (remember, the guy hires an ontourage of body guards to protect him against 'militant linux extremists' whenever he leaves his little enclave). If slashdot readers ruthlessly prank this number without confirming who it is they are really calling, its likely to backfire when open source supporters are portrayed as unjust lunatics.
And if they guy isn't the Darl Mcbride, hasn't he suffered enough already by sharing the same name?
Linux has succeeded where billion-dollar software developments have failed
And since Lemma 1.7 says "no communist is worth his own weight in dog excrement," it naturally follows that Linux must have originated elsewhere.
I propose one of the following:
Space aliens implanted Linux into Linus as a trojan horse against humanity. After Linux becomes ubiquitous, our intergalactic "friends" will return to harvest our bodies for food.
Soviets stole AT&T Unix, used hybrid nordic programmers to improve it with stealth soviet cold-war technology, and unleashed it as a trojan horse against humanity. After Linux becomes ubiquitous, our Russian "friends" will return to harvest our bodies for food.
Artificial Intelligence experiments from MIT escaped the lab and created Linux. They then implanted Linux into Linus as a trojan horse against humanity. After Linux becomes ubiquitous, our AI "friends" will return to harvest our bodies for food.
Darl McBride created a pile of cotton swabs. He named them "Georgie" and claimed that Georgie was a new type of advanced television technology for watching reruns of the Smurfs. Good for him!
I think you'll see the logic in all of this immediately.
I agree that touting this primarily as a security feature is the wrong approach. Yes, it is a bit "better" than omni, but the real benefit is "getting more bandwidth by allowing more transmitters in the same region, minimizing interference, minimizing radiation output, etc."
The thing that is exciting about this is the field of research that it opens up. Of course, directional comm antennas have been around for quite some time, but building networks out of them is relatively new. Do a literature search on "active topology formation" and you'll see what I mean -- not a whole lot has been done here, and yet this dramatically changes how you architect networks.
The exact issues involved are what spurred DARPA to recently proclaim that it was time to revise the 7-layer OSI model into something that appears more like a mesh. And that really is what will be required -- for example, waiting 10s of seconds for OSPF to figure out whether a link is up or down just will not cut it when you can form links on the order of milliseconds.
The point is, that in order to use the card, you have to have it in your hand. You can't steal my credit card simply by writing down the magic numbers (just like RSA's SecurID).
What does adding voice input from the card's owner do? Not a whole lot, except that now, instead of only needing to physically have the card in your hand, you also have to physically sound like the owner (or have a good recording of the owner speaking his password).
Is this beatable? Absolutely. But the thing to remember is that it is significantly less beatable than the current system. And since there is no such thing as unbreakable authentication, that's about the best you can ask for. If this system works, it eliminates most fraud, because most credit card theft is performed by complete strangers, i.e., people that don't have access to recording my voice or physically swiping my card.
Slashdot Mirror
"self-similarity" and "fractal" are often used interchangeably in the literature. Do a literature search on either term -- it applies generally to a lot of things: NFS traffic, web site traffic, cars passing under an overpass, number of people in a bathroom stall at any moment, number of disks that fail in a massive storage array, etc...
Suppose each disk has a MTBF (mean time before failure) of 500,000 hours. That means that the average disk is expected to have a failure about every 57 years. Sounds good, right? Now, suppose you have 1000 disks. How long before the first one fails? Chances, are, not 57 years. If you assume that the failures are spread out evenly across time, a 1000-disk system will have a failure every 500 hours, or about every 3 weeks!
Now, of course the failures won't be spread out evenly, which makes this even trickier. A lot of your disks will be dead on arrival, or fail within the first few hundred hours. A lot will go for a long time without failure. The failure rates, in fact, will likely be fractal -- you'll have long periods without failures, or with few failures, and then a bunch of failures will occur in a short period of time, seemingly all at once.
You absolutely must plan on using some redundancy or erasure coding to store data on such a system. Some of the filesystems you mentioned do this. This allows the system to keep working under X number of failures. Redundancy/coding allows you to plan on scheduled maintanence, where you simply go in and swap out drives that have gone bad after the fact, rather than running around like a chicken with its head cut off every time a drive goes belly up.
My question is: how in the world does Amber LIT survive? Sure, they plead with users on their website to only use the conversion tool on content that they have legally purchased, but according to DMCA et al, users don't have the right to convert content to other formats. And creating tools to circumvent DRM is explicitly criminal according to the DMCA. So how long before Amber gets a cease-and-desist? And then what will our friend Mr. Gartenberg do?
