Exactly my point... but so can closed source application save/edit documents in XML format. The fact that a particular application is closed source vs. open source has nothing to do with the app's support of open standards for file formats and such. Therefore, you cannot say that in order to support open standards you must use open source and/or Linux.
So... if a closed source application reads and writes file formats according to open standards, wouldn't that be OK by that wording? You don't need to see the application code as long as the output(s) are compliant to an open standard. Also, I didn't think Linux was an "open standard", at least no one has ever given me the standards document nor have I ever seen the ISO, ANSI, IEEE, etc. approval of it.
Re:are all the reviews by idiots?
on
Athlon 64 Debuts
·
· Score: 1
Well... you are simply (and I'm not oversimplifying) wrong. If you care to continue discussing your point, perhaps you would attempt to prove yourself by giving examples of these amazing speedups of which you speak.
An Athlon64 (or FX-51 or Opteron) running a 64-bit version of Linux and running apps compiled as native 64-bit x86-64 is truly a 64-bit system... all the "pieces" are 64-bit. You can test your claims today and report back. The UT2003 tests show ~20% speedup. That is one point of data for your research already.
Well.... it does say "*A* new technology" and not "*This* new technology". I had to look over that a couple times, for the same reasons. I am only worried if they really meant "*This* new technology".
In my experience as well, you write it on one platform and have to do a bit of debugging to get it to run on another, more debugging for a 3rd, etc. This generally isn't so bad if you use the same JVM on the different hardware/OS but switch JVM and you are out in left field again.
What is the theoretical size limit of a beowulf cluster? 255? 35xxx? 10^20 ?
This is usually limited by the amount of resources on the platforms. At times, this has been governed by such things as the number of open sockets the OS supported and/or how long it took to open all the connections to all the machines before RSH or the like started timing out and closing the connections.
Has anyone thought of making a beowulf cluster of beowulf clusters yet?
Yes. Grid computing encompasses this idea (things like Globus, etc.)
Not really. As a developer, we downloaded the.NET Runtime, include it on the CD, and do the install if a.NET CLR isn't detected as already present. It is very easy to do using an installer program (like 15 minutes worth of work, including the download of the CLR from Microsoft and you only have to do it once).
Well... it draws attention to it. Let me give an example. Sure, it isn't that big of a deal, but here it is.
If you don't know what the offending code is, and it is removed in a subsequent kernel release, then you might not ever even care to try to find out what it was. Sure, some folks with nothing better to do will do diffs and post it all over, but the majority of folks would just patch and move on.
Now, enumerate the lines that are offending and folks will all know.
Sort of like that rumor about Tom Clancy and his books. Supposedly Clancy found out about stuff that was "classified" and put it in his books. He maintained that all his information was public information, you might just have to dig for it. Supposedly the CIA told Clancy that he has classified information in his books and would have to remove it. Clancy said that he would remove it no problem, just point out which information was classified and he'd remove it. The CIA figured it was best to just leave it alone since removing the information would reveal that it was potentially true whereas leaving the information in could hide it among the other data that was simply fiction.
Granted, if there IS code that has to be removed, there are plenty of folks with nothing better to do than to do diffs on every release to see what the code was.
Since the code is now visible, the cat's out of the bag. However, if this SCO issue IS true and no real penalties can be assessed (actually, it doesn't matter that the person who put the code in is penalized since this type of thing can, in some folks' opinions, potentially ruin companies and cause lots of folks to lose their jobs), it is still a bad thing in that it rewards behavior that is found unacceptable (in the legal world at least). One fear is that it simply encourages others to "steal" code from their employer and release it in an OSS product and letting that cat out of the bag as well - either through vindictive behavior or for some other reason like it would make the person doing this have some "heroic" reputation among the community.
If there is no penalty for doing so, then this could be a big issue to commercial software and I'm still not convinced that any of the OSS folks have thought about what that could mean in the long term.
- It was never said that the drivers wouldn't work for everything else. - If you ever paid for a driver update from either nVidia or ATI, you got ripped.
What I think we'll find is that only the games that have benchmark capability and are, or are going to be, popular enough for large market appeal - the HalfLife2s, the Quake/Dooms, etc. - are the ones that will get optimizations by the card vendors. This is because it will help them sell cards.
