Translation is hard and ungrateful work, and the people who need it doesn't know how to do it. That's very different from kernel hacking, where there is people who know how to do what is lacking, and even from documentation, where people write stuff based on their learning experiences (which can be based on irc/mailing lists).
Alas, that's not the whole problem. Each time a programmer uses a message string embedded in the program (in the C source), he makes it harder to translate the software. What are people supposed to do? Go through all source files and change each message by hand? And then what, have a "pt-BR" distribution, with different binaries?
If the OSS operating systems are ever going to get on par with the commercial operating systems, they need to standarize on message tables, so that changing language does not require recompiling anything, and supports even multiple languages at the same time (with the time-honored LANG environment variable).
I do not think the results confusing at all. The poll shows how the user thinks, instead of pointing the most wanted feature. "Most wanted" things do not work in this field, you need a well-balanced solution.
Software classes refer to how important the users rate each "class" (type, kind) of software.
"But I'm not rejoicing for the ascendancy of Linux or its penguin mascot."
In other words, if it's not BSD, he doesn't want it to succeed. We don't want users to have a choice. No, we just want them to have BSD.
He never said he was loathing the Linux ascendancy. He just doesn't care.
Observation 1: Linux zealots have a "if you are not with us, you are against us" mentality.
" If I had to pick out the single difference between the BSD community as a whole and the proponents of Linux, I would say it is something called ''correctness.''"
Ok...if that is true, then that is a very important point. But is it true? How can we know? Not from this article. It doesn't cite any ways in which BSD is more "correct" than Linux. I guess we're just supposed to trust Mr. Garfinkel.
Oh, but it does. He explains that the *BSD development teams are loath to accept code that "just works" but does not actually correct problems, but just get around them. Not in so many words, I admit, but I'm just translating the concept for you benefit.
I personally think this might be a biased view, but it is the same view that results in the "*BSD development is closed, Linux development is not" myth. If you do not accept the former, then you ought not to accept latter.
Alas, this is not relevant. He claimed *BSD development model is superior because it puts greater emphasis in getting it right instead of getting it working. If you are interested in this, then you should get off your ass and do further research for yourself. The development models are all documented, you just have to go and compare them. The article has no more obligation of explaining this sort of thing in details than it has of explaining what an Operating System is.
Observation 2: Whenever confronted with uncomfortable statements, Linux zealots will talk a lot to confuse the issue, and then state that nothing was really explained, and so it doesn't merit further investigation (or, at least, they try the very best to dismiss the issue without further investigation).
" This may be one reason proponents of Linux are frequently slow to admit the debt they owe to the Computer Science Research Group at Berkeley, which created BSD, and to the Free Software Foundation at MIT"
Oh, now I get it. I know Mr. Garfinkel's true identity. *wink* But why is he calling it Linux rather than GNU/Linux?
If you "get it", then why don't you share it with the rest of us?
He is calling it Linux because that's the name it is known by the public. This was not a slashdot article, it was a general press article directed to a wider public.
Observation 3: Whenever a chance comes by, Linux zealots will try to instigate suspicions that the "opponent" (ie, the one who is not "with us") has ulterior motives, instead of actually being objective/factual/honest.
Now, I can agree with this point. However, that "Linux is a fine operating system" is not the impression that I get from the rest of the article.
So? It is not an article about Linux, it's an article about *BSD. He is not talking about Linux strengths, he is talking about *BSD strengths. He is not even making a formal comparision about Linux and *BSD.
Observation 4: Linux zealots think the world revolves around them, and thus treat any comments as being in relation to Linux/Linuxen.
This could have been a really good article. It included some real meat, unlike most BSD articles in mainstream mags. However, it is ruined by the author's arrogance. An arrogance which unfortunately seems to be abundant among BSD users.
It only seems arrogant to those that fall into the "observation 4" category. "Hey, he is not talking good things about us! The arrogant bastard!".
Hotmail runs Solaris *and* FreeBSD (Solaris only being responsible for the database backends).
The greater availability of cracks for Linux in comparision to *BSD most surely does imply that there is a greater chance of a Linux box being compromised than a *BSD one. Simple mathematics.
So, the author was correct in all respects, and you were not.
Is the copyright notice "as it is in paragraph 1 and paragraph 2 section c" identical to the two clauses of the small BSD license (I assume you are excluding the BSD licenses with the claim-credit clauses)? If they are not ipsis literis identical, then they are not the same.
