Right, but as with virtually every other slashdot poster, you make the mistake of assuming you are representative of the marjority of computer users. You very definitely aren't.
MOST people on dialup who see a dialog box pop up saying "Microsoft thinks you should download a 100 meg update right now and restart your computer" click "go away right now and never bother me again"
If everyone was like you, we wouldn't see CNN stories about massive worm outbreaks, and the percentage of slashdot posts bitching about Microsoft security problems would plunge from 25% all the way down to 20%.
Being or not installed by default, in corporate networks you are not machine administrator, such, your automated update goes to hell...
The machine should be initially set up by the administrator(s). During that initial setup, the update policy should be established. If you choose anything other than "download and install automatically", then you as the administrator are accepting the burden of ensuring that the machine stays updated. The fact that the average user shouldn't log in as their machine's administrator doesn't mean that their machine shouldn't have an administrator.
if there is an XP-home (without "domain" concept) why has this version the netbios open and running (think: if it's for home alone PC's, why oh why open the shared resources???)
"Home" doesn't mean "Home alone". It's not at all uncommon for an ordinary household to contain more than one computer (desktop and laptop, or multiple desktops in different rooms). These users will frequently want to share their files, and as average home users, they are less informed about how to managed a permissions-based file sharing setup than "professional" users would be. Also, it's more reasonable to assume that other computers in you home are trusted than it is to assume that other computers in your office are trusted.
I'm not saying it doesn't present security risks, but for most people (and any OS which is trying to be the dominant desktop OS should be marketed towards "most people") convenience far outweighs security.
I generally agree with you about the importance of studying Calculus, but I still think this guy was right. If the reason to take Calculus is that it's really important for everyone, not that it's necessary to study Comp Sci, then it should be a required class, not a Comp Sci prerequisite.
Yes, it's important. Yes, everyone should learn this. But no, these two subjects are not dependent, and your requirement fails to catch the large percentage of people who don't take Comp Sci.
Computer Science is not about programming. It's about finding solutions to problems using computers
Nope, that's Computer Engineering.
Computer Science is about studying computation (feel free to Google "definition of Computer Science"). A good analysis is "Comp Sci is to Physics as Computer Engineering is to Mechanical Engineering"
Science is studying how the world works, Engineering is taking the knowledge that Science has given you and using it to solve problems.
In theory, no acadmeic field is supposed to prepare you to be a skilled worker upon graduation. The theory is, you study basic principles in school (be that high school, undergrad, or grad study) then you go to work somewhere in an entry-level (by definition, requiring no specialized skills and experience) position, and your foundation of knowledge better prepares you to understand and absorb the training and experience you get at the specific job.
That was the theory. In reality, employers need employees, and students need jobs, so there's pressure on schools to teach their students practical work-related skills. This is addressed somewhat by having different programs of study, one for those who want to study the science and theory, then another for those want to learn foundations which will prepare them for employment. Again, take Physics - which tends to lead its graduates to further grad study or employment in research programs, and Mechanical Engineering - which tends to lead its graduates to work for engineering firms.
Unfortunately, because Computing is a newer field than, say, Physics and Chemistry, it hasn't really found this kind of academic equilibrium yet. Several schools are starting to offer Computer Engineering degrees (I myself graduated with a B.S. in Computer and Systems Engineering) and courses in specific programming. This is not to say that people shouldn't study Computer Science, simply that most of those studying it today expected it to be about programming, not about Computer Science.
Bottom line, Computer Science as a field is not for people seeking to become programmers. It's just that all those people have been going to Comp Sci because they either didn't know better, or because the alternatives weren't there yet. As the field matures, I would expect more students to learn the difference between Computer Science, Computer Engineering, and Programming, and then to choose their field accordingly.
Copyright can't stop them, since it is a fact that the answer to question #115 is letter D, and you cannot copyright a fact.
That's some quality reasoning there. I should publish a book:
Page 1: It is a fact that page 1 of Harry Potter book 1 reads thusly - [verbatim copied first page of Harry Potter book 1]
Page 2: It is a fact that page 2 of Harry Potter book 1 reads thusly - [verbatim copied second page of Harry Potter book 1]
I could sell it for just 50 cents above the printing and distribution costs. I'm sure at that price it would sell pretty well. And when J.K. Rowling tries to sue me, I'll just tell the judge that I read on slashdot that you can't copyright facts.
Wouldn't it be better if the cost of the average computer came down instead of the minimum hardware spec going up?
The cost of the average computer is coming down. Back in '98 it was considered a good deal to find a retail machine (think Best Buy, Dell, Gateway, not assembling your own machine from parts you bought on eBay) for under $1,000, and that was with no monitor. These days you can get a Dell machine with a 17" flat panel and a P4 (not Celeron) for $500.
