It's inspiring that the GPL is becoming the standard license for companies who release free software. This is great news because it removes a lot of the complications for users and developers.
Thanks Troll Tech! I look forward to finally using KDE!
Maybe the G4 can theoretically achieve 2 gigaflops, but I challenge you to prove it to me on a useful application. I'd be proud of you if you could get 25% of that number.
The main reason you would want this instead of a G4 is that it's memory is much faster than a G4's. The C90's main memory (4GB) is like your G4's cache (not 4GB) in terms of speed. Even when your faithful CPU meter reads "100%", your CPU is often idle. It has to sit and wait for memory requests which take quite a while. The cache can only do so much.
It is true that this C90 is getting old and most of them have probably been decommissioned. However, I am sure that there are still C90's out there in use and they can't be replaced by Macs.
We could have the first community-owned supercomputer. Imagine the
possibilities...
Well this is kind of a good idea, but you will be happy to know that it has already been done.
The US government spends millions of dollars on supercomputers every year. Some of the computers are for Classified projects, but many of them are for research purposes. These research computers are like a national computing resource. You paid the tax money for them and, if you're so inclined, you can probably use them. If you think you have a project that would benefit from a supercomputer, you can apply for time on one.
If you buy your own Cray you'll be guaranteed time on it, but you'll also be burdened with maintenance and upgrades.
The C90 uses a liquid refridgerant. No fans here. 35k seems like a good price, but you have to remember that the power requirements of this thing will make your electric meter spin. Plus, if it broke you would be screwed. (Cray service isn't cheap.)
Why do you think athletes still make pitful salaries?
Maybe they arn't unionized. They should join the screen actors guild. Actors and atheletes arn't much different from each other these days. It would be pretty funny to see the atheletes striking too.
Maybe most civilizations don't last long after discovering radio communication. It seems like our candle has almost been snuffed several times in the last century.
The whole problem with this system is that it assumes that the only valid purpose of publishing is to make money by delivering information from the producer to the consumer. But publishing does more than that. It adds to the collective body of knowledge owned by humanity.
I have dozens of textbooks on the shelf behind me, and I don't intend to buy another circuits book because Addison Wesley thinks it's out of date.
The problem with the world wide web, and the reason publishers like it, is that information can be revoked at whim. Sure, CNN publishes hundreds of stories every week on the web, and you can access them quite a ways back. But what if CNN realizes that one of the stories it wrote last year was very embarrasing, but only because of some new information that has come to light. There is a strong incentive for them to remove the story from their website. (Remember they did this with a DeCSS story a while back, removing a very embarrassing link.) When a story is published on paper, the publisher can't recall the paper for a bit of editing. They have to live with their mistakes.
If all information is published "WWW" style, this starts to look like 1984. Now books are moving in to this territory. Today they're on CDROM, tomorrow they'll just be downloaded by a proprietary browser. Imagine a world where the page in the history book you're reading today is different than it was yesterday. Maybe the publisher updated it with "value added" content, but maybe they just crossed out a paragraph.
I recommend that people reject this kind of digital publishing. If publishers really want to publish a book electronically it should never be licensed in a way that limits the time that information can be used. If they want to put their information on the internet, they should use the Freenet, or some similar means to ensure that the information is not controlled by anyone once it has been released.
We have trouble defining an OS now, so it's hard to say what the first one was. Early system software was just device drivers that hung around in memory. Timesharing and other cool stuff didn't happen for a while. Looking at the timeline in "Operating System Concepts" by Silberschatz and Galvin we see:
1958 Atlas, virtual memory, University of Manchester, England
1962 CTSS, Compatible Time-Sharing System, by Corbato at MIT
These seem to be early milestones in the evolution of Operating Systems.
I would moderate this article -1, flamebait. I don't see what is wrong with a compnay that is trying to make money off of free software. It's better than a company making money off of commercial software. The article implies that Helix Code will twist Gnome to it's own evil purposes. How can it when Gnome is GPLed?
"Gnome's stated purpose, its whole reason for existence, is to kill KDE."
This implies that all the Gnome developers care about is bashing KDE. Once KDE is dead, they'll all just stop working on Gnome and go to do something else. Gnome exists to provide a free alternative to KDE. If KDE dies because people prefer freedom, well that's too bad for KDE.
Licenses do matter, and they're worth nitpicking over. It's the licensing of Gnome that causes companies to prefer it over KDE. It's not any technical reason. If technical issues were all that mattered for software, Bill Gates would not be a rich man.