We do not claim that less people can touch FreeBSD sources thus, ergo, etc, etc. We claim *more* people can touch FreeBSD source than can touch Linux.
It's just that we have a method, a source control system and a philosophy of work that results in a superior source code, on average.
Linux does have more features, because it has more people developing for it (but *not* more people who can change the source code), and has some pretty good stuff because it has some pretty good developers.
Anyone thinking FreeBSD (in particular) doesn't "release" often enough... well, just doesn't get it. FreeBSD's releases are just points at which it goes into cds.
If you want to know at which pace FreeBSD progresses, send a message to majordomo@freebsd.org subscribing to cvs-all.
Cool? As of today, we can sleep in peace[2], knowing that our scientists can create diseases to protect us from our enemies, if need be.
Sure, it also means we have opened a door to much greater understanding of diseases, and new ways to fight them. But it is a two-edged sword, and we better not forget that.
[1] Read Reality Disfunction, by Peter F. Hamilton. You'll thank me.
[2] "Man, let me tell you, if they ever devise a way to generate energy from sarcarsm, you'll all owe me big time!" -- Bill Paul
Nice theory from someone who doesn't actually have a clue about the present situation in Japan. The last thing one would imagine is that computers are limited to "techies" here. On the other hand, *home* use seems to be very limited.
As for Turbo Linux numbers, it probably doesn't include Windows 98 OEM, and includes Turbo Linux distributed with magazines. Only way to explain these numbers.
I think the guy had a valid point -- while FreeBSD may run many Linux binaries, it is by no means as easy to do for non-trivial programs as some people like to say.
First, that's not what the guy said, nor what I replied to. He said "no apps", I pointed to a large collection of apps.
Second, the problem with applixware is installing the damn thing, not running it. As for Wordperfect, I never heard of any problem at all with it.
Actually, two. First, even in the real world, getting the run queue in Linux to get more than 2 things in it is HARD.
My real world is obviously different from yours.:-)
Second, you usually don't want to run your server peaked. ftp.cdrom.com is not peaked even when its putting out 1TB a day.
Of course you don't run your server "peaked". That would be a contradiction in terms. Peak is what happens when the workload reaches it's highest point.
Now, you can size your hardware to be able to handle the peaks, and idly throw away cycles during the averages (perhaps running SETI@Home?:), or you can size your hardware to handle the average, and depend on your operating system not to screw up at the peaks.
Take your pick.
Now, if FreeBSD performed worse than Linux under the Mindcraft test, we can also deduce that its TCP/IP stack is worse than Linux's stack.
And why would that be? As I recall, it was reported that *Linux* did not perform well because of TCP/IP stack caps. What that has to do with the reasons of FreeBSD bad performance?
The reason why 386BSD was "splitted" into NetBSD, FreeBSD and OpenBSD is because the developers had a different view of what should go into an operating system, and what the priorities ought to be.
That is the same reason why there are different Linux distributions. People have different ideas about what should go in the operating system, and have different priorities.
btw, I thought Red Hat was distributed with a kernel using some Red Hat specific patches...
Sure. The Gartner Group had a test with NT, Linux and FreeBSD. NT sucked, of course. FreeBSD was first, with Linux a somewhat close second. The test, unfortunately, is not available on the web anymore.
What they benchmarked was the ability of the system to degrade gracefully underload. In other words, forget how fast your system is when everything is cached in RAM... when things get ugly, you have a lot of processes, and swap is heavy, during the "peaks" of your workload, how well does the system handle that? FreeBSD showed the best scores by far under the heavier loads, with the scores (err, FreeBSD and Linux) getting closer under lighter loads, and both NT and Linux surpassing FreeBSD under no load to speak of.
As it goes, they somehow reached the conclusion that Linux was the prefered server under a number of environments analysed, even though FreeBSD performed better and had none of the drawnbacks they found in Linux. Go figure.
Then, there is the Mindcraft benchmark. I'm told Linux performed significantly better than FreeBSD, with NT performed way better, and Solaris beating everyone. Alas, this test measured the perfomance of the system under very few processes, and everything cached. At least, as far as NT went. Gartner results say that if the environment of the test changed so that number of processes increased and swapping became necessary, NT would soon find itself trailing everyone.
Contrary to what previous repliers said, you cannot do it. BSD license says it's terms must be reproduced, and GPL license requires that no other restrictions (such as reproducing the terms of the original license) must be made.