I suppose it would be nice if the minimum hardware spec didn't rise, and I didn't have to buy new computers, and if the prices plunged. But right now I want "better" software (games with better graphics, browsers that run more tabs and windows, more features, etc...) and I want a faster machine to run my existing software. I can't really be in the market for a faster machine while complaining that the average machine is getting faster. I'm helping to drive that trend, and not because Microsoft forced me to.
If the present trend that I've observed continues, however, we won't see 6Ghz in 2006.
1) The slashdot problem of anecdotal evidence. If you're looking to refute someone else's assertion, don't look at your data (aka "one data point") look at something like the mid-level Dell desktops. Of course, this is hypocritical of me, because I posted a similar progression of my personal PC specs in a different comment.
2) The new Prescott core. Intel has been using the Northwood core for a few years now, and was running up against challenges to clock-rate increases. The Prescott core, while slightly less efficient on ops/cycle, has way more headroom for speed increases. So the 3.4 Prescott may run like your 3.2 Northwood, but I imagine they'll have 5-6 GHz Prescotts in not too long.
Oh it's inevitable sure, but for now it's rediculous.
Right, but I'm pretty sure they're not releasing Longhorn now, they're releasing it later. "Later" being the time when things which are inevitable happen.
They're saying the first beta for next year, which says to me they're shooting for a full release two years from now. History shows it's pretty reasonable to expect the deadline to be missed, so we're talking about seeing a true retail/OEM release around the end of 2006.
Look at average computer specs 2.5 years ago. Compare with today's. Extrapolate forward. Their numbers don't seem unreasonable to me. In 2000, I considered my desktop to be high-end, with its 128 MB of RAM. 2002 I bought a machine with 256 MB, thinking that would be adequate. End of last year, I upgraded my desktop to 512 MB, to make it adequate again. Now I'm considering the purchase of a new machine, and I want a gig of RAM to be high-end again.
From where I'm sitting, their numbers seem fairly reasonable. The terabyte disk space seems unnecessary, since most people these days use less than 40 gigs, but what do I know. At worst, their guess is off by six months to a year, it's certainly not flat out ridiculous.
I'm not seeing a lot of technical details on the implementation here, but the key seems to be that it has dynamic properties where conventional antennas have static ones. The properties in question seem to be directionality and bandwidth, but I'm not an expert in antenna design so I can't speak to the details on the band-shifting.
Looking at it generically though, I find myself wondering if this could be used to implement a similar concept to port-knocking: have the device receving only on a particular band, then jump to another one after receiving a communcation. Repeat a few times, finally settling on the desired communications band for full conversation. Again, not being an expert in this field, I can't speak to the value of that procedure, but I figured the idea was worth mentioning. (cue the posts telling me it wasn't)
A somewhat directional antenna may or may not increase security slightly, but not at an interesting cost/performance ratio.
I agree that this is not efficient for low-end security, and thus will almost certainly not be implemented in home APs or even most corporate networks, but there are some organizations (think CIA/NSA/Army) who have applied a large amount of cost already, and are happy to get more performance even at extreme price. Yes, secure protocols and encryption are a much more efficent way to prevent people inserting themselves into your communication, but they don't prevent someone from recording all your transmissions and cracking them later, using massive resources (again, think NSA).
Given a choice of "$200 buys you 80% security, $2,000 buys you 90%, and $2,000,000,000 buys you 99%" most home users will pick the $200 option and most companies will pick the $2,000 option. That doesn't mean there's nobody who picks the two billion dollar option.
A security system that is 99% effective is not much better than a system which is 10% effective
Actually, it's 89% better, which is a lot. If you think there is any security system that is 100% effective, you're naive.
There are assumptions like "if the intruder can do [X], then they can break in" and to improve your security you try to make [X] exceptionally difficult and unlikely. If you're really good, you can narrow [X] down to "knows the password", but even that isn't good enough for most organizations. Even password protected machines are usually placed behind firewalls, so that random people distributed out in the world can't even communicate with the machine in question.
That's analogous to the directional transmission: it adds a physical necessity, much as the "behind the firewall" and "local network vs. internet" layout adds the necessity of being phyisically in the building (assuming the firewall and VPN are well-secured). Yes, it's true, if the machines on the internal network have no password protection at all and allow anonymous logins from any addresses, then they are still unsecure and exposed to attack, but that doesn't mean that the firewall adds no security value.
Some content filterng programs use lists of sites, categorized. This allows you to implement the kind of nuanced policy that you're talking about, (say block Sex/Acts but allow Sex/Education) but has the weakness of allowing anything that hasn't been put on the lists. If someone starts a new porn site, it won't be filtered until the fitering software list is updated. Also, this runs into trouble when a site (say, Geocities or images.google) hosts a huge vareity of content, and blocking the site will block all of it.
