The best parallel algothims get sorting time of O(n log(n))
Wrong. Radix and bucket sort algorithms sort in O(n). And, you can parallellize them using divide and conquer method. Thus, you can get O(n/m) + some overhead (which turns out to be O(n/m) as well) for merge where m is the number of parallel cores.
Dear flynt, no need for personal insult. Let me start with the word 'illiterate'. English is not my native language (how many languages do you speak?), as you may guess, but even I know what it means. One of the meaning stated by the Oxford Dictionary is 'ignorant in a particular subject or activity.' And, that is exactly what I mean. They give such statistics every now and then, they should consult to an authority about statistics before making the news public.
As for the error analysis: I do this for living (got PhD from MIT in experimental physics - yep I was a keen observer for a while). When you present data, you should always evaluate its error. If you don't have the error, then don't show the data, OR simply approximate it. At that point all we can do is approximation. Then, we have to think about 'the' measurements, about the 'system'. There are precise measurements and there are accurate measurements. They are not the same. This is the first rule of the error analysis. If the individual measurements are not accurate (just think about the leap year, weather conditions, etc..), increasing the sample size won't help you. It is like measuring a distance thousand times with a poorly calibrated meter-stick; you will never get an accurate measurement. The simplest I could do with the data at hand is assuming that we are counting the number of miles an individual travels during the course of the year. Is it good? Probably not. Can I (you) do better: not with the data at hand, or without more assumptions.
However, as an experimentalist, I can bet that the statistical fluctuations of these data will be at the order of 100 miles.
It is amazing that how news reporter are so illiterate when it comes to math and science. The figures they report indicates no conclusive reduction (change) in driving habits at all. When we measure a quantity there is always some error in that measurement. Driving habit is pretty much a random event when average over millions of people. Thus, a good approximation (I don't claim it to be the most accurate) of its associated error is 1/sqrt(x), and thus x +- x/sqrt(x) is a proper way for indicating the average x. In this specific case, it is 13600 +- ~110. Thus the change (~50) is smaller then a sigma away from the average. The proper way to report this would be: Americans don't care about the environment, and they have the money to burn the same amount of fossil fuel as they did last year.
Not only because of political o economic reasons, but also plain geographical ones: Turkey is not in Europe, but in Asia. I have always wondered why is there so much fuss about Turkey joining the EU and not about Morocco or Egypt, for example (same proximity, similar cultural ties).
The odds are you've never been in all of these three countries:
Turkey, Egypt, and Morocco, and/or you have no idea about their caltural ties. There is a saying by a famous Turkish writer (let me try to translate for you):
"Those can not have an opinion before having
the knowledge, first."
As for the membership, as a Turkish citizen, I'm equally against Turkey's membership, but, because of entirely different set of reasons. I think we should strengthen our ties to the east, starting from China...
Anyone have any ideas on how to back up 1TB in a home environment? i.e., not $3000 tape drives & $200 tapes
Somebody already mentioned this on this thread: Lacie - Bigger Disk is a TB disk with a price tag of $1200.
It is more expensive than the system proposed above, but
it is less hassle (a nice case), and can easily be chained by
several units as your demand increases. It is probably faster than software RAID with ATA/133. And, it is cheaper
than a true raid (say Apple Xserve RAID - 4 times cheaper!), and its performance -even with a few chained units- is definitely sufficient for a home user who has less than a few running processes.
And, don't forget the labor of putting 10-20 HDs together, also, include a possible CPU/board price if you have to put it in a
separate case. in addition to this, add the possible cost of hardware failure due to not very well designed system (air flow, power requirements, etc...)
And, you can software-mirror those units and construct software RAID if you wish.
I think it is the best solution for a serious user who values its data.
b) Have a separate and intelligent module for rendering badly coded websites that dont follow specs
Off course, it is considered to be good that a browser try to correct the badly written pages.
However, one should also consider the case that if no browser supports that kind of ugly pages, the authors of those pages will be forced to correct their mistakes all the time, once they try and see that their code is just plain junk.
Then, the browser maintainers will not try to write intelligent codes, whose definition is open ended, and that will not always work depending on the badness of the HTML syntax they may face.