What the author is missing is that the resentment from the BSD community comes not from the success of GPL license itself, but from the slanting GPL community makes of FreeBSD.
GNU org itself has been part of this. The effect is so widespread that even the author, who seems to be quite unbiased, shows signs of it!
You can see that the way he casually references the claim-credit clause as "advertising clause", an expression intended to reinforce the false notion that advertisement of a software containing BSD-licensed code must include the credits of that code. Not only this is false, but GPL advocacy sites have use this argument to downplay the BSD license even though not all of them even contain such clause!
The strong reaction seen nowadays from the BSD community toward the licensing issue is trying to dispell this and others false claims.
The author, otherwise, is very objective and precise in it's arguments. One thing, though, caught my attention and needs to be corrected.
Though it's true that there are people in the BSD community that associate the GPL with the "communist" ideas expressed in the GNU Manifesto, there is a couple of problems GPL imposes on commercial software which BSD doesn't have.
One problem is that GPL code not only insures that the GPLed code will remain available, but also calls for all other code added to that to be licensed as GPL. For business which rely on proprietary code as their raison d'etre, this prevents the use of GPL code. Notice that many on the BSD community do work on such companies, and while they actively contribute to BSD-centered efforts, they do need to keep some of the code to themselves.
Another problem is that while GPL code can easily be used on commercial "package" softwares, that is not true of turn-key and rommable systems, such as net appliances. This is not the same thing as distributing documentation... you have to ensure that the product is made available with the correct version of the source code, which can be quite expensive.
I hope this helps further clarify the issue. I'd like to thank the author for what is, otherwise, an unbiased, concise and very clear description of this "conflict".
Linux VM system doesn't scale. You can tune it to a given load, but it just won't be able to adapt to different loads.
Though I suspect this issue is not relevant for wcarchive. For one thing, since it's sole purpose is serving files through ftp, it doesn't need to adapt to different load usages. For another thing, I doubt it ever swaps... Not that Linux can handle 4 Gb of RAM on ia32, but...
Nevertheless, there are still gains in the handling of memory on the FreeBSD. David Greeman _is_ one of the responsibles for the vm subsystem in FreeBSD, and has been so for a long time.
The second you have to access two disks simultaneously, IDE looses. If you make each one a master, then you are really putting them on two different controllers.
If you have lots of disk access, lots of disk, there is no way IDE can compete. Even more if you want your CPU time for other things. Ever heard of tag queues?
"Unscientific"? What is it in your definition of "scientific" that makes this claim "unscientific"?
This is *NOT* like a golfer alone at midnight. WCArchive is accessible 24 hours a day, and you can see it's logs at any time. Moreover, there are a lot of people out there partially responsible for this record.
FreeBSD 3.2 supports nfs3, and it's pretty stable as long as you keep to UDP. NFS/TCP still have it's share of problems.
As for performance, you may have to tune nfs sysctl settings. On a recent thread on one of the FreeBSD lists, for instance, it became clear that OS/2 clients interacted badly with FreeBSD's NFS write-gathering.
Also, Linux (by default, at least), violates NFS2 specs by immediatly acking write requests, instead of waiting until they are committed to disk. FreeBSD has this tunable, with the default being not violating the specification. But if you are going to use nfs3, that's a non-issue.
You forgot to mention that after the Microsoft flooded the market with the "proprietary fork", and all Linux users started using it for unknown reasons, we would end up in a world where the dominant OS didn't actually s*ck.
There are two things *I* would like to know, though:
1) Why is it such a bad thing if MS sold a decent OS? and 2) Why is it that Linux users seem absolutely convinced they would adopt the Microsoft version if Microsoft released a proprietary fork of it?
Slashdot Mirror
Alas, that's not the whole problem. Each time a programmer uses a message string embedded in the program (in the C source), he makes it harder to translate the software. What are people supposed to do? Go through all source files and change each message by hand? And then what, have a "pt-BR" distribution, with different binaries?
If the OSS operating systems are ever going to get on par with the commercial operating systems, they need to standarize on message tables, so that changing language does not require recompiling anything, and supports even multiple languages at the same time (with the time-honored LANG environment variable).
I do not think the results confusing at all. The poll shows how the user thinks, instead of pointing the most wanted feature. "Most wanted" things do not work in this field, you need a well-balanced solution.