The alternative is dynamically filtering pages based on words that appear in it. That's the approach being used here, but sometimes it's tricky to differentiate between a medical site talking about penises and a porn site talking about penises, just based on which words are on the page. Some material is ambiguous enough that different humans have a hard time agreeing whether or not it's porn. Good luck getting a fast algorithm to do better.
If you're going to re-trample this same ground, at least think about it a bit more. It's pathetic that there are so many posts modded up which are one or two lines saying "Gates is really rich, so 800k doesn't mean much to him." and a few more posts saying how we should fine him in proportion to his net worth, so it'll actually discourage him.
This is idiotic. 90% of the posts don't even refer to what rule he violated, simply to the quantity of the fine. You don't give 10 years in prison for a parking ticket, regardless of how much you dislike the offender. His "crime" here wasn't that he launched some anti-competitive hostile takeover of some open-source small-business, it's that he bought some stock (some, not a controlling interest) and didn't officially notify the government about it. This is a mistake, but not an offense that warrants docking someone a week's pay, regardless of what their salary is.
Of course, those complaining about the size of the fine aren't at all interested in the law, they're simply happy to see someone they hate getting penalized and wish it was more.
See, it's like the predictions about how many more programmer jobs will be in India 5, 10, and 15 years from now. I'll wager $1,000 that the same group generating those figures would have had a wildly different number for their 15 year projection 10 years ago than they have for their 5 year projection today.
When you're trying to predict the outcome from an action, it becomes less and less accurate as more factors become involved. When dealing with anything the size of the entire human species or the entire planet, making predictions about events this far off based on massive unpredictable shifts is absurd.
Imagine 8 years from now Pakistan gains control of a large-yield nuclear warhead and uses it against India. Let's see if the job projection is still on target. Things happen. Things change. Things you cannot possibly expect or predict.
Even assuming no major suprising events, (which of course there will be) then that change is gradual. It's not like the coastline shift means that in February everything is normal and in June California is gone. I don't expect that the situation 50 years from now will be anywhere near as dire as you say. But then, as I stated, I don't have anywhere near enough evidence to back up that expectation. Since you don't either, I think I'll opt to not totally freak out. Sanguine works for me. Feel free to choose your own path, but don't be surprised when 90% of the population declines to slash their CO2 outputs just because you predict doom and destruction.
Re:Modest proposal: Run it on Diebold's hardware?
on
Open Voting at OSCON
·
· Score: 1
Not a bad idea, but don't underestimate the cost to make the changes. Just because its free software doesn't mean it's cheap to install. I work for a government department, and on large projects the cost of the hardware being installed is often less than the cost of the guys who go out and install it. You can't ignore what it would cost to have guys reconfigure every voting station, especially since you would want to make sure those guys are reliable to avoid uncertainty of tampering.
Assuming it's a file and not an actual stream, then this will work, yes (thought I still wonder what are "fake connections").
Many sites stream from a file, not an "actual" stream. It's funny you mock his use of the term "fake" connections, then you imply that streaming an audio file isn't an "actual" stream. The term "streaming" is most commonly used to refer to the system wherein media is sent at approximately 1x to a client, which then plays that media. It does not imply that the media has to be live. Some streams allow seek commands to be sent to the server, others do not. The poster specifically refers to Shoutcast, which streams from files.
When he calls it a "fake" connection, I assume he means it's fake because you're not actually listening to the media on that connection, which is the intent of the stream in the first place. You are effectively telling the server "I want to listen to that station now" when all you really want to do is grab a chunk of data, so you can assemble it and listen to it later. That's the aspect of it that is fake.
It is hard avoid being sarcastic when someone posts a detailed technical explanation about something they apparently know nothing about.
You're posting a detailed technical explanation about something you apparently know nothing about! You just flatly assume that there aren't any servers out there that would allow the technique he names, despite the fact that there are. That indicates to me that you don't know enough about this field to be debunking the specifics offered by another poster who does. Which is not to say that you couldn't have posted:
"Since it only streams at 1x, what's the point of disconnecting and reconnecting? How could that possibly be better than just dumping the whole stream to one file and chopping it up later?"
That way, when someone answered that question with either
"Well, many servers allow you to seek farther into the stream, and fill buffers faster than 1x, even if you only playback at 1x"
or
"It's not better, the original poster made a mistake about this"
Ah, but you're missing something: If they can make such a huge profit, selling their product made with really cheap labor, then being greedy, they'll want to make more. So they'll use those profits to hire more and more cheap workers, to make more and more profits.
Now you have an entire country full of workers with better opportunities than they had before. Then as all of them improve their standard of living and all of them are employed, they start to demand slightly higher wages. The gradual improvement continues, until eventually the company decides it should go find a different country where labor is cheaper. Then that country goes through the same gradual improvement process.
Meanwhile, the first country is bitching about huge sections of their valuable sweat shop industry being outsourced to foreigners, but at this point they're also hiring their own workers to build new housing and infrastructure, which they can afford now because of all the money that the corporation brought into their country.