I would prefer this kind of approach; punish the bad syntax, and I also realize that this is just a naive way of thinking, because there will be always evil software companies (everyone know who it is!) that will try to support the non-standard to gain more costomer, and even more, they will introduce their set of standards on top of this, at which point the others will have to follow it in order not to let this company to be the sole player.
Our department has Konica 7165 copier, which has the scan-to-email capability. It can e-mail the scanned document as multi-page-pdf or tiff files, thus you don't need to convert it to pdf page by page.
And, use a low resolution setting (say 100dpi) for handwritten documents. It will do just fine. Pdf (depending on the driver though) compresses the image. If you use a machine something like the Konica, try to set the threshold/brightness to a level such that the empty portion of the pages will appear as plain white; this will increase the compression ratio significantly.
So, my recommendation is that try to find a Kinkos which has this type of machine. If you can't find, just tell the professors that it is simply not a reasonable task that can be done in finite time.
Re:His comment on Slashdot:
on
More From Tanenbaum
·
· Score: 5, Interesting
Minix... wasn't really good for learning the architecture of an OS.... and people could goof around with it for free
I think, if a professor of mine would write an operating system for me to have me the freedom of goofing around, I would call it a very good source.
And, even if the OS itself was not so good, it doesn't matter, because people often learn the subject from the mistakes, shortcomings. After all, Linus wrote the Linux to improve Minix on Minix.
How nice to see this discussion at this point! 5-10 years ago people were dreaming digital images that are as good as 35mm films. Now, we dare to compare the digital images to the large format (4x6 - 8x10") films.
I think, this alone summarizes what is going on here, and what will be the position of the digital imaging in 5-10 years!
This is the hidden problem with debit card fraud. Until the matter is resolved, you *are* out
Not true. Once I withdrawn $200 using my debit card from a 3rd party ATM (belongs to well known bank other than my bank), but it didn't eject any cash although it charged my account (broken mechanics, broken software ?)
I called my bank the same day, and they reversed the transaction immediately and told me that it will stay this way, until the issue is investigated. 2 weeks later I received a letter stating that my claim was approved.
Maybe we are at the mercy of the banking institute, but I should note that my bank (Fleet Bank) is not so innocent considering how they are using their monopoly power in our area. So, I would be surprised if this kind of solution is not mandated by law.
By the time 1000000000 users have asked for the 10 kg barcode being sent to them, they will probably consider more reasonable means of distribution.
Nope! That's the point. They can charge for the cost of media and the shipping, I think that much is clear from the GPL (correct me if I'm wrong.) And, the purpose is not to distribute the code. If they "decide" to distribute it on punch-cards, it is going to cost 5,000,000 cards, and the mailing costs and the (punch-card) equipment maintenance to punch that many cards. A rough estimate gives at least a million dollar! It is big enough to discourage the people.
There is an invention called PEN which can be used to make modification on a printed material, it can even be considered as a machine, given the fact that they became pretty sophisticated these days! One can convince the judge with this... In general I find GPL childish... If you want to make it free, make it really FREE.
There are two solutions for the spam mails. One, new laws that punishes the spammer significantly, two, technology.
The first solution may shift the business to off-shore, that only reduces the spam rate. One of the technology which prevents the spam that comes to my mind first is the public/private key exchange. It is a little hassle, but people can get used to it quickly if they have to, with a nice interface. It is the initial forced-switch that makes it hard to realize.
Today, the only way to make one of these happening is making the e-mail COMPLETELY in-useful, that way it may get the attention of those who do not use it often, or who don't use at all, but have the power.
For example, if everybody (say most of us) becomes a spammer, and starts to send (at least try to send) a million spam mails a day, nobody can check the real mails they may get within hundreds of spam mail in their mailboxes. And, due to the load on the internet, all systems start to slow down. At this point a solution becomes ABSOLUTELY necessary, and I am sure that that kind of situation accelerates a search for a better system.
So, the general solution is making the current e-mail protocol out of control to the highest degree possible.
Ooo! There is no doubt that ECC memory is useful. What I tried to stress is the fact that poison distribution doesn't apply here if the program you're using is smart enough. And, when there is ECC, and we have it by the way, the overhead of non-correctable memory errors is zero for all practical purposes.