This is absolutely true. As sharing becomes easier people will be more likely to do so. When unmetered broadband and stable OSs become widespread, there won't be any reason why people won't share lots of files.
Look at the free software movement which is considered relatively successful. How many users are also authors? A small percentage even though the source is availible and anyone can make modifications. It's hard to write software and yet it happens often enough that many of us never have to use proprietary programs.
By comparison, file sharing is much easier and will become even more so soon. I don't think you can legitimately criticize a dial-up user for being reluctant to give up bandwidth to sharefiles.
An interesting bit was the cc:NUMA architecture for high end clustering. I wonder what will become of it?
Maybe you've never heard of a company called SGI. They're down right now, but not necessarily out. Their new Orign 3000 server line uses a cc:NUMA architecure. They hope to run Linux on these someday, but Intel has been jerking them around with Itanium delays.
I suspect that Cray is also using cc:NUMA technology (which they got from SGI) in their next computer, the SV2.
I do find it ironic, though, that it is *Gnome* taking this step. Could anybody have possibly imagined this
when Gnome started? Weren't they the "hacker desktop"? Didn't they have all the "desktop for the
people" principles? Hmm... times change, I guess.
Yeah, I get a special feeling using code that's 100% pure hacker encoded. It has that pleasant "caffine" smell to it. Using code that came from paid programmers just isn't the same. I guess I'll have to ditch using Gnome now that it has "sold out" to "the man".
Seriously, Gnome is GPL'd! It doesn't matter who codes it. This has been a good month for free software. Even if Sun and HP did nothing but leech the Gnome source without contributing any new features, the publicity alone would be very helpful. I hope this trend continues.
I'm not surprised that Microsoft is doing this, but I will be surprised if they actually decide to sell it.
With a company the size of Microsoft, developing a piece of software is not the same as making a product. My guess is that this is a bit of an experiment. If it works out well and Linux keeps growing on the desktop, they might market it, but if they had MS Office for Linux ready to go right now, I doubt they would put in on the shelves given the current market. There is a very small market for Linux desktop applications and with the anti-microsoft sentiment around here I don't think it would be warmly recieved by most Linux users.
One thing about the software industry is there is a lot of intellectual property compared to other industries.
If I take the source code from a software company, I've basically got the entire product and the means to produce it.
If I take one of GM's cars, I might get some good ideas about how to make a car, but I need dies, tools, and a whole assembly line if I want to duplicate it.
>I'd rather pay an extra $200 and get the OTHER really small computer.
Or maybe a Sony Vaio!
The tiquet is pretty impressive, but when you consider all of the things it needs to get running like a keyboard, monitor, battery, etc. you might as well just get a laptop which will at least be fast and have a decent sized disk.
Come up with, by consensus or decree, a project which the class will work on. Something that's too big for most of the students to tackle on their own, but doable by a group in a few weeks. A good example would be a game like tetris.
Then step them through the software design process.
1. State the goal 2. Decide what components are necessary for the program 3. Define specifications for each component 4. Write code 5. Test 6. Integrate 7. Test 8. Enjoy
Steps 3-5 could be split into different groups and done in parallel. (ie. you do input, you do output, you make a title screen, etc.)
This might be getting a little complicated for HS, and you'd need some functional programming skills as well as a decent language, but it would be a good team project and you'd have something interesting at the end.
If it's a big or competant class, break it up into groups and make several programs. (Competition!) The key is to keep it fairly simple so that you get a finished product.
It sounds to me like this is a tool for people who have lots of money. So I guess this means that an independant artist couldn't just whip up a version of their album encoded with "watermarks" for distribution.
If music players all force you to only play music that is "signed" by this technology, guess who gets to decide what kind of music people listen to? Only music labels.
The promise of digital technology was that it could make things like producing and distributing things like books and music accessible to anyone with a computer and microphone. But I guess we can't have that if it threatens the music industry. If this kind of technology spreads from the music industry to other places like software and publishing, we're really going to be in trouble.
Music publishers want to have it both ways. The freedom of the internet, and the control of copyright enforcement. These two forces are opposed to each other. Something's gotta give, and I hope it isn't freedom.
Artists will make great music even if they aren't paid. I have the recordings to prove it.
120 or so 400mhz PII systems and thinking just how much CPU power was wasted (probably close to as much as the CrayT3E in the next building has).
You must have a really cheap Cray.