Given a completely open global market, and no issues or government or laws, (some countries would offer really cheap labor, but are avoided by companies because they have unstable governments or lack the rule of law) the tendency will be a raise in standard of living across the board for everyone, especially those at the very bottom. Now, in the real world, there are lots of things standing between us and that idealized version of the global market, and there are some "growing pains" present in the rapid adjustments. But that's a whole 'nother discussion...
the problem arrives when you fire electrons at a magnetic storage material fast enough (approaching the speed of light) they stop behaving in the expected way, and start producing random results.
-geoff313
if you take the HD out of the equation the limit goes away (for a while, perhaps)./dev/random to/dev/null isnt all that limited by HD speed.
-doormat
Actually, I bet dev/random would also have the problem of producing random results...
This is why a lot of the new high speed buses are serial (USB, firewire etc.)
and don't forget Serial ATA.
Your use of the word "easier" is appropriate though. If we're talking about theoretical physical limits, you'll get more bandwidth out of more conductors. If we're talking practical implementation based on current technology, it's easier to push the clock-rates on a single pipe than to push the clock-rate, architecture-efficiency, and bus-size of multiple-conductor busses.
As those clock-rates begin to approach their physical limitations, it will become more effective to improve parallel technologies than serial ones, but right now we're a long way away from those limits.
You're missing the point here. Bandwidth, or data transfer rate, isn't measured in meters per second. When we talk about data transfer speed, we're not talking about how quickly a bit can go from NYC to LA. That will just determine latency. We're talking about how rapidly you can put bits on "the line".
Obviously, if I have a message of even the tiniest size, I can't get it from NYC to LA faster than light. The real question is how MUCH data can I get there in a particular slice of time, say one second. If I have 100 pairs of fiber, then it's possible for my bandwidth across distance X to exceed (1 bit * c)/ X meters.
The article is slightly mislabeled, because it's not about data TRANSFER so much as data ACCESS. They're saying that even if you have 100 OC-192's between your sites, you can only read from disk at one end and write to disk at the other at a certain rate, which they cap at 1,000 times current speeds. Their point is that there's no point in transferring faster than you can access, 'cause you'd just overflow your buffers and drop data.
Just like the multiple fiber pairs solution I proposed, this solution could be addressed with multiple disks. Given an ideal bus, you could hook up 100 drives to one end of this pipe and 100 drives to the other end, and the transfer could occur at 100 times the cap they claim.
And even that still settles for the assumption that we're using conventional magnetic storage, and with only one head at a time. This reseach shows a very specific limit, rate at which a magnetic bit can be flipped, which is then extrapolated falsely to impose limits on other things like data transfer. Read their facts, not their conclusions.
The volume of the DVDs would probably fill the plane before the weight limit was exceeded, (sorry I don't have "exact" figures for that) but you could carry them externally, possibly mounted on top.
So at this point, my only disagreement with your figures is that you rule out double-sided, dual-layered discs. I suppose that's because the average user can't make them. Still, it's not like research on the absolute peak for data transfer is meant to reflect the average user. I say if we're proposing our 747-300 protocol to compete with their magnetic storage, we ought to go all out and use the dual-double-discs.
Of course, this isn't even considering the new blu-ray stuff...
Hmmmm... maybe I should consider building a Line In for my car cassette deck, so I can hook my portable MP3 player directly into it.
Actually, I've been wondering about this myself. Why don't more car stereos have a simple Line In? I wouldn't think it would be that difficult, or take up much space, and if would allow so much flexibility.
Sure, if you've got a tape deck, you can get one of the adapters, but what if you don't? What if you've just got a CD player (I do) and want to play a tape deck, or mp3 player, or maybe even a Discman that does mp3s. Maybe something crazier, like a police radio scanner or a an instrument (electric guitar or somesuch). My examples might be flawed, but the point is that there's a vast and varied selection of devices out there that output audio signals, and practically all of them either use 1/8 inch "headphone jack" or something easily adapted to that. I'd think it would be worth having the capability to play those things on your car stereo...
I believe he's referring to a common technique of download acceleration. Your stream will be at 1x. If you want to grab songs at 4x, you open four streams, and tell the second one to skip 1/4 of the way through the song, the third stream to skip 1/2 way in, and the fourth to skip 3/4 of the way in. Then, when each one finishes, you just join them together.
There are several programs that do this with FTP or HTTP file downloads, but I'm not familiar with the common stream ripper programs, so I'm just speculating that this is the same approach they use.
Try to keep the level of sarcasm low until you're really certain that you're right and he's wrong. Even then, it wouldn't hurt to post an informative correction rather than a bitter sarcastic jab.
Slashdot Mirror
Right, but as with virtually every other slashdot poster, you make the mistake of assuming you are representative of the marjority of computer users. You very definitely aren't.