And, without ECC when an error occurs in the executable part or in the logic-decision data, it is more likely that the application will crash, which is perfectly fine as we can continue from the last saved state. And, just for amusement, even the error occurs in the data part of the memory, at least for our case, it would be just another source of randomness we appreciate for a Monte Carlo simulation:)
Thus, the worse case requires a factor of two correction even with the worse memory chips that one can get, not the poison statistics. And, my guess is that, most of the users will prefer to run the application only once, as the executable part of the application often requires small amount of memory (means smaller possibility of errors), and the large data is mostly prone to single bit errors.
Though, I understand that it may depends on the application. And, the thermally well design G5 system makes this risk worth to take.
And, although I can't imagine how, Varadarajan says that "We are planning on moving to ECC systems in the future." Does anyone know if the G5 memory subsystem supports ECC modules? Or, did he mean to replace the existing G5 boxes with newer generation ones that can support ECC, G5-server maybe?
This is all in theory. In practice life is not so bad. I often do long calculation using multiple nodes. And, I tell you this; memory errors is not the only source of problem when you deal with a supercomputer.
For this purpose we write programs that saves its COMPLETE state periodically. Thus, when an error occurs or we need to shutdown for some other good reason, we continue crunching from the latest good saved state forward. That way, the inefficiency is minimal.
And, all the serious programmers who writes codes for long jobs addresses this issue beforehand.
Thus, their benchmark at its face value is good for me. It is one hell of a machine!
And, an extra question for the bonus point: What would be the average time of completion of a job that requires 1 week of computation, and saves its state every 10 minutes with a "2 days per error" memory?
I have a PhD degree in Physics, I work in the area of nuclear/particle physics, and I do a lot of Monte Carlo simulation, data crunching, theoretical calculations, whatever you can think of that require mathematics at any level, and yet, I didn't need to use my TI89 for the last 10 months, I think. The last time I used it was for filling the (damn) tax forms.
When the arithmetic is simple (say "18*2^(-8/9) makes 10; an order of magnitude larger -- by assuming that 8/9 is slightly smaller than unity -> 2^(~-1) is slightly greater than 1/2 which makes the result a little more than slightly greater than 9 which is closer to 10 -- with a precision better than the experiment/theory"), we don't need any device, estimations are good enough. We need computers (well actually "super computers") when we have a tougher problem.
Although, I didn't see the HP49g+ yet, I don't think the Fourier/integral transform can be good enough for a real work. Those kind of things may be helpful only for homework unless you don't have access to a computer.
My point is that it has nothing to do with the civilized world. I prefer a sleeker calculator that does the basic operations (say simple arithmetic + trigonometry + complex numbers), rather than a clumsy TI89. Sometimes, I miss my pre-historic Casio with beautiful aluminum casing design, which I lost long ago. The only advantage of the modern calculators of today is the complex numbers, which is needed once in a while...
And, as for the area of mathematics, which we physicists use intensively, a calculator is even less needed. I often saw the amazed facial expressions of my non-technical friends who are referring the notes in my hand as: "Is it really mathematics? I don't see any number here! And, what are does curly symbols for?" Got the picture...
...features many improvements... including spam filters,... Palm synchronization and more...
I hate this. I am a relatively smart person, above average for sure, and I live within computers all day, yet these densely packed programs manage to confuse me. Does it have to support Palm synchronization or do we really need to combine the web browser and the mail client? If they feel that they are good at these areas as well, they can produce a separate product(s) which can communicate with each other, but can also work independently.
Adding more properties is not the challenge, adding them transparent to the user is the challenge. I want to see neither millions of buttons in toolbar nor millions of options when i press the 2nd mouse button on the canvas. Ex. can someone tell me why do I need to press the 2nd mouse button and choose the "back/forward" in the pop-up menu, which may appear in different locations w.r.t. the pointer depending on the pointer position, instead of just clicking the "back/forward" icon in the toolbar?