The problem you're seeing here is not how many FLOPS you have. It's how many you can use. If one man can dig a post hole in 10 minutes does that mean that 10 men can dig it in one?
The postal workers are getting a little out of control. With the competition from overnight services and email, the USPS's role in binding the nation togeather is becoming less and less important.
So do they accept this? No. Like most beauracracies they have too much momentum to stop when they arn't needed. USPS online bill paying service is first. Now USPS email. Maybe next they'll compete with PayPal. Hell, they might as well just get into the online commerce buisness and call themselves "eUSPS.com". But they'll never IPO. They'll just keep raking in taxpayer money and raise taxes whenever they need more.
The provision for the Postal Service in the US Constituion was a mistake. It made sense that a postal service was needed at the time, but the writers of the constitution didn't forsee that there might be a time when it wasn't needed. (Or at least a time when it wasn't needed and it wouldn't be politically feasible to eliminate it.) Now we're saddled with this beast which has nothing better to do but compete with the private sector. If there actions with snail mail are any indication, they will probably setup barriers against other companies, like ISPs, who compete with their services.
Their claim of privacy is a crock too. Do you think the USPS would refuse an FBI request to install Carnivore? I highly doubt it.
I'm not saying that we're ready to get rid of the USPS yet, but that day could be coming soon. The last thing we need is for it to be expanding it into areas where it competes with legitimate buisnesses which are doing a perfectly good job already. When the government is given something (or takes it) it is very reluctant to give it back. Let's keep email with the people.
NUMA (non-uniform memory access) is very tricky to program. The attraction of a distributed OS is the idea that the operating system will somehow magically take care of all of the network implementation details. The application programmer would then just write a multi-threaded application and watch it run efficiently across many computers. Most of the time this doesn't work.
If the OS presents a uniform view of memory, then the system is easy to program, but won't be efficient. When you read some memory it might be on your computer or it might be on another one. If its on another one and you are accessing it from an inner loop, you application will be very slow.
If the OS presents a non-uniform view of memory, then the programmer is burdened with the the decisions about what should go where. In many applications it is difficult or impossible to pick a "best" location for your data. So you're stuck doing a lot of work, but maybe not getting much benefit from it.
Consider that MPP machines like the Cray T3E are essentially distributed OS machines. They have a very high bandwidth interconnect network, and they still suffer severe performance penalties from the problems described above.
Some applications can benefit from a distributed OS, but for many others the communication overhead seriously impacts performance. Don't expect to hook up a pile of PCs to a 10-baseT network and start solving anything less parallel than SETI@Home.
One way to solve the latency problem is to create enough threads per node to hide the communication delay. Unfortunatly most applications that normal users are interested in arn't big enough to use very many threads.
Seriously, alternatives are under development (freenet.sourceforge.net) but they won't be generally availible for a while.
I think the Napster protocol is it for a now, but we don't have to use Napster, Inc. for that. Perhaps this weeks scare has prompted more people to setup and use OpenNap servers.
Slashdot Mirror
It's inspiring that the GPL is becoming the standard license for companies who release free software. This is great news because it removes a lot of the complications for users and developers.
Thanks Troll Tech! I look forward to finally using KDE!
Maybe the G4 can theoretically achieve 2 gigaflops, but I challenge you to prove it to me on a useful application. I'd be proud of you if you could get 25% of that number.
The main reason you would want this instead of a G4 is that it's memory is much faster than a G4's. The C90's main memory (4GB) is like your G4's cache (not 4GB) in terms of speed. Even when your faithful CPU meter reads "100%", your CPU is often idle. It has to sit and wait for memory requests which take quite a while. The cache can only do so much.
It is true that this C90 is getting old and most of them have probably been decommissioned. However, I am sure that there are still C90's out there in use and they can't be replaced by Macs.
Well this is kind of a good idea, but you will be happy to know that it has already been done.
The US government spends millions of dollars on supercomputers every year. Some of the computers are for Classified projects, but many of them are for research purposes. These research computers are like a national computing resource. You paid the tax money for them and, if you're so inclined, you can probably use them. If you think you have a project that would benefit from a supercomputer, you can apply for time on one.
If you buy your own Cray you'll be guaranteed time on it, but you'll also be burdened with maintenance and upgrades.
Try these links:
NPACI Allocations
www.sdsc.edu
www.ncsa.edu
The C90 uses a liquid refridgerant. No fans here. 35k seems like a good price, but you have to remember that the power requirements of this thing will make your electric meter spin. Plus, if it broke you would be screwed. (Cray service isn't cheap.)
