These are also the same people who enjoy particpating in system administration discussions when their system administration experience only stems from the 4 boxes they have at home.
Yes, and that's how SunOS itself originally became big, then Windows NT, then Linux. Face it, people start small and then both their projects and their needs grow.
If a company can't offer good value and good performance on the low end, they are doomed because only a few, select people who happen to be running big systems will buy them. That kind of arrogance killed most mainframe vendors and nearly did IBM in.
Of course, even as far as big systems and large networks are concerned, I think Sun has very little to offer that's competitive, but that's a separate discussion.
It's never a suprised that people on slashdot just don't get Sun equipment.
I used to buy Sun machines by the dozens--back when they gave me good bang-for-the-buck and when they were the best of the UNIX workstation bunch (of course, even back then, it was GNU software that made Solaris tolerable). Today, PCs give me more bang-for-the-buck and Linux and BSD have become far better operating systems, so there is no reason to like or advocate Sun workstations anymore.
Much like Apple, companies (I'd wager extremely few people buy Sun's for personal everyday use) that buy these boxes are buying them for the OS and rarely for the groundbreaking hardware.
Sure: they buy them for the OS. But that's not because there is anything particularly great about Solaris, it's because they want backwards compatibility.
Note that people used to use Sun workstations for personal everyday use, back when they were reasonably priced, performed well, and still represented the state-of-the-art in UNIX-like desktop systems. No more, however.
I seriously doubt that there are many businesses who switch to Solaris--Sun is living off repeat business now. Almost anybody who has a choice for a new project starts off with Linux, BSD, or NT these days.
Maybe on sheer performance it will be beaten by x86 however for crunching big data sets the UltraSparc is just more effecient.
Oh, and how exactly is it going to be "just more efficient"? Crunching big data sets may be limited by CPU or I/O. In terms of CPU, this thing is slow no matter which way you look at it. And in terms of I/O, it uses the same IDE drives as everybody else, and its Ethernet isn't going to be any faster either.
15 years ago, Sun workstations were a great deal: for under $2000, you could get something that no PC, no Mac, no other machine could compete with. Today, they are overpriced underperformers. Sun is milking their remaining customers who are unlucky enough to have committed to a Solaris environment.
Also some software only runs on Solaris so for that this box is good.
Yup, and that's why people buy these things. Of course, you can get equivalent or better software for other platforms, so anybody starting out now would be a fool to buy into Solaris, but those unlucky few who are stuck on Solaris software will just have to make do with this.
Ultimately, it comes down to the fact that when you attend a university, you have a contractual agreement with the university. And that contract probably permits the university to do this sort of thing. In fact, they probably also own any research results you obtain while at the university and any software you write. Public universities are a little more restricted in the kinds of conditions they can impose on students, but they can impose conditions as well.
I would ask two other questions about this, though. First, if the professor created new, interesting problem sets every year, then students couldn't plagiarize from the Internet, and plagiarism within a reasonable class size should be obvious.
Second, this kind of effort really doesn't need a private company; universities should and could do it cooperatively among themselves for less money. And the software for that isn't hard to write either.
I suspect the reason why it isn't happening is because good professors don't need this sort of thing and professors that do need this sort of thing probably are the same ones that aren't up to setting up large, cross-university software and data collection efforts.
As a student, you have a simple choice: just don't take classes from professors that engage in this sort of nonsense. And if it is part of mandatory courses, either grin and bear it or consider whether your choice of university was less-than-optimal after all.
Too bad they didn't come up with that list in 1958 when it mattered
A "kilo" stood for 1000 in 1958 and even in 1858.
and could have actually been successfully adpoted. It's simply too late to just arbitrarily start calling 1024 bytes a Kibibyte or whatever the fuck they decided to call it
This studpidity and problem is entirely due to computer engineers and they need to fix it. As computers are moving into other areas of daily life, the conflicting terminology will only get worse and the physical prefixes, with centuries of tradition behind them and wide use throughout even the computer industry, won't change.
Using color filters for high quality digital color photography is an old technology (and even older for analog). Its obvious limitation is that the subject has to be still.
Bigger pixels at lower resolution are not necessarily a good tradeoff: you can do almost as well in terms of noise and sensitivity by using more smaller sensors and performing the averaging in software.
Compositing lots of low resolution images into a single high resolution image is also completely standard: you can get both free and commercial software to do it.
Altogether, I suspect that if you take something like the new Sony 8Mpixel camera and take raw pictures with it, and reduce it to 1024x768 using good software, you are going to be pretty close to the measured quality and sensitivity of Spirit's sensor (in practice, you'll see little or no difference under normal circumstances, however). Then, you can use compositing software to composit multiple images for panoramas.