Well I am sure that someone will find an absurd reason for this, thus let me respond it beforehand: Then, I need the "Sort the lines shown on this page w.r.t. the second word on each line" item in the same pop-up window. I need this once in every about 3 months. I can hear another reply. Here is my answer: Be realistic. Most of the users do not disable the toolbar. Thus, instead of making it complex to keep the minority happy, it can be programed as a dynamic menu depending on the toolbar status. (Well, I must admit, I didn't install Netscape 7.1 yet, but this is the defacto behavior for almost all browsers I saw -- Himm... I'm not sure about Safari I should check this next time I use it.) If Netscape 7.1 addresses these kind of issues, can someone point this out for me?
The bottom line is that we need simple looking but powerful software, not a messy software with kitchen-sink included.
Slashdot Mirror
low-level 3D graphics hardware based on the OpenGL ES 2.0 API (application programming interface)...
Since when Slashdoters are not aware what "API" means? It is a changing world....
In 1954 the European Laboratory for Particle Physics was founded
That is not true as well. The official name is "European Organization for Nuclear Research".
"European Laboratory for Particle Physics" is just a common nickname which serves better.
See, http://public.web.cern.ch/public/en/About/Name-en.html for details.
So, you mean people are guilty, until proven innocent.
Wrong. Radix and bucket sort algorithms sort in O(n). And, you can parallellize them using divide and conquer method. Thus, you can get O(n/m) + some overhead (which turns out to be O(n/m) as well) for merge where m is the number of parallel cores.
> she said, 'Well, why should we believe you? You're a criminal. Criminals lie all the time.'
Constitution says, people are innocent until proven guilty. I guess, the idiot was confused "suspect" with "criminal"
As for the error analysis: I do this for living (got PhD from MIT in experimental physics - yep I was a keen observer for a while). When you present data, you should always evaluate its error. If you don't have the error, then don't show the data, OR simply approximate it. At that point all we can do is approximation. Then, we have to think about 'the' measurements, about the 'system'. There are precise measurements and there are accurate measurements. They are not the same. This is the first rule of the error analysis. If the individual measurements are not accurate (just think about the leap year, weather conditions, etc..), increasing the sample size won't help you. It is like measuring a distance thousand times with a poorly calibrated meter-stick; you will never get an accurate measurement. The simplest I could do with the data at hand is assuming that we are counting the number of miles an individual travels during the course of the year. Is it good? Probably not. Can I (you) do better: not with the data at hand, or without more assumptions.
However, as an experimentalist, I can bet that the statistical fluctuations of these data will be at the order of 100 miles.
It is amazing that how news reporter are so illiterate when it comes to math and science. The figures they report indicates no conclusive reduction (change) in driving habits at all. When we measure a quantity there is always some error in that measurement. Driving habit is pretty much a random event when average over millions of people. Thus, a good approximation (I don't claim it to be the most accurate) of its associated error is 1/sqrt(x), and thus x +- x/sqrt(x) is a proper way for indicating the average x. In this specific case, it is 13600 +- ~110. Thus the change (~50) is smaller then a sigma away from the average. The proper way to report this would be: Americans don't care about the environment, and they have the money to burn the same amount of fossil fuel as they did last year.
Don't panic. It has built-in "Omnidirectional Microphone" as the MacBookPros have. Check out the comparison chart for this: http://store.apple.com/Catalog/US/Images/compariso n_chart.html
The odds are you've never been in all of these three countries: Turkey, Egypt, and Morocco, and/or you have no idea about their caltural ties. There is a saying by a famous Turkish writer (let me try to translate for you):
As for the membership, as a Turkish citizen, I'm equally against Turkey's membership, but, because of entirely different set of reasons. I think we should strengthen our ties to the east, starting from China...
Whether there's a grand unified theory that explains everything in our universe I do not know (or remember).
Seriously; how can we distinguish whether we don't know or remember? If we can't remember, we don't know it anymore!
Somebody already mentioned this on this thread: Lacie - Bigger Disk is a TB disk with a price tag of $1200.
It is more expensive than the system proposed above, but it is less hassle (a nice case), and can easily be chained by several units as your demand increases. It is probably faster than software RAID with ATA/133. And, it is cheaper than a true raid (say Apple Xserve RAID - 4 times cheaper!), and its performance -even with a few chained units- is definitely sufficient for a home user who has less than a few running processes.