Nope... just saying that the fighter analogy is a poor one. That and I don't want a USA-OS and a USA-CPU, not that we can control what other countries do. Also that it is wrong for the USA Government to compete with the private/commercial sector according to our (USA) laws.
Actually, it is illegal in the USA for the government to compete with commercial industry. That's why fighter planes are defined by a big list of specifications (has to fly mach 1+ at 90% power, carry a payload of X, etc.) which are given to a number of airplane designers/manufactuerers and those companies bid on the cost of making the plane. Some review board then looks at the prototypes and such and decides which meet the specifications and which do not, and decides which company gets the contract. The difference in the previous scenario is that the new fighter plane has requirements that are not met by an existing plane in inventory. That's why they need a new one.
If the requirements of a government software project are met by an existing set of software from a commercial business, they aren't supposed to give money or commission something else. If their requirements are met by some software on MSWindows or Linux, then they are supposed to use what is there. If the software doesn't exist, then they provide specs and get bids for the work. etc.
Having a USA OS (doesn't matter what the thing is) is a VERY, VERY bad thing. Same as having the USA-2680 CPU (some made up processor). Other countries can do what they want, but in the USA there are laws (and good reasons for them) why this should not be.
Re:Slow cumbersome process
on
Java vs .NET
·
· Score: 1
Care to put your name behind your words? Didn't think so...
My experience has been that Java was slower than C#.NET apps on the same hardware and the same level of optimizations.... sometimes not by much, sometimes by a bit.
But what happens in two years when MS changes to something completely different and your hojillion lines of.NET code are worthless?
The same thing that happens now? You keep your hardware/software as it is and keep using the thing. I've seen companies that have controller software written for Apple ][e machine that use those things *to this day* because it works and does the job. They have to buy all the Apple ][e machines/parts they can from eBay and such to make sure they can keep their hardware working. How long ago was the Apple ][e put on the unsupported list?
You'd be surprised just how many companies still use MSDOS programs as their main software platform, I guess. Just because it is there is not sufficient reason for you to have to upgrade to it. If it becomes unsupported from the company that makes it, you can still live on.
Heh... we had a Java project and we decided to use JBuilder for it. I had a PentiumIII @ 850MHz and 768M of memory and the thing took eons to do anything. I found it somewhat inconsistent. While it basically worked, it was extremely slow (I don't know if this is because it was all written in Java or what) to the point of breaking my flow at times. I'd not want to use it again.
Not to mention that the checksum calculations will destroy the cache and your performance will go into the crapper. I can't imagine this as being a viable option.
The only real option is to have a voter system where multiple nodes (or subclusters) all calculate the same thing at the same time. You compare all the answers and take the answer set that has the most nodes (or subclusters) with that result.
what did people do before there were web browsers at work? Oh, yeah, they did THEIR JOB.
Not necessarily more or less than after web browsers came along. I usually think that people who have to go outside for 10 minutes every hour to smoke a cigarette spend more time away from their desk than the time I spend browsing the web or whatever. And those people tend to browse the web as much as I do on top of that. Before browsers, people would hang around the water cooler, in the break room, or in the copy room and talk. There might have been some increase in "non-productive" time since web browsers but probably 1/2 or more of the stuff I browse can potentially apply to work (I visit a number of tech sites that frequently have information that is useful to me as a part of my daily ritual).
Heh... so he likes the fact that they can't download stuff to use on their machines. Sounds kind of draconian in the end. You will use your machine for nothing but work! I have had bosses that tried to do similar things in the past and all they really did is cause folks to leave the company because things were so strict and you basically couldn't take breaks anymore other than pushing back from your desk and staring at the wall for 10 minutes. Of course, the people who are most able to get other jobs move on first and the people who can't, stay. People will find a way, and even pride themselves in their creativity, of jacking the system using breaks and finding ways to goof off. Sure. The amount of time these folks may work may go up but the quality/productivity may go down so the net result will be a loss.
not necessarily of the Linux kernel, but of the apps on the platform.
1. I want to download an app and it puts me in dependency hell. I have to download 10 other packages in order to actually use that app, and 8 of those packages break other apps I have installed.