The Spirit tradeoffs make sense for a Mars rover, also taking into account power and weight requirements, but they do not result in a level of picture quality that you couldn't achieve with the digital cameras you can buy at the local store.
People in AI have worked on automated scientific discovery for decades, and some of their systems have also had robotic components. This seems like a tweak and a good sales job, not a breakthrough.
The lesson wasn't lost on the Republicans: buy votes with taxpayer money.
That argument is bogus because the Republicans are cutting like crazy in programs whose benefits voters actually experience.
What Republicans are buying isn't "votes", which would be the right thing to buy in a democracy, what they are buying is campaign contributions which they then use to influence the media. And that is not a good thing.
The traditional "neural feedback" for ADHD involves a whack with a ruler, administered by a Catholic nun. I suspect it is rather effective, although it may have unwanted (if temporary) side-effects.
This has to be 3-4 drives in a box without replication or redundancy (since you can't swap anything). That means you just greatly increased your risk of losing a whole lot of data at once because if any one drive goes, all your data is gone.
Get a real RAID drive or separate disks and you'll have more safety and more flexibility.
It's pretty ironic for republicans to portray democrats as fiscally irresponsible: republicans have been far, far worse over the last few decades, starting with Reagan. Republicans love towaste huge amounts of money unproductively, foremost on the military and propping up unproductive industries.
Large amounts of government spending can do wonders for the economy, if citizens are willing to make the sacrifices (i. e. pay the taxes). And manned space travel, useless as it is, is at least less destructive to foreign relations and industry than wars, Bush's other favorite economic activity.
However, tax cuts and massive spending don't work. And private industry is unlikely to go into space anytime soon--it's not profitable.
Well, he's right, and it's not like Apple's monopolistic impulses are new.
To me, neither Apple nor Sun seem like a good alternative to MS: they are trying to be like MS, they are just less good at it. And being smaller, they are less restrained and more vicious once they manage to sink their teeth into a market.
How can they "fall back" on OSX? OSX is Apple-proprietary. IBM can't sell it, they can' make their own machines for it, etc. And why would they want to anyway?
If you want to measure performance, let's consider time from "hey, we need some software" to "ok, data processing complete". That's the sort of speed that matters - maybe the benchmark should have checked for that..
You are so right that programmer productivity matters a great deal. Isn't it too bad that Java doesn't do that well either?
For performance critical and numerical code, programmer productivity in Java is horrendously bad because the language has so little support for it; C++ and C# beat Java hands down in those areas in terms of programmer productivity in my experience. Trying to write numerical or high performance code in Java is an exercise in frustration and takes many times as long as in languages like C++ or C#.
And for code where performance doesn't matter so much, Java seems barely better than C++ (it's slightly better mostly because it has garbage collection) and worse than C#. In my experience, highly productive languages are dynamically typed languages (Python, Smalltalk, Scheme, etc.), functional languages (SML, Scheme, Haskell, etc.), and logic programming languages (Prolog, etc.). Java just doesn't have the type system or means of abstraction to make programmers highly productive.
Most of what I want to do with Java can already be done with it; the rest tends to be specific business logic that I'd have to write myself anyway. Since Java meets my needs, and supplies many advantages other than execution speed, it is the language of choice for the business I work for.
How nice for you that Java meets your needs, but that is completely and utterly besides the point. The news item is about claims that Java implementations have become really fast. I'm pointing out that the benchmarks test only a tiny fraction of what you would need to test for in a meaningful benchmark and that Java still isn't anywhere near competitive with C/C++ on the kind of code people write in C/C++.
Incidentally, it seems likely that you process a lot of strings in your work--guess what, Java is also very inefficient at that.
The news has been welcomed by gravitational wave hunters, since it boosts their hopes for detecting the gravitational waves [...] General relativity predicts that the two stars will slowly wobble like spinning tops allowing new tests of the theory.
Alternative theories to general relativity usually also predict such effects, including gravitational waves. So, these results, even if confirmed, don't actually tell you a whole lot. What they do tell you is that Newtonian mechanics isn't quite right, but, then, we already knew that.
If you want to use a different language that locks your binary into an specific compilation you are more than welcome to do that, but as for me, my application will experience no OS lockin.
I see: instead of being "locked into" an open, free standard with dozens of implementations, like POSIX and UNIX, you prefer to be locked into a single, proprietary language and library with only proprietary implementations, created by an ailing hardware vendor. Yeah, real smart.