And, don't forget the labor of putting 10-20 HDs together, also, include a possible CPU/board price if you have to put it in a separate case. in addition to this, add the possible cost of hardware failure due to not very well designed system (air flow, power requirements, etc...)
And, you can software-mirror those units and construct software RAID if you wish.
I think it is the best solution for a serious user who values its data.
Off course, it is considered to be good that a browser try to correct the badly written pages.
However, one should also consider the case that if no browser supports that kind of ugly pages, the authors of those pages will be forced to correct their mistakes all the time, once they try and see that their code is just plain junk.
Then, the browser maintainers will not try to write intelligent codes, whose definition is open ended, and that will not always work depending on the badness of the HTML syntax they may face.
I would prefer this kind of approach; punish the bad syntax, and I also realize that this is just a naive way of thinking, because there will be always evil software companies (everyone know who it is!) that will try to support the non-standard to gain more costomer, and even more, they will introduce their set of standards on top of this, at which point the others will have to follow it in order not to let this company to be the sole player.
Our department has Konica 7165 copier, which has the scan-to-email capability. It can e-mail the scanned document as multi-page-pdf or tiff files, thus you don't need to convert it to pdf page by page.
And, use a low resolution setting (say 100dpi) for handwritten documents. It will do just fine. Pdf (depending on the driver though) compresses the image. If you use a machine something like the Konica, try to set the threshold/brightness to a level such that the empty portion of the pages will appear as plain white; this will increase the compression ratio significantly.
So, my recommendation is that try to find a Kinkos which has this type of machine. If you can't find, just tell the professors that it is simply not a reasonable task that can be done in finite time.
I think, if a professor of mine would write an operating system for me to have me the freedom of goofing around, I would call it a very good source.
And, even if the OS itself was not so good, it doesn't matter, because people often learn the subject from the mistakes, shortcomings. After all, Linus wrote the Linux to improve Minix on Minix.
I think, this alone summarizes what is going on here, and what will be the position of the digital imaging in 5-10 years!
Not true. Once I withdrawn $200 using my debit card from a 3rd party ATM (belongs to well known bank other than my bank), but it didn't eject any cash although it charged my account (broken mechanics, broken software ?)
I called my bank the same day, and they reversed the transaction immediately and told me that it will stay this way, until the issue is investigated. 2 weeks later I received a letter stating that my claim was approved.
Maybe we are at the mercy of the banking institute, but I should note that my bank (Fleet Bank) is not so innocent considering how they are using their monopoly power in our area. So, I would be surprised if this kind of solution is not mandated by law.
Nope! That's the point. They can charge for the cost of media and the shipping, I think that much is clear from the GPL (correct me if I'm wrong.) And, the purpose is not to distribute the code. If they "decide" to distribute it on punch-cards, it is going to cost 5,000,000 cards, and the mailing costs and the (punch-card) equipment maintenance to punch that many cards. A rough estimate gives at least a million dollar! It is big enough to discourage the people.
I think it is a pretty good idea.
There is an invention called PEN which can be used to make modification on a printed material, it can even be considered as a machine, given the fact that they became pretty sophisticated these days! One can convince the judge with this... In general I find GPL childish... If you want to make it free, make it really FREE.
The first solution may shift the business to off-shore, that only reduces the spam rate. One of the technology which prevents the spam that comes to my mind first is the public/private key exchange. It is a little hassle, but people can get used to it quickly if they have to, with a nice interface. It is the initial forced-switch that makes it hard to realize.
Today, the only way to make one of these happening is making the e-mail COMPLETELY in-useful, that way it may get the attention of those who do not use it often, or who don't use at all, but have the power.
For example, if everybody (say most of us) becomes a spammer, and starts to send (at least try to send) a million spam mails a day, nobody can check the real mails they may get within hundreds of spam mail in their mailboxes. And, due to the load on the internet, all systems start to slow down. At this point a solution becomes ABSOLUTELY necessary, and I am sure that that kind of situation accelerates a search for a better system.
So, the general solution is making the current e-mail protocol out of control to the highest degree possible.