MOST people on dialup who see a dialog box pop up saying "Microsoft thinks you should download a 100 meg update right now and restart your computer" click "go away right now and never bother me again"
If everyone was like you, we wouldn't see CNN stories about massive worm outbreaks, and the percentage of slashdot posts bitching about Microsoft security problems would plunge from 25% all the way down to 20%.
Being or not installed by default, in corporate networks you are not machine administrator, such, your automated update goes to hell...
The machine should be initially set up by the administrator(s). During that initial setup, the update policy should be established. If you choose anything other than "download and install automatically", then you as the administrator are accepting the burden of ensuring that the machine stays updated. The fact that the average user shouldn't log in as their machine's administrator doesn't mean that their machine shouldn't have an administrator.
if there is an XP-home (without "domain" concept) why has this version the netbios open and running (think: if it's for home alone PC's, why oh why open the shared resources???)
"Home" doesn't mean "Home alone". It's not at all uncommon for an ordinary household to contain more than one computer (desktop and laptop, or multiple desktops in different rooms). These users will frequently want to share their files, and as average home users, they are less informed about how to managed a permissions-based file sharing setup than "professional" users would be. Also, it's more reasonable to assume that other computers in you home are trusted than it is to assume that other computers in your office are trusted.
I'm not saying it doesn't present security risks, but for most people (and any OS which is trying to be the dominant desktop OS should be marketed towards "most people") convenience far outweighs security.
I generally agree with you about the importance of studying Calculus, but I still think this guy was right. If the reason to take Calculus is that it's really important for everyone, not that it's necessary to study Comp Sci, then it should be a required class, not a Comp Sci prerequisite.
Yes, it's important. Yes, everyone should learn this. But no, these two subjects are not dependent, and your requirement fails to catch the large percentage of people who don't take Comp Sci.
Computer Science is not about programming. It's about finding solutions to problems using computers
Nope, that's Computer Engineering.
Computer Science is about studying computation (feel free to Google "definition of Computer Science"). A good analysis is "Comp Sci is to Physics as Computer Engineering is to Mechanical Engineering"
Science is studying how the world works, Engineering is taking the knowledge that Science has given you and using it to solve problems.
In theory, no acadmeic field is supposed to prepare you to be a skilled worker upon graduation. The theory is, you study basic principles in school (be that high school, undergrad, or grad study) then you go to work somewhere in an entry-level (by definition, requiring no specialized skills and experience) position, and your foundation of knowledge better prepares you to understand and absorb the training and experience you get at the specific job.
That was the theory. In reality, employers need employees, and students need jobs, so there's pressure on schools to teach their students practical work-related skills. This is addressed somewhat by having different programs of study, one for those who want to study the science and theory, then another for those want to learn foundations which will prepare them for employment. Again, take Physics - which tends to lead its graduates to further grad study or employment in research programs, and Mechanical Engineering - which tends to lead its graduates to work for engineering firms.
Unfortunately, because Computing is a newer field than, say, Physics and Chemistry, it hasn't really found this kind of academic equilibrium yet. Several schools are starting to offer Computer Engineering degrees (I myself graduated with a B.S. in Computer and Systems Engineering) and courses in specific programming. This is not to say that people shouldn't study Computer Science, simply that most of those studying it today expected it to be about programming, not about Computer Science.
Bottom line, Computer Science as a field is not for people seeking to become programmers. It's just that all those people have been going to Comp Sci because they either didn't know better, or because the alternatives weren't there yet. As the field matures, I would expect more students to learn the difference between Computer Science, Computer Engineering, and Programming, and then to choose their field accordingly.
Copyright can't stop them, since it is a fact that the answer to question #115 is letter D, and you cannot copyright a fact.
That's some quality reasoning there. I should publish a book:
Page 1: It is a fact that page 1 of Harry Potter book 1 reads thusly - [verbatim copied first page of Harry Potter book 1]
Page 2: It is a fact that page 2 of Harry Potter book 1 reads thusly - [verbatim copied second page of Harry Potter book 1]
I could sell it for just 50 cents above the printing and distribution costs. I'm sure at that price it would sell pretty well. And when J.K. Rowling tries to sue me, I'll just tell the judge that I read on slashdot that you can't copyright facts.
Wouldn't it be better if the cost of the average computer came down instead of the minimum hardware spec going up?
The cost of the average computer is coming down. Back in '98 it was considered a good deal to find a retail machine (think Best Buy, Dell, Gateway, not assembling your own machine from parts you bought on eBay) for under $1,000, and that was with no monitor. These days you can get a Dell machine with a 17" flat panel and a P4 (not Celeron) for $500.