Instead, write a complex *array* class, and an FFT using that complex array class.
Oh, and what do you propose array subscripting return from that complex array class? In fact, what do you propose "Complex.add(Complex)" return? Sorry, but you just haven't thought this through; there is no reasonable workaround for the lack of value classes in Java. If there were, language designers for languages like C++, Fortran, and C# wouldn't have gone through the considerable trouble of defining value classes.
I contest this point, simply because the same could be said of C++. If C++ were built on top of an interpretive VM, you'd argue that C++ was a poor language?
No. The problem isn't the implementation, it really is the language. C++ and C# both have been implemented on top of a VM (the CLR) and they don't share Java's limitations.
There exist Java -> native machine compilers today that generate binary code similar to what a C++ compiler would do. Java->native compilers are much less sophisticated than the optimizing compilers of Intel or Sun, but are getting better.
gcj is an excellent Java-to-native batch compiler. But it has the same limitations as JIT-based Java implementations. If anything, the JIT helps with avoiding some of the problems resulting from Java's language design.
Would you accuse a C++ garbage collecting memory manager of being "not good in high performance computing"?
Why would I? Garbage collection is generally more efficient than manual storage management. Java's problem isn't garbage collection, it's limtiations of the language itself. C# doesn't have those limitations, and neither does C++ (with or without garbage collection).
The C++ designers were very careful to ensure that you could write a "complex" number class, a "string" class, and an "array" class that were as efficient as anything built into the C++ compiler. Try implementing those classes efficiently in Java and you'll see that you can't.
If you stay away from trig functions, your code should be as fast as VisualC++, and faster then code compiled with GCC. How is that not "high performance"?
No, only code of the exact form used in that setl of carefully selected microbenchmarks is "as fast". Once you actually start using constructs like classes or multidimensional arrays, Java performance falls apart. Try implementing a complex number class in Java and compare its performance with a complex number class in C++.
Slashdot Mirror
The Intel High Definition Audio solution will have increased bandwidth that allows for 192 kHz, 32-bit, multi-channel audio
This is so that my eight-eared mutant pet bat from outer space can finally have a full high-fidelity experience.
For regular humans, of course, CD-quality audio is already overkill.
These are also the same people who enjoy particpating in system administration discussions when their system administration experience only stems from the 4 boxes they have at home.
Yes, and that's how SunOS itself originally became big, then Windows NT, then Linux. Face it, people start small and then both their projects and their needs grow.
If a company can't offer good value and good performance on the low end, they are doomed because only a few, select people who happen to be running big systems will buy them. That kind of arrogance killed most mainframe vendors and nearly did IBM in.
Of course, even as far as big systems and large networks are concerned, I think Sun has very little to offer that's competitive, but that's a separate discussion.
It's never a suprised that people on slashdot just don't get Sun equipment.
I used to buy Sun machines by the dozens--back when they gave me good bang-for-the-buck and when they were the best of the UNIX workstation bunch (of course, even back then, it was GNU software that made Solaris tolerable). Today, PCs give me more bang-for-the-buck and Linux and BSD have become far better operating systems, so there is no reason to like or advocate Sun workstations anymore.
Much like Apple, companies (I'd wager extremely few people buy Sun's for personal everyday use) that buy these boxes are buying them for the OS and rarely for the groundbreaking hardware.
Sure: they buy them for the OS. But that's not because there is anything particularly great about Solaris, it's because they want backwards compatibility.
Note that people used to use Sun workstations for personal everyday use, back when they were reasonably priced, performed well, and still represented the state-of-the-art in UNIX-like desktop systems. No more, however.
I seriously doubt that there are many businesses who switch to Solaris--Sun is living off repeat business now. Almost anybody who has a choice for a new project starts off with Linux, BSD, or NT these days.
Maybe on sheer performance it will be beaten by x86 however for crunching big data sets the UltraSparc is just more effecient.
Oh, and how exactly is it going to be "just more efficient"? Crunching big data sets may be limited by CPU or I/O. In terms of CPU, this thing is slow no matter which way you look at it. And in terms of I/O, it uses the same IDE drives as everybody else, and its Ethernet isn't going to be any faster either.
15 years ago, Sun workstations were a great deal: for under $2000, you could get something that no PC, no Mac, no other machine could compete with. Today, they are overpriced underperformers. Sun is milking their remaining customers who are unlucky enough to have committed to a Solaris environment.
Also some software only runs on Solaris so for that this box is good.