There has also been some discussion about all this n the p2p-hackers mailing list.
What will you do when you 2yr old turns into a 12yr old who robs little old ladies at gunpoint for heroine money, all because they saw Daddy do it to the movie studies when they were little? hmm?
Kademlia isn't a protocol per-se. If you read the Kademlia paper, you'll see that they do outline the system very well, but as for specifying the way bits/bytes are packaged up into messages, etc., there's no information. So if two different authors implement a kademlia-like DHT, you can be pretty confident that they won't talk to each other unless they've both agreed on the low-level details of the messages.
Why not use Shamir secret sharing to hold onto the private key? You can choose N people to hold pieces of the private key and choose K = N such that [any] K individuals can reconstruct the secret (but not any less than K).
So why is this not easily accessible at www.fsf.org?
Does anyone know what this thing looks like? Surely it involves more than emailing the maintainer and saying "I assign the copyright of my contribution to the FSF?"
In that light, what does LSB buy me? An easy escape route for my customers to switch distros.
(used Redhat in this example, but I think it pretty much applies across the board).
Would anyone at MIT have entertained, even just for a moment, traveling to California to do something similar to the California Institute of Technology? I doubt it.
Not to say that Caltech isn't one of the top engineering schools in the country: of course it is. But it doesn't enjoy MIT's prestige, and these pranks just go to enlarge that prestige.
For example, consider the case of directional antennas (which 802.11s doesn't concern itself with). A directional antenna must be pointed at another antenna. If there is more than one available place to point it, the decision of which antenna to point to should take many factors into consideration, including: 1) probability of being able to make a connection, 2) signal strength, which corresponds to the bandwidth that will be available, 3) the destination of traffic traveling from either end, 4) the priority of traffic destined for a particular address, etc.
Note that the act of pointing the antenna is a link layer responsibility. Also notice that while #1 and #2 could be done entirely at the physical/data link layers, #3 and #4 live way up the stack. Finally, note that #3 and #4 probably should have more to say than #1 or #2 -- it does us no good to point at B if we only have data destined for A (unless B can route it for us). In a traditional network stack, there is now way layers 1 or 2 could get the information they need to make this decision. In a mesh network there is.
Similar issues exist in omnidirectional networks. The RoofNet group at MIT, for instance, has a nice paper explaining that shortest path is about the worst thing you can do in a dense wireless network (you lose signal strength/power by making a direct link across the network, and thus have much less bandwidth than if you instead transmitted to the closest antenna between the two endpoints and had it relay to the next closest, etc)
Backspace didn't work out of the box? What did you install, "Larry's Linux Distro Extraordinaire"?
Sorry, I don't mean to trivialize your problems, but I am seriously shocked -- no redhat, SuSE, or Debian distro that I know of in the past 4 years would have any of the problems you describe.
On the other hand, I have had to press magic F8 buttons in the past 2 years while installing NT, just to get a standard 3com NIC driver installed (and the Dell instructions warned that if I installed the video driver first, the NIC driver wouldn't install properly!).
I'll admit though, that the last time I installed XP, it was fairly painless. And I do have to jump through a hoop or two to get my cheapo scanner working under Linux. But for the most part, it's stick in the CD, click on my locale and install class, and let 'er rip.
1) Linux device drivers are a big problem
and
2) Putting Windows PnP in Linux would be an easy task
I have a problem with #1 because, well, I haven't had a problem with device drivers for years. The first thing I do with a new computer (and I've gone through 5, from Dell and HP, in the last few years) is reformat, install Windows, and then install Linux. Guess which one is easier to install? Guess which one requires special driver disks and arcane "press-F8-at-the-right-time-during-the-install" crazieness to get things working? That's right: windows. With Linux, stick the CD in, click a few buttons, and done.
The problem with #2 should be obvious to everyone: one of the main tasks of an OS is to manage devices. Look at the code in the kernel that does this. Sure, there's other important stuff (vfs, memory management, process management, etc), but if you count the lines, the heaviest piece of the OS is device driver management. Ripping this out and sticking in Redmonds garbage would be disastrous.
Now, user-mode linux is a different beast. Even virtualizing the hardware could get things to work correctly under Dvorak's scheme without so much effort. But what he suggests is not only ludicrous, its outright silly, and really illustrates how out of touch he is with how technology works.
I do like his vision of robots run amok trying to destroy humanity a la Will Smith, though. There's some good thinking.