I suppose it would be nice if the minimum hardware spec didn't rise, and I didn't have to buy new computers, and if the prices plunged. But right now I want "better" software (games with better graphics, browsers that run more tabs and windows, more features, etc...) and I want a faster machine to run my existing software. I can't really be in the market for a faster machine while complaining that the average machine is getting faster. I'm helping to drive that trend, and not because Microsoft forced me to.
If the present trend that I've observed continues, however, we won't see 6Ghz in 2006.
1) The slashdot problem of anecdotal evidence. If you're looking to refute someone else's assertion, don't look at your data (aka "one data point") look at something like the mid-level Dell desktops. Of course, this is hypocritical of me, because I posted a similar progression of my personal PC specs in a different comment.
2) The new Prescott core. Intel has been using the Northwood core for a few years now, and was running up against challenges to clock-rate increases. The Prescott core, while slightly less efficient on ops/cycle, has way more headroom for speed increases. So the 3.4 Prescott may run like your 3.2 Northwood, but I imagine they'll have 5-6 GHz Prescotts in not too long.
Oh it's inevitable sure, but for now it's rediculous.
Right, but I'm pretty sure they're not releasing Longhorn now, they're releasing it later. "Later" being the time when things which are inevitable happen.
They're saying the first beta for next year, which says to me they're shooting for a full release two years from now. History shows it's pretty reasonable to expect the deadline to be missed, so we're talking about seeing a true retail/OEM release around the end of 2006.
Look at average computer specs 2.5 years ago. Compare with today's. Extrapolate forward. Their numbers don't seem unreasonable to me. In 2000, I considered my desktop to be high-end, with its 128 MB of RAM. 2002 I bought a machine with 256 MB, thinking that would be adequate. End of last year, I upgraded my desktop to 512 MB, to make it adequate again. Now I'm considering the purchase of a new machine, and I want a gig of RAM to be high-end again.
From where I'm sitting, their numbers seem fairly reasonable. The terabyte disk space seems unnecessary, since most people these days use less than 40 gigs, but what do I know. At worst, their guess is off by six months to a year, it's certainly not flat out ridiculous.
I'm not seeing a lot of technical details on the implementation here, but the key seems to be that it has dynamic properties where conventional antennas have static ones. The properties in question seem to be directionality and bandwidth, but I'm not an expert in antenna design so I can't speak to the details on the band-shifting.
Looking at it generically though, I find myself wondering if this could be used to implement a similar concept to port-knocking: have the device receving only on a particular band, then jump to another one after receiving a communcation. Repeat a few times, finally settling on the desired communications band for full conversation. Again, not being an expert in this field, I can't speak to the value of that procedure, but I figured the idea was worth mentioning.
(cue the posts telling me it wasn't)
A somewhat directional antenna may or may not increase security slightly, but not at an interesting cost/performance ratio.
I agree that this is not efficient for low-end security, and thus will almost certainly not be implemented in home APs or even most corporate networks, but there are some organizations (think CIA/NSA/Army) who have applied a large amount of cost already, and are happy to get more performance even at extreme price. Yes, secure protocols and encryption are a much more efficent way to prevent people inserting themselves into your communication, but they don't prevent someone from recording all your transmissions and cracking them later, using massive resources (again, think NSA).
Given a choice of "$200 buys you 80% security, $2,000 buys you 90%, and $2,000,000,000 buys you 99%" most home users will pick the $200 option and most companies will pick the $2,000 option. That doesn't mean there's nobody who picks the two billion dollar option.
A security system that is 99% effective is not much better than a system which is 10% effective
Actually, it's 89% better, which is a lot. If you think there is any security system that is 100% effective, you're naive.
There are assumptions like "if the intruder can do [X], then they can break in" and to improve your security you try to make [X] exceptionally difficult and unlikely. If you're really good, you can narrow [X] down to "knows the password", but even that isn't good enough for most organizations. Even password protected machines are usually placed behind firewalls, so that random people distributed out in the world can't even communicate with the machine in question.
That's analogous to the directional transmission: it adds a physical necessity, much as the "behind the firewall" and "local network vs. internet" layout adds the necessity of being phyisically in the building (assuming the firewall and VPN are well-secured). Yes, it's true, if the machines on the internal network have no password protection at all and allow anonymous logins from any addresses, then they are still unsecure and exposed to attack, but that doesn't mean that the firewall adds no security value.
Some content filterng programs use lists of sites, categorized. This allows you to implement the kind of nuanced policy that you're talking about, (say block Sex/Acts but allow Sex/Education) but has the weakness of allowing anything that hasn't been put on the lists. If someone starts a new porn site, it won't be filtered until the fitering software list is updated. Also, this runs into trouble when a site (say, Geocities or images.google) hosts a huge vareity of content, and blocking the site will block all of it.