Software classes refer to how important the users rate each "class" (type, kind) of software.
Let's see how a *BSD partisan sees you.
"But I'm not rejoicing for the ascendancy of Linux or its penguin mascot."
In other words, if it's not BSD, he doesn't want it to succeed. We don't want users to have a choice. No, we just want them to have BSD.
He never said he was loathing the Linux ascendancy. He just doesn't care.
Observation 1: Linux zealots have a "if you are not with us, you are against us" mentality.
" If I had to pick out the single difference between the BSD community as a whole and the proponents of Linux, I would say it is something called ''correctness.''"
Ok...if that is true, then that is a very important point. But is it true? How can we know? Not from this article. It doesn't cite any ways in which BSD is more "correct" than Linux. I guess we're just supposed to trust Mr. Garfinkel.
Oh, but it does. He explains that the *BSD development teams are loath to accept code that "just works" but does not actually correct problems, but just get around them. Not in so many words, I admit, but I'm just translating the concept for you benefit.
I personally think this might be a biased view, but it is the same view that results in the "*BSD development is closed, Linux development is not" myth. If you do not accept the former, then you ought not to accept latter.
Alas, this is not relevant. He claimed *BSD development model is superior because it puts greater emphasis in getting it right instead of getting it working. If you are interested in this, then you should get off your ass and do further research for yourself. The development models are all documented, you just have to go and compare them. The article has no more obligation of explaining this sort of thing in details than it has of explaining what an Operating System is.
Observation 2: Whenever confronted with uncomfortable statements, Linux zealots will talk a lot to confuse the issue, and then state that nothing was really explained, and so it doesn't merit further investigation (or, at least, they try the very best to dismiss the issue without further investigation).
" This may be one reason proponents of Linux are frequently slow to admit the debt they owe to the Computer Science Research Group at Berkeley, which created BSD, and to the Free Software Foundation at MIT"
Oh, now I get it. I know Mr. Garfinkel's true identity. *wink* But why is he calling it Linux rather than GNU/Linux?
If you "get it", then why don't you share it with the rest of us?
He is calling it Linux because that's the name it is known by the public. This was not a slashdot article, it was a general press article directed to a wider public.
Observation 3: Whenever a chance comes by, Linux zealots will try to instigate suspicions that the "opponent" (ie, the one who is not "with us") has ulterior motives, instead of actually being objective/factual/honest.
Now, I can agree with this point. However, that "Linux is a fine operating system" is not the impression that I get from the rest of the article.
So? It is not an article about Linux, it's an article about *BSD. He is not talking about Linux strengths, he is talking about *BSD strengths. He is not even making a formal comparision about Linux and *BSD.
Observation 4: Linux zealots think the world revolves around them, and thus treat any comments as being in relation to Linux/Linuxen.
This could have been a really good article. It included some real meat, unlike most BSD articles in mainstream mags. However, it is ruined by the author's arrogance. An arrogance which unfortunately seems to be abundant among BSD users.
It only seems arrogant to those that fall into the "observation 4" category. "Hey, he is not talking good things about us! The arrogant bastard!".
Sigh.
The only misinformation was your own message.
Yahoo doesn't run Linux.
Hotmail runs Solaris *and* FreeBSD (Solaris only being responsible for the database backends).
The greater availability of cracks for Linux in comparision to *BSD most surely does imply that there is a greater chance of a Linux box being compromised than a *BSD one. Simple mathematics.
So, the author was correct in all respects, and you were not.
Is the copyright notice "as it is in paragraph 1 and paragraph 2 section c" identical to the two clauses of the small BSD license (I assume you are excluding the BSD licenses with the claim-credit clauses)? If they are not ipsis literis identical, then they are not the same.
We do not claim that less people can touch FreeBSD sources thus, ergo, etc, etc. We claim *more* people can touch FreeBSD source than can touch Linux.
It's just that we have a method, a source control system and a philosophy of work that results in a superior source code, on average.
Linux does have more features, because it has more people developing for it (but *not* more people who can change the source code), and has some pretty good stuff because it has some pretty good developers.
Anyone thinking FreeBSD (in particular) doesn't "release" often enough... well, just doesn't get it. FreeBSD's releases are just points at which it goes into cds.
If you want to know at which pace FreeBSD progresses, send a message to majordomo@freebsd.org subscribing to cvs-all.