2. Inconsistencies in GUI among apps. In some apps, right click pulls up menus, in others it doesn't - or it is some other random sequence of mouse buttons. In some apps, the "normal" menu items are scattered all over hell's half acre compared to where the "normal" menu items are for another application. ("Normal" being file open/save, edit, copy/paste, etc.) In some apps, ones with multiple windows, which window actually contains the menu you want may have no correlation with what you are actually doing in that window.
3. Usability of GUI among apps. In many applications, the GUI appears to have been put together either simply by order of implementation of the features or by some obscure/strange workflow that makes no sense to anyone other than the person who developed the GUI. Many times the GUI seems very haphazardly assembled with no regard for workflow patterns which are normally used in the type of processing performed by the application.
I understand that for you, no flame intended, the O/S is more important than the data you generate with your "own" computer.
Actually... interoperability with data I produce is a big concern. It is silly of me to generate data that someone else cannot use because I'm using some non-standard and/or off-the-wall app. Knowing that a certain platform has built-in support for a certain type of data eliminates concerns that I have of interoperability.
Knowing that a platform has certain features built-in means that I can make use of that in software that I write. For example, for help files, I *know* that a particular platform has an HTML browser (built-in or bundled) so I can write all my help files in HTML without worry of some user not being able to access the help files I produce. I do not have to generate a big long list of dependencies that users must satisfy before they can use the software I built. (Remember my previous statements of how users like to find out after-the-fact that they have to add more things to their system in order to use something they just got. This is one of the top three most infuriating things to me about Linux... I find some piece of software that I want to use only to find out that I have to spend half a day downloading and installing the dependencies for it before I can use it. The other most infuriating things being inconsistencies and usability of UIs.)
(offtopic)
Man, I miss my Atari1040ST. Unfortunately, it gave up the ghost a few years back and I didn't need it enough to warrant trying to fix it.
I liked the Amigas a lot even though I was never an owner.
I used to sometimes wonder what computing would be like today had IBM gone with the 68k instead of the 8088 for the IBM PC... we can all dream I guess.
Slashdot Mirror
Exactly my point... but so can closed source application save/edit documents in XML format. The fact that a particular application is closed source vs. open source has nothing to do with the app's support of open standards for file formats and such. Therefore, you cannot say that in order to support open standards you must use open source and/or Linux.
So... if a closed source application reads and writes file formats according to open standards, wouldn't that be OK by that wording? You don't need to see the application code as long as the output(s) are compliant to an open standard. Also, I didn't think Linux was an "open standard", at least no one has ever given me the standards document nor have I ever seen the ISO, ANSI, IEEE, etc. approval of it.
Well... you are simply (and I'm not oversimplifying) wrong. If you care to continue discussing your point, perhaps you would attempt to prove yourself by giving examples of these amazing speedups of which you speak.
An Athlon64 (or FX-51 or Opteron) running a 64-bit version of Linux and running apps compiled as native 64-bit x86-64 is truly a 64-bit system... all the "pieces" are 64-bit. You can test your claims today and report back. The UT2003 tests show ~20% speedup. That is one point of data for your research already.
So you think DRAM access time is 40ns (or so) because of the system bus?
L1 cache typically found on today's processors and DRAM are two different things with different design targets. Pick up a VLSI book.
Says a lot about folks who mod, that's for sure...
Well.... it does say "*A* new technology" and not "*This* new technology". I had to look over that a couple times, for the same reasons. I am only worried if they really meant "*This* new technology".
In my experience as well, you write it on one platform and have to do a bit of debugging to get it to run on another, more debugging for a 3rd, etc. This generally isn't so bad if you use the same JVM on the different hardware/OS but switch JVM and you are out in left field again.
What is the theoretical size limit of a beowulf cluster? 255? 35xxx? 10^20 ?
This is usually limited by the amount of resources on the platforms. At times, this has been governed by such things as the number of open sockets the OS supported and/or how long it took to open all the connections to all the machines before RSH or the like started timing out and closing the connections.
Has anyone thought of making a beowulf cluster of beowulf clusters yet?
Yes. Grid computing encompasses this idea (things like Globus, etc.)
"allows you to scale your processing pretty much linearly."
Interesting promise. I guess it depends on what you mean by "pretty much"...