The alternative is dynamically filtering pages based on words that appear in it. That's the approach being used here, but sometimes it's tricky to differentiate between a medical site talking about penises and a porn site talking about penises, just based on which words are on the page. Some material is ambiguous enough that different humans have a hard time agreeing whether or not it's porn. Good luck getting a fast algorithm to do better.
If you're going to re-trample this same ground, at least think about it a bit more. It's pathetic that there are so many posts modded up which are one or two lines saying "Gates is really rich, so 800k doesn't mean much to him." and a few more posts saying how we should fine him in proportion to his net worth, so it'll actually discourage him.
This is idiotic. 90% of the posts don't even refer to what rule he violated, simply to the quantity of the fine. You don't give 10 years in prison for a parking ticket, regardless of how much you dislike the offender. His "crime" here wasn't that he launched some anti-competitive hostile takeover of some open-source small-business, it's that he bought some stock (some, not a controlling interest) and didn't officially notify the government about it. This is a mistake, but not an offense that warrants docking someone a week's pay, regardless of what their salary is.
Of course, those complaining about the size of the fine aren't at all interested in the law, they're simply happy to see someone they hate getting penalized and wish it was more.
See, it's like the predictions about how many more programmer jobs will be in India 5, 10, and 15 years from now. I'll wager $1,000 that the same group generating those figures would have had a wildly different number for their 15 year projection 10 years ago than they have for their 5 year projection today.
When you're trying to predict the outcome from an action, it becomes less and less accurate as more factors become involved. When dealing with anything the size of the entire human species or the entire planet, making predictions about events this far off based on massive unpredictable shifts is absurd.
Imagine 8 years from now Pakistan gains control of a large-yield nuclear warhead and uses it against India. Let's see if the job projection is still on target. Things happen. Things change. Things you cannot possibly expect or predict.
Even assuming no major suprising events, (which of course there will be) then that change is gradual. It's not like the coastline shift means that in February everything is normal and in June California is gone. I don't expect that the situation 50 years from now will be anywhere near as dire as you say. But then, as I stated, I don't have anywhere near enough evidence to back up that expectation. Since you don't either, I think I'll opt to not totally freak out. Sanguine works for me. Feel free to choose your own path, but don't be surprised when 90% of the population declines to slash their CO2 outputs just because you predict doom and destruction.
Not a bad idea, but don't underestimate the cost to make the changes. Just because its free software doesn't mean it's cheap to install. I work for a government department, and on large projects the cost of the hardware being installed is often less than the cost of the guys who go out and install it. You can't ignore what it would cost to have guys reconfigure every voting station, especially since you would want to make sure those guys are reliable to avoid uncertainty of tampering.
That's right folks, you heard it here on slashdot first:
Unix boxes are superior to Windows boxes because they're more waterproof
Assuming it's a file and not an actual stream, then this will work, yes (thought I still wonder what are "fake connections").
Many sites stream from a file, not an "actual" stream. It's funny you mock his use of the term "fake" connections, then you imply that streaming an audio file isn't an "actual" stream. The term "streaming" is most commonly used to refer to the system wherein media is sent at approximately 1x to a client, which then plays that media. It does not imply that the media has to be live. Some streams allow seek commands to be sent to the server, others do not. The poster specifically refers to Shoutcast, which streams from files.
When he calls it a "fake" connection, I assume he means it's fake because you're not actually listening to the media on that connection, which is the intent of the stream in the first place. You are effectively telling the server "I want to listen to that station now" when all you really want to do is grab a chunk of data, so you can assemble it and listen to it later. That's the aspect of it that is fake.
It is hard avoid being sarcastic when someone posts a detailed technical explanation about something they apparently know nothing about.
You're posting a detailed technical explanation about something you apparently know nothing about! You just flatly assume that there aren't any servers out there that would allow the technique he names, despite the fact that there are. That indicates to me that you don't know enough about this field to be debunking the specifics offered by another poster who does.
Which is not to say that you couldn't have posted:
"Since it only streams at 1x, what's the point of disconnecting and reconnecting? How could that possibly be better than just dumping the whole stream to one file and chopping it up later?"
That way, when someone answered that question with either
"Well, many servers allow you to seek farther into the stream, and fill buffers faster than 1x, even if you only playback at 1x"
or
"It's not better, the original poster made a mistake about this"
you don't look like quite as much of an ass.
Ah, but you're missing something:
If they can make such a huge profit, selling their product made with really cheap labor, then being greedy, they'll want to make more. So they'll use those profits to hire more and more cheap workers, to make more and more profits.
Now you have an entire country full of workers with better opportunities than they had before. Then as all of them improve their standard of living and all of them are employed, they start to demand slightly higher wages. The gradual improvement continues, until eventually the company decides it should go find a different country where labor is cheaper. Then that country goes through the same gradual improvement process.
Meanwhile, the first country is bitching about huge sections of their valuable sweat shop industry being outsourced to foreigners, but at this point they're also hiring their own workers to build new housing and infrastructure, which they can afford now because of all the money that the corporation brought into their country.