Yup, and that's why people buy these things. Of course, you can get equivalent or better software for other platforms, so anybody starting out now would be a fool to buy into Solaris, but those unlucky few who are stuck on Solaris software will just have to make do with this.
Or, alternatively, just learn lucid dreaming and become master of your dream world.
Ultimately, it comes down to the fact that when you attend a university, you have a contractual agreement with the university. And that contract probably permits the university to do this sort of thing. In fact, they probably also own any research results you obtain while at the university and any software you write. Public universities are a little more restricted in the kinds of conditions they can impose on students, but they can impose conditions as well.
I would ask two other questions about this, though. First, if the professor created new, interesting problem sets every year, then students couldn't plagiarize from the Internet, and plagiarism within a reasonable class size should be obvious.
Second, this kind of effort really doesn't need a private company; universities should and could do it cooperatively among themselves for less money. And the software for that isn't hard to write either.
I suspect the reason why it isn't happening is because good professors don't need this sort of thing and professors that do need this sort of thing probably are the same ones that aren't up to setting up large, cross-university software and data collection efforts.
As a student, you have a simple choice: just don't take classes from professors that engage in this sort of nonsense. And if it is part of mandatory courses, either grin and bear it or consider whether your choice of university was less-than-optimal after all.
Too bad they didn't come up with that list in 1958 when it mattered
A "kilo" stood for 1000 in 1958 and even in 1858.
and could have actually been successfully adpoted. It's simply too late to just arbitrarily start calling 1024 bytes a Kibibyte or whatever the fuck they decided to call it
This studpidity and problem is entirely due to computer engineers and they need to fix it. As computers are moving into other areas of daily life, the conflicting terminology will only get worse and the physical prefixes, with centuries of tradition behind them and wide use throughout even the computer industry, won't change.
Using color filters for high quality digital color photography is an old technology (and even older for analog). Its obvious limitation is that the subject has to be still.
Bigger pixels at lower resolution are not necessarily a good tradeoff: you can do almost as well in terms of noise and sensitivity by using more smaller sensors and performing the averaging in software.
Compositing lots of low resolution images into a single high resolution image is also completely standard: you can get both free and commercial software to do it.
Altogether, I suspect that if you take something like the new Sony 8Mpixel camera and take raw pictures with it, and reduce it to 1024x768 using good software, you are going to be pretty close to the measured quality and sensitivity of Spirit's sensor (in practice, you'll see little or no difference under normal circumstances, however). Then, you can use compositing software to composit multiple images for panoramas.
The Spirit tradeoffs make sense for a Mars rover, also taking into account power and weight requirements, but they do not result in a level of picture quality that you couldn't achieve with the digital cameras you can buy at the local store.
People in AI have worked on automated scientific discovery for decades, and some of their systems have also had robotic components. This seems like a tweak and a good sales job, not a breakthrough.
The family that frags together, stays together.
And remember: Uncle Sam is telling you that only a genuine Microsoft Xbox will do.
The lesson wasn't lost on the Republicans: buy votes with taxpayer money.
That argument is bogus because the Republicans are cutting like crazy in programs whose benefits voters actually experience.
What Republicans are buying isn't "votes", which would be the right thing to buy in a democracy, what they are buying is campaign contributions which they then use to influence the media. And that is not a good thing.
The traditional "neural feedback" for ADHD involves a whack with a ruler, administered by a Catholic nun. I suspect it is rather effective, although it may have unwanted (if temporary) side-effects.
This has to be 3-4 drives in a box without replication or redundancy (since you can't swap anything). That means you just greatly increased your risk of losing a whole lot of data at once because if any one drive goes, all your data is gone.
Get a real RAID drive or separate disks and you'll have more safety and more flexibility.
It's pretty ironic for republicans to portray democrats as fiscally irresponsible: republicans have been far, far worse over the last few decades, starting with Reagan. Republicans love towaste huge amounts of money unproductively, foremost on the military and propping up unproductive industries.
Maybe you have some trouble with the meaning of the word "can"? Look it up.
Large amounts of government spending can do wonders for the economy, if citizens are willing to make the sacrifices (i. e. pay the taxes). And manned space travel, useless as it is, is at least less destructive to foreign relations and industry than wars, Bush's other favorite economic activity.
However, tax cuts and massive spending don't work. And private industry is unlikely to go into space anytime soon--it's not profitable.
Well, he's right, and it's not like Apple's monopolistic impulses are new.
To me, neither Apple nor Sun seem like a good alternative to MS: they are trying to be like MS, they are just less good at it. And being smaller, they are less restrained and more vicious once they manage to sink their teeth into a market.