Not really. As a developer, we downloaded the .NET Runtime, include it on the CD, and do the install if a .NET CLR isn't detected as already present. It is very easy to do using an installer program (like 15 minutes worth of work, including the download of the CLR from Microsoft and you only have to do it once).
Well... it draws attention to it. Let me give an example. Sure, it isn't that big of a deal, but here it is.
If you don't know what the offending code is, and it is removed in a subsequent kernel release, then you might not ever even care to try to find out what it was. Sure, some folks with nothing better to do will do diffs and post it all over, but the majority of folks would just patch and move on.
Now, enumerate the lines that are offending and folks will all know.
Sort of like that rumor about Tom Clancy and his books. Supposedly Clancy found out about stuff that was "classified" and put it in his books. He maintained that all his information was public information, you might just have to dig for it. Supposedly the CIA told Clancy that he has classified information in his books and would have to remove it. Clancy said that he would remove it no problem, just point out which information was classified and he'd remove it. The CIA figured it was best to just leave it alone since removing the information would reveal that it was potentially true whereas leaving the information in could hide it among the other data that was simply fiction.
Granted, if there IS code that has to be removed, there are plenty of folks with nothing better to do than to do diffs on every release to see what the code was.
Since the code is now visible, the cat's out of the bag. However, if this SCO issue IS true and no real penalties can be assessed (actually, it doesn't matter that the person who put the code in is penalized since this type of thing can, in some folks' opinions, potentially ruin companies and cause lots of folks to lose their jobs), it is still a bad thing in that it rewards behavior that is found unacceptable (in the legal world at least). One fear is that it simply encourages others to "steal" code from their employer and release it in an OSS product and letting that cat out of the bag as well - either through vindictive behavior or for some other reason like it would make the person doing this have some "heroic" reputation among the community.
If there is no penalty for doing so, then this could be a big issue to commercial software and I'm still not convinced that any of the OSS folks have thought about what that could mean in the long term.
- It was never said that the drivers wouldn't work for everything else.
- If you ever paid for a driver update from either nVidia or ATI, you got ripped.
What I think we'll find is that only the games that have benchmark capability and are, or are going to be, popular enough for large market appeal - the HalfLife2s, the Quake/Dooms, etc. - are the ones that will get optimizations by the card vendors. This is because it will help them sell cards.
This is good to hear. I ordered a Dell Inspiron 8600 last week w/ 1.4GHz Pentium-M.
Nope... just saying that the fighter analogy is a poor one. That and I don't want a USA-OS and a USA-CPU, not that we can control what other countries do. Also that it is wrong for the USA Government to compete with the private/commercial sector according to our (USA) laws.
Actually, it is illegal in the USA for the government to compete with commercial industry. That's why fighter planes are defined by a big list of specifications (has to fly mach 1+ at 90% power, carry a payload of X, etc.) which are given to a number of airplane designers/manufactuerers and those companies bid on the cost of making the plane. Some review board then looks at the prototypes and such and decides which meet the specifications and which do not, and decides which company gets the contract. The difference in the previous scenario is that the new fighter plane has requirements that are not met by an existing plane in inventory. That's why they need a new one.
If the requirements of a government software project are met by an existing set of software from a commercial business, they aren't supposed to give money or commission something else. If their requirements are met by some software on MSWindows or Linux, then they are supposed to use what is there. If the software doesn't exist, then they provide specs and get bids for the work. etc.
Having a USA OS (doesn't matter what the thing is) is a VERY, VERY bad thing. Same as having the USA-2680 CPU (some made up processor). Other countries can do what they want, but in the USA there are laws (and good reasons for them) why this should not be.
Care to put your name behind your words? Didn't think so...
My experience has been that Java was slower than C#.NET apps on the same hardware and the same level of optimizations.... sometimes not by much, sometimes by a bit.
But what happens in two years when MS changes to something completely different and your hojillion lines of .NET code are worthless?
The same thing that happens now? You keep your hardware/software as it is and keep using the thing. I've seen companies that have controller software written for Apple ][e machine that use those things *to this day* because it works and does the job. They have to buy all the Apple ][e machines/parts they can from eBay and such to make sure they can keep their hardware working. How long ago was the Apple ][e put on the unsupported list?
You'd be surprised just how many companies still use MSDOS programs as their main software platform, I guess. Just because it is there is not sufficient reason for you to have to upgrade to it. If it becomes unsupported from the company that makes it, you can still live on.