Maybe they arn't unionized. They should join the screen actors guild. Actors and atheletes arn't much different from each other these days. It would be pretty funny to see the atheletes striking too.
It wouldn't stop bad patents, but it would reduce the damage.
Maybe most civilizations don't last long after discovering radio communication. It seems like our candle has almost been snuffed several times in the last century.
I have dozens of textbooks on the shelf behind me, and I don't intend to buy another circuits book because Addison Wesley thinks it's out of date.
The problem with the world wide web, and the reason publishers like it, is that information can be revoked at whim. Sure, CNN publishes hundreds of stories every week on the web, and you can access them quite a ways back. But what if CNN realizes that one of the stories it wrote last year was very embarrasing, but only because of some new information that has come to light. There is a strong incentive for them to remove the story from their website. (Remember they did this with a DeCSS story a while back, removing a very embarrassing link.) When a story is published on paper, the publisher can't recall the paper for a bit of editing. They have to live with their mistakes.
If all information is published "WWW" style, this starts to look like 1984. Now books are moving in to this territory. Today they're on CDROM, tomorrow they'll just be downloaded by a proprietary browser. Imagine a world where the page in the history book you're reading today is different than it was yesterday. Maybe the publisher updated it with "value added" content, but maybe they just crossed out a paragraph.
I recommend that people reject this kind of digital publishing. If publishers really want to publish a book electronically it should never be licensed in a way that limits the time that information can be used. If they want to put their information on the internet, they should use the Freenet, or some similar means to ensure that the information is not controlled by anyone once it has been released.
1958 Atlas, virtual memory, University of Manchester, England
1962 CTSS, Compatible Time-Sharing System, by Corbato at MIT
These seem to be early milestones in the evolution of Operating Systems.
"Gnome's stated purpose, its whole reason for existence, is to kill KDE."
This implies that all the Gnome developers care about is bashing KDE. Once KDE is dead, they'll all just stop working on Gnome and go to do something else. Gnome exists to provide a free alternative to KDE. If KDE dies because people prefer freedom, well that's too bad for KDE.
Licenses do matter, and they're worth nitpicking over. It's the licensing of Gnome that causes companies to prefer it over KDE. It's not any technical reason. If technical issues were all that mattered for software, Bill Gates would not be a rich man.
Look at the free software movement which is considered relatively successful. How many users are also authors? A small percentage even though the source is availible and anyone can make modifications. It's hard to write software and yet it happens often enough that many of us never have to use proprietary programs.
By comparison, file sharing is much easier and will become even more so soon. I don't think you can legitimately criticize a dial-up user for being reluctant to give up bandwidth to sharefiles.
Maybe you've never heard of a company called SGI. They're down right now, but not necessarily out. Their new Orign 3000 server line uses a cc:NUMA architecure. They hope to run Linux on these someday, but Intel has been jerking them around with Itanium delays.
I suspect that Cray is also using cc:NUMA technology (which they got from SGI) in their next computer, the SV2.
Yeah, I get a special feeling using code that's 100% pure hacker encoded. It has that pleasant "caffine" smell to it. Using code that came from paid programmers just isn't the same. I guess I'll have to ditch using Gnome now that it has "sold out" to "the man".
Seriously, Gnome is GPL'd! It doesn't matter who codes it. This has been a good month for free software. Even if Sun and HP did nothing but leech the Gnome source without contributing any new features, the publicity alone would be very helpful. I hope this trend continues.
With a company the size of Microsoft, developing a piece of software is not the same as making a product. My guess is that this is a bit of an experiment. If it works out well and Linux keeps growing on the desktop, they might market it, but if they had MS Office for Linux ready to go right now, I doubt they would put in on the shelves given the current market. There is a very small market for Linux desktop applications and with the anti-microsoft sentiment around here I don't think it would be warmly recieved by most Linux users.
If I take the source code from a software company, I've basically got the entire product and the means to produce it.
If I take one of GM's cars, I might get some good ideas about how to make a car, but I need dies, tools, and a whole assembly line if I want to duplicate it.
-jlg
ps. use Debian! www.debian.org
But I bet it has a great linpack score. Benchmarks don't lie.
-jlg
ps. use Debian! www.debian.org
Or maybe a Sony Vaio!
The tiquet is pretty impressive, but when you consider all of the things it needs to get running like a keyboard, monitor, battery, etc. you might as well just get a laptop which will at least be fast and have a decent sized disk.