And, without ECC when an error occurs in the executable part or in the logic-decision data, it is more likely that the application will crash, which is perfectly fine as we can continue from the last saved state. And, just for amusement, even the error occurs in the data part of the memory, at least for our case, it would be just another source of randomness we appreciate for a Monte Carlo simulation :)
Thus, the worse case requires a factor of two correction even with the worse memory chips that one can get, not the poison statistics. And, my guess is that, most of the users will prefer to run the application only once, as the executable part of the application often requires small amount of memory (means smaller possibility of errors), and the large data is mostly prone to single bit errors. Though, I understand that it may depends on the application. And, the thermally well design G5 system makes this risk worth to take.
And, although I can't imagine how, Varadarajan says that "We are planning on moving to ECC systems in the future." Does anyone know if the G5 memory subsystem supports ECC modules? Or, did he mean to replace the existing G5 boxes with newer generation ones that can support ECC, G5-server maybe?
For this purpose we write programs that saves its COMPLETE state periodically. Thus, when an error occurs or we need to shutdown for some other good reason, we continue crunching from the latest good saved state forward. That way, the inefficiency is minimal.
And, all the serious programmers who writes codes for long jobs addresses this issue beforehand.
Thus, their benchmark at its face value is good for me. It is one hell of a machine!
And, an extra question for the bonus point: What would be the average time of completion of a job that requires 1 week of computation, and saves its state every 10 minutes with a "2 days per error" memory?
Good work!
When the arithmetic is simple (say "18*2^(-8/9) makes 10; an order of magnitude larger -- by assuming that 8/9 is slightly smaller than unity -> 2^(~-1) is slightly greater than 1/2 which makes the result a little more than slightly greater than 9 which is closer to 10 -- with a precision better than the experiment/theory"), we don't need any device, estimations are good enough. We need computers (well actually "super computers") when we have a tougher problem.
Although, I didn't see the HP49g+ yet, I don't think the Fourier/integral transform can be good enough for a real work. Those kind of things may be helpful only for homework unless you don't have access to a computer.
My point is that it has nothing to do with the civilized world. I prefer a sleeker calculator that does the basic operations (say simple arithmetic + trigonometry + complex numbers), rather than a clumsy TI89. Sometimes, I miss my pre-historic Casio with beautiful aluminum casing design, which I lost long ago. The only advantage of the modern calculators of today is the complex numbers, which is needed once in a while...
And, as for the area of mathematics, which we physicists use intensively, a calculator is even less needed. I often saw the amazed facial expressions of my non-technical friends who are referring the notes in my hand as: "Is it really mathematics? I don't see any number here! And, what are does curly symbols for?" Got the picture...
Here is my back of the envelope calculation:
- $65,994 2.52 TB Xserve RAID (6x$10999)
Where did I make the mistake?$45,117 into someone else's pocket
--------
$111,111 Subtotal
-$11,111 Education and volume discount (10%)
--------
$100,000 Total
I hate this. I am a relatively smart person, above average for sure, and I live within computers all day, yet these densely packed programs manage to confuse me. Does it have to support Palm synchronization or do we really need to combine the web browser and the mail client? If they feel that they are good at these areas as well, they can produce a separate product(s) which can communicate with each other, but can also work independently.
Adding more properties is not the challenge, adding them transparent to the user is the challenge. I want to see neither millions of buttons in toolbar nor millions of options when i press the 2nd mouse button on the canvas. Ex. can someone tell me why do I need to press the 2nd mouse button and choose the "back/forward" in the pop-up menu, which may appear in different locations w.r.t. the pointer depending on the pointer position, instead of just clicking the "back/forward" icon in the toolbar?
Well I am sure that someone will find an absurd reason for this, thus let me respond it beforehand: Then, I need the "Sort the lines shown on this page w.r.t. the second word on each line" item in the same pop-up window. I need this once in every about 3 months. I can hear another reply. Here is my answer: Be realistic. Most of the users do not disable the toolbar. Thus, instead of making it complex to keep the minority happy, it can be programed as a dynamic menu depending on the toolbar status. (Well, I must admit, I didn't install Netscape 7.1 yet, but this is the defacto behavior for almost all browsers I saw -- Himm... I'm not sure about Safari I should check this next time I use it.) If Netscape 7.1 addresses these kind of issues, can someone point this out for me?
The bottom line is that we need simple looking but powerful software, not a messy software with kitchen-sink included.