This was nice for when X locked up (which it did occasionally) or for when the network was down or if I didn't have another box to get network access from -- I could still telnet in through serial, kill X, and be good as new.
The only hitch was that most distros didn't come with the serial port enabled for this in the first place.
Actually, that isn't correct. Hofstadter makes a lot out of self reference, and certainly Zen makes a big deal out of it. But here's the kicker: there are an infinite number of incompleteness proofs, and many of them don't rely on self-reference. For instance, determining that a particular Diophantine equation doesn't have a solution for P(n)=0 is provably undecidable. Also, Chaitin's Number (which he calls Omega) also provides an interesting proof of Incompleteness using Algorithmic Information Theory instead of self-reference. Turing's Halting Problem is also a proof of incompleteness, though you could make a case that it uses some variant of self-reference (a computer program examining its own code).
The point is, mathmatical incompleteness is not a result of a paradoxical statement. In fact, the surprising result of all of this is that some mathematical facts are true by pure chance, and that the vast majority of facts in math (or in any rational system) occur randomly, and not because of derivation from some primary truth.
Math exists only in the mind. The symbols we use to do math may represent quantities in the real world, but we use symbolic manipulation to turn these symbols into more symbols, which we then project onto the world. But the world operates without these symbols. The stones or oranges or apples that you count to try to prove 1+1=2 are completely ignorant of the math involved, as is all substance that makes up those objects.
But I'm just rambling. It is true, however, that integer arithmetic is provably incomplete (which means that there are infinite "truths" in such a system which are not provable or derivable from any sort of rational process), or provably inconsistent (in which case the system is just broken, producing erroneous or conflicting results). Godel's Incompleteness proofs considered just arithmetic operations on the natural numbers and showed, irrevocably, that this was the case.
I once believed that math was intrinsically true, and that not even God could refute 1+1=2. Now I refute it.
The universe is unreasonable, the world is irrational, and there is no "Theory of Everything" (Grand Unification Theory, etc) or simple explanation for how the whole thing really works.
And, I can prove it. Or, better stated, Godel already did. Others have also provided proofs (Turing, Chaitain) or at least alluded to the incompleteness of rational thought (Rusell, Epimedes, etc).
BYU Internet Security Research Lab
Urbana-Champaigne Database and Information Systems Laboratory
I'm convinced that this isn't really a problem for most uses of identification technology -- the problem is not that one guy can get 10 or 15 or 40 IDs, its that one guy and some software can get 1,000 or 100,000 or 100,000,000 IDs. An online unique identification system that solves the former and not the latter is "good enough" for most online collaborative applications that require any sort of integrity of "voting."
But that's just my opinion.
Can someone please confirm that this is the Darl McBride of SCO (what is the first letter of Darl's middle name, for starters)? For more than one reason, I believe that if it really were him, he wouldn't have his number listed (remember, the guy hires an ontourage of body guards to protect him against 'militant linux extremists' whenever he leaves his little enclave). If slashdot readers ruthlessly prank this number without confirming who it is they are really calling, its likely to backfire when open source supporters are portrayed as unjust lunatics.
And if they guy isn't the Darl Mcbride, hasn't he suffered enough already by sharing the same name?
And since Lemma 1.7 says "no communist is worth his own weight in dog excrement," it naturally follows that Linux must have originated elsewhere.
I propose one of the following:
I think you'll see the logic in all of this immediately.
The thing that is exciting about this is the field of research that it opens up. Of course, directional comm antennas have been around for quite some time, but building networks out of them is relatively new. Do a literature search on "active topology formation" and you'll see what I mean -- not a whole lot has been done here, and yet this dramatically changes how you architect networks.
The exact issues involved are what spurred DARPA to recently proclaim that it was time to revise the 7-layer OSI model into something that appears more like a mesh. And that really is what will be required -- for example, waiting 10s of seconds for OSPF to figure out whether a link is up or down just will not cut it when you can form links on the order of milliseconds.
What does adding voice input from the card's owner do? Not a whole lot, except that now, instead of only needing to physically have the card in your hand, you also have to physically sound like the owner (or have a good recording of the owner speaking his password).
Is this beatable? Absolutely. But the thing to remember is that it is significantly less beatable than the current system. And since there is no such thing as unbreakable authentication, that's about the best you can ask for. If this system works, it eliminates most fraud, because most credit card theft is performed by complete strangers, i.e., people that don't have access to recording my voice or physically swiping my card.