-jlg
ps. use Debian! www.debian.org
Then step them through the software design process.
1. State the goal
2. Decide what components are necessary for the program
3. Define specifications for each component
4. Write code
5. Test
6. Integrate
7. Test
8. Enjoy
Steps 3-5 could be split into different groups and done in parallel. (ie. you do input, you do output, you make a title screen, etc.)
This might be getting a little complicated for HS, and you'd need some functional programming skills as well as a decent language, but it would be a good team project and you'd have something interesting at the end.
If it's a big or competant class, break it up into groups and make several programs. (Competition!) The key is to keep it fairly simple so that you get a finished product.
-jlg
ps. use Debian! www.debian.org
If music players all force you to only play music that is "signed" by this technology, guess who gets to decide what kind of music people listen to? Only music labels.
The promise of digital technology was that it could make things like producing and distributing things like books and music accessible to anyone with a computer and microphone. But I guess we can't have that if it threatens the music industry. If this kind of technology spreads from the music industry to other places like software and publishing, we're really going to be in trouble.
Music publishers want to have it both ways. The freedom of the internet, and the control of copyright enforcement. These two forces are opposed to each other. Something's gotta give, and I hope it isn't freedom.
Artists will make great music even if they aren't paid. I have the recordings to prove it.
-jlg
ps. use Debian! www.debian.org
You must have a really cheap Cray.
The problem you're seeing here is not how many FLOPS you have. It's how many you can use. If one man can dig a post hole in 10 minutes does that mean that 10 men can dig it in one?
-jlg
ps. use Debian! www.debian.org
So do they accept this? No. Like most beauracracies they have too much momentum to stop when they arn't needed. USPS online bill paying service is first. Now USPS email. Maybe next they'll compete with PayPal. Hell, they might as well just get into the online commerce buisness and call themselves "eUSPS.com". But they'll never IPO. They'll just keep raking in taxpayer money and raise taxes whenever they need more.
The provision for the Postal Service in the US Constituion was a mistake. It made sense that a postal service was needed at the time, but the writers of the constitution didn't forsee that there might be a time when it wasn't needed. (Or at least a time when it wasn't needed and it wouldn't be politically feasible to eliminate it.) Now we're saddled with this beast which has nothing better to do but compete with the private sector. If there actions with snail mail are any indication, they will probably setup barriers against other companies, like ISPs, who compete with their services.
Their claim of privacy is a crock too. Do you think the USPS would refuse an FBI request to install Carnivore? I highly doubt it.
I'm not saying that we're ready to get rid of the USPS yet, but that day could be coming soon. The last thing we need is for it to be expanding it into areas where it competes with legitimate buisnesses which are doing a perfectly good job already. When the government is given something (or takes it) it is very reluctant to give it back. Let's keep email with the people.
-jlg
ps. use Debian! www.debian.org
NUMA (non-uniform memory access) is very tricky to program. The attraction of a distributed OS is the idea that the operating system will somehow magically take care of all of the network implementation details. The application programmer would then just write a multi-threaded application and watch it run efficiently across many computers. Most of the time this doesn't work.
If the OS presents a uniform view of memory, then the system is easy to program, but won't be efficient. When you read some memory it might be on your computer or it might be on another one. If its on another one and you are accessing it from an inner loop, you application will be very slow.
If the OS presents a non-uniform view of memory, then the programmer is burdened with the the decisions about what should go where. In many applications it is difficult or impossible to pick a "best" location for your data. So you're stuck doing a lot of work, but maybe not getting much benefit from it.
Consider that MPP machines like the Cray T3E are essentially distributed OS machines. They have a very high bandwidth interconnect network, and they still suffer severe performance penalties from the problems described above.
Some applications can benefit from a distributed OS, but for many others the communication overhead seriously impacts performance. Don't expect to hook up a pile of PCs to a 10-baseT network and start solving anything less parallel than SETI@Home.
One way to solve the latency problem is to create enough threads per node to hide the communication delay. Unfortunatly most applications that normal users are interested in arn't big enough to use very many threads.
-jlg
ps. use Debian! www.debian.org
-jlg
ps. use Debian! www.debian.org
I'm putting my blues collection back up!
Seriously, alternatives are under development (freenet.sourceforge.net) but they won't be generally availible for a while.
I think the Napster protocol is it for a now, but we don't have to use Napster, Inc. for that. Perhaps this weeks scare has prompted more people to setup and use OpenNap servers.
-jlg
ps. use Debian! www.debian.org