Given a completely open global market, and no issues or government or laws, (some countries would offer really cheap labor, but are avoided by companies because they have unstable governments or lack the rule of law) the tendency will be a raise in standard of living across the board for everyone, especially those at the very bottom. Now, in the real world, there are lots of things standing between us and that idealized version of the global market, and there are some "growing pains" present in the rapid adjustments. But that's a whole 'nother discussion...
the problem arrives when you fire electrons at a magnetic storage material fast enough (approaching the speed of light) they stop behaving in the expected way, and start producing random results.
/dev/random to /dev/null isnt all that limited by HD speed.
-geoff313
if you take the HD out of the equation the limit goes away (for a while, perhaps).
-doormat
Actually, I bet dev/random would also have the problem of producing random results...
Well, they always said, "Gigs will never fly"
They must not have seen the posts above about 747's full of DVDs...
This is why a lot of the new high speed buses are serial (USB, firewire etc.)
and don't forget Serial ATA.
Your use of the word "easier" is appropriate though. If we're talking about theoretical physical limits, you'll get more bandwidth out of more conductors. If we're talking practical implementation based on current technology, it's easier to push the clock-rates on a single pipe than to push the clock-rate, architecture-efficiency, and bus-size of multiple-conductor busses.
As those clock-rates begin to approach their physical limitations, it will become more effective to improve parallel technologies than serial ones, but right now we're a long way away from those limits.
You're missing the point here.
Bandwidth, or data transfer rate, isn't measured in meters per second. When we talk about data transfer speed, we're not talking about how quickly a bit can go from NYC to LA. That will just determine latency. We're talking about how rapidly you can put bits on "the line".
Obviously, if I have a message of even the tiniest size, I can't get it from NYC to LA faster than light. The real question is how MUCH data can I get there in a particular slice of time, say one second. If I have 100 pairs of fiber, then it's possible for my bandwidth across distance X to exceed
(1 bit * c)/ X meters.
The article is slightly mislabeled, because it's not about data TRANSFER so much as data ACCESS. They're saying that even if you have 100 OC-192's between your sites, you can only read from disk at one end and write to disk at the other at a certain rate, which they cap at 1,000 times current speeds. Their point is that there's no point in transferring faster than you can access, 'cause you'd just overflow your buffers and drop data.
Just like the multiple fiber pairs solution I proposed, this solution could be addressed with multiple disks. Given an ideal bus, you could hook up 100 drives to one end of this pipe and 100 drives to the other end, and the transfer could occur at 100 times the cap they claim.
And even that still settles for the assumption that we're using conventional magnetic storage, and with only one head at a time. This reseach shows a very specific limit, rate at which a magnetic bit can be flipped, which is then extrapolated falsely to impose limits on other things like data transfer. Read their facts, not their conclusions.
The volume of the DVDs would probably fill the plane before the weight limit was exceeded, (sorry I don't have "exact" figures for that) but you could carry them externally, possibly mounted on top.
So at this point, my only disagreement with your figures is that you rule out double-sided, dual-layered discs. I suppose that's because the average user can't make them. Still, it's not like research on the absolute peak for data transfer is meant to reflect the average user. I say if we're proposing our 747-300 protocol to compete with their magnetic storage, we ought to go all out and use the dual-double-discs.
Of course, this isn't even considering the new blu-ray stuff...
Hmmmm... maybe I should consider building a Line In for my car cassette deck, so I can hook my portable MP3 player directly into it.
Actually, I've been wondering about this myself. Why don't more car stereos have a simple Line In? I wouldn't think it would be that difficult, or take up much space, and if would allow so much flexibility.
Sure, if you've got a tape deck, you can get one of the adapters, but what if you don't? What if you've just got a CD player (I do) and want to play a tape deck, or mp3 player, or maybe even a Discman that does mp3s. Maybe something crazier, like a police radio scanner or a an instrument (electric guitar or somesuch). My examples might be flawed, but the point is that there's a vast and varied selection of devices out there that output audio signals, and practically all of them either use 1/8 inch "headphone jack" or something easily adapted to that. I'd think it would be worth having the capability to play those things on your car stereo...
I believe he's referring to a common technique of download acceleration. Your stream will be at 1x. If you want to grab songs at 4x, you open four streams, and tell the second one to skip 1/4 of the way through the song, the third stream to skip 1/2 way in, and the fourth to skip 3/4 of the way in. Then, when each one finishes, you just join them together.
There are several programs that do this with FTP or HTTP file downloads, but I'm not familiar with the common stream ripper programs, so I'm just speculating that this is the same approach they use.
Try to keep the level of sarcasm low until you're really certain that you're right and he's wrong. Even then, it wouldn't hurt to post an informative correction rather than a bitter sarcastic jab.