I guess it takes one to know one...
How can they "fall back" on OSX? OSX is Apple-proprietary. IBM can't sell it, they can' make their own machines for it, etc. And why would they want to anyway?
If you want to measure performance, let's consider time from "hey, we need some software" to "ok, data processing complete". That's the sort of speed that matters - maybe the benchmark should have checked for that..
You are so right that programmer productivity matters a great deal. Isn't it too bad that Java doesn't do that well either?
For performance critical and numerical code, programmer productivity in Java is horrendously bad because the language has so little support for it; C++ and C# beat Java hands down in those areas in terms of programmer productivity in my experience. Trying to write numerical or high performance code in Java is an exercise in frustration and takes many times as long as in languages like C++ or C#.
And for code where performance doesn't matter so much, Java seems barely better than C++ (it's slightly better mostly because it has garbage collection) and worse than C#. In my experience, highly productive languages are dynamically typed languages (Python, Smalltalk, Scheme, etc.), functional languages (SML, Scheme, Haskell, etc.), and logic programming languages (Prolog, etc.). Java just doesn't have the type system or means of abstraction to make programmers highly productive.
Most of what I want to do with Java can already be done with it; the rest tends to be specific business logic that I'd have to write myself anyway. Since Java meets my needs, and supplies many advantages other than execution speed, it is the language of choice for the business I work for.
How nice for you that Java meets your needs, but that is completely and utterly besides the point. The news item is about claims that Java implementations have become really fast. I'm pointing out that the benchmarks test only a tiny fraction of what you would need to test for in a meaningful benchmark and that Java still isn't anywhere near competitive with C/C++ on the kind of code people write in C/C++.
Incidentally, it seems likely that you process a lot of strings in your work--guess what, Java is also very inefficient at that.
The news has been welcomed by gravitational wave hunters, since it boosts their hopes for detecting the gravitational waves [...] General relativity predicts that the two stars will slowly wobble like spinning tops allowing new tests of the theory.
Alternative theories to general relativity usually also predict such effects, including gravitational waves. So, these results, even if confirmed, don't actually tell you a whole lot. What they do tell you is that Newtonian mechanics isn't quite right, but, then, we already knew that.
If you want to use a different language that locks your binary into an specific compilation you are more than welcome to do that, but as for me, my application will experience no OS lockin.
I see: instead of being "locked into" an open, free standard with dozens of implementations, like POSIX and UNIX, you prefer to be locked into a single, proprietary language and library with only proprietary implementations, created by an ailing hardware vendor. Yeah, real smart.
Instead, write a complex *array* class, and an FFT using that complex array class.
Oh, and what do you propose array subscripting return from that complex array class? In fact, what do you propose "Complex.add(Complex)" return? Sorry, but you just haven't thought this through; there is no reasonable workaround for the lack of value classes in Java. If there were, language designers for languages like C++, Fortran, and C# wouldn't have gone through the considerable trouble of defining value classes.
I contest this point, simply because the same could be said of C++. If C++ were built on top of an interpretive VM, you'd argue that C++ was a poor language?
No. The problem isn't the implementation, it really is the language. C++ and C# both have been implemented on top of a VM (the CLR) and they don't share Java's limitations.
There exist Java -> native machine compilers today that generate binary code similar to what a C++ compiler would do. Java->native compilers are much less sophisticated than the optimizing compilers of Intel or Sun, but are getting better.
gcj is an excellent Java-to-native batch compiler. But it has the same limitations as JIT-based Java implementations. If anything, the JIT helps with avoiding some of the problems resulting from Java's language design.
Would you accuse a C++ garbage collecting memory manager of being "not good in high performance computing"?
Why would I? Garbage collection is generally more efficient than manual storage management. Java's problem isn't garbage collection, it's limtiations of the language itself. C# doesn't have those limitations, and neither does C++ (with or without garbage collection).
The C++ designers were very careful to ensure that you could write a "complex" number class, a "string" class, and an "array" class that were as efficient as anything built into the C++ compiler. Try implementing those classes efficiently in Java and you'll see that you can't.
Erm, did you read the article?
Erm, did you read what I wrote?
If you stay away from trig functions, your code should be as fast as VisualC++, and faster then code compiled with GCC. How is that not "high performance"?
No, only code of the exact form used in that setl of carefully selected microbenchmarks is "as fast". Once you actually start using constructs like classes or multidimensional arrays, Java performance falls apart. Try implementing a complex number class in Java and compare its performance with a complex number class in C++.