Heh... we had a Java project and we decided to use JBuilder for it. I had a PentiumIII @ 850MHz and 768M of memory and the thing took eons to do anything. I found it somewhat inconsistent. While it basically worked, it was extremely slow (I don't know if this is because it was all written in Java or what) to the point of breaking my flow at times. I'd not want to use it again.
Not to mention that the checksum calculations will destroy the cache and your performance will go into the crapper. I can't imagine this as being a viable option.
The only real option is to have a voter system where multiple nodes (or subclusters) all calculate the same thing at the same time. You compare all the answers and take the answer set that has the most nodes (or subclusters) with that result.
what did people do before there were web browsers at work? Oh, yeah, they did THEIR JOB.
Not necessarily more or less than after web browsers came along. I usually think that people who have to go outside for 10 minutes every hour to smoke a cigarette spend more time away from their desk than the time I spend browsing the web or whatever. And those people tend to browse the web as much as I do on top of that. Before browsers, people would hang around the water cooler, in the break room, or in the copy room and talk. There might have been some increase in "non-productive" time since web browsers but probably 1/2 or more of the stuff I browse can potentially apply to work (I visit a number of tech sites that frequently have information that is useful to me as a part of my daily ritual).
Heh... so he likes the fact that they can't download stuff to use on their machines. Sounds kind of draconian in the end. You will use your machine for nothing but work! I have had bosses that tried to do similar things in the past and all they really did is cause folks to leave the company because things were so strict and you basically couldn't take breaks anymore other than pushing back from your desk and staring at the wall for 10 minutes. Of course, the people who are most able to get other jobs move on first and the people who can't, stay. People will find a way, and even pride themselves in their creativity, of jacking the system using breaks and finding ways to goof off. Sure. The amount of time these folks may work may go up but the quality/productivity may go down so the net result will be a loss.
Yet-another-person who fails to grasp emoticons.
Sounds like every Mac user pre-ordered then ;)
not necessarily of the Linux kernel, but of the apps on the platform.
1. I want to download an app and it puts me in dependency hell. I have to download 10 other packages in order to actually use that app, and 8 of those packages break other apps I have installed.
2. Inconsistencies in GUI among apps. In some apps, right click pulls up menus, in others it doesn't - or it is some other random sequence of mouse buttons. In some apps, the "normal" menu items are scattered all over hell's half acre compared to where the "normal" menu items are for another application. ("Normal" being file open/save, edit, copy/paste, etc.) In some apps, ones with multiple windows, which window actually contains the menu you want may have no correlation with what you are actually doing in that window.
3. Usability of GUI among apps. In many applications, the GUI appears to have been put together either simply by order of implementation of the features or by some obscure/strange workflow that makes no sense to anyone other than the person who developed the GUI. Many times the GUI seems very haphazardly assembled with no regard for workflow patterns which are normally used in the type of processing performed by the application.
I understand that for you, no flame intended, the O/S is more important than the data you generate with your "own" computer.
Actually... interoperability with data I produce is a big concern. It is silly of me to generate data that someone else cannot use because I'm using some non-standard and/or off-the-wall app. Knowing that a certain platform has built-in support for a certain type of data eliminates concerns that I have of interoperability.
Knowing that a platform has certain features built-in means that I can make use of that in software that I write. For example, for help files, I *know* that a particular platform has an HTML browser (built-in or bundled) so I can write all my help files in HTML without worry of some user not being able to access the help files I produce. I do not have to generate a big long list of dependencies that users must satisfy before they can use the software I built. (Remember my previous statements of how users like to find out after-the-fact that they have to add more things to their system in order to use something they just got. This is one of the top three most infuriating things to me about Linux... I find some piece of software that I want to use only to find out that I have to spend half a day downloading and installing the dependencies for it before I can use it. The other most infuriating things being inconsistencies and usability of UIs.)
(offtopic)
Man, I miss my Atari1040ST. Unfortunately, it gave up the ghost a few years back and I didn't need it enough to warrant trying to fix it.
I liked the Amigas a lot even though I was never an owner.
I used to sometimes wonder what computing would be like today had IBM gone with the 68k instead of the 8088 for the IBM PC... we can all dream I guess.