Oh, come on now... Cooler mascot??? How can a daemon be cooler than a penguin??? Logic clearly dictates FreeBSD's mascot is hotter.
Cool? As of today, we can sleep in peace[2], knowing that our scientists can create diseases to protect us from our enemies, if need be.
Sure, it also means we have opened a door to much greater understanding of diseases, and new ways to fight them. But it is a two-edged sword, and we better not forget that.
[1] Read Reality Disfunction, by Peter F. Hamilton. You'll thank me.
[2] "Man, let me tell you, if they ever devise a way to generate energy from sarcarsm, you'll all owe me big time!" -- Bill Paul
Nice theory from someone who doesn't actually have a clue about the present situation in Japan. The last thing one would imagine is that computers are limited to "techies" here. On the other hand, *home* use seems to be very limited.
As for Turbo Linux numbers, it probably doesn't include Windows 98 OEM, and includes Turbo Linux distributed with magazines. Only way to explain these numbers.
[yes, I live in Japan]
Weird language, english.
I think the guy had a valid point -- while FreeBSD may run many Linux binaries, it is by no means as easy to do for non-trivial programs as some people like to say.
First, that's not what the guy said, nor what I replied to. He said "no apps", I pointed to a large collection of apps.
Second, the problem with applixware is installing the damn thing, not running it. As for Wordperfect, I never heard of any problem at all with it.
Actually, two. First, even in the real world, getting the run queue in Linux to get more than 2 things in it is HARD.
:-)
:), or you can size your hardware to handle the average, and depend on your operating system not to screw up at the peaks.
My real world is obviously different from yours.
Second, you usually don't want to run your server peaked. ftp.cdrom.com is not peaked even when its putting out 1TB a day.
Of course you don't run your server "peaked". That would be a contradiction in terms. Peak is what happens when the workload reaches it's highest point.
Now, you can size your hardware to be able to handle the peaks, and idly throw away cycles during the averages (perhaps running SETI@Home?
Take your pick.
Now, if FreeBSD performed worse than Linux under the Mindcraft test, we can also deduce that its TCP/IP stack is worse than Linux's stack.
And why would that be? As I recall, it was reported that *Linux* did not perform well because of TCP/IP stack caps. What that has to do with the reasons of FreeBSD bad performance?
The reason why 386BSD was "splitted" into NetBSD, FreeBSD and OpenBSD is because the developers had a different view of what should go into an operating system, and what the priorities ought to be.
That is the same reason why there are different Linux distributions. People have different ideas about what should go in the operating system, and have different priorities.
btw, I thought Red Hat was distributed with a kernel using some Red Hat specific patches...
No apps? I suppose you never typed cd /usr/ports, did you?
Sure. The Gartner Group had a test with NT, Linux and FreeBSD. NT sucked, of course. FreeBSD was first, with Linux a somewhat close second. The test, unfortunately, is not available on the web anymore.
What they benchmarked was the ability of the system to degrade gracefully underload. In other words, forget how fast your system is when everything is cached in RAM... when things get ugly, you have a lot of processes, and swap is heavy, during the "peaks" of your workload, how well does the system handle that? FreeBSD showed the best scores by far under the heavier loads, with the scores (err, FreeBSD and Linux) getting closer under lighter loads, and both NT and Linux surpassing FreeBSD under no load to speak of.
As it goes, they somehow reached the conclusion that Linux was the prefered server under a number of environments analysed, even though FreeBSD performed better and had none of the drawnbacks they found in Linux. Go figure.
Then, there is the Mindcraft benchmark. I'm told Linux performed significantly better than FreeBSD, with NT performed way better, and Solaris beating everyone. Alas, this test measured the perfomance of the system under very few processes, and everything cached. At least, as far as NT went. Gartner results say that if the environment of the test changed so that number of processes increased and swapping became necessary, NT would soon find itself trailing everyone.
By the way, the BSD license allows you to apply the GPL to a modified BSD work.
No, it doesn't. GPL requires that no other restrictions be made, BSD license requires that it's terms be preserved (which is a restriction).
Thus, you can't apply GPL over BSD.
BSD license requires that it's terms be preserved. This is a restriction to the distribution of the code and, as such, forbidden by the GPL.
BSD license, with or without claim-credit clause, CANNOT be GPLed.
Contrary to what previous repliers said, you cannot do it. BSD license says it's terms must be reproduced, and GPL license requires that no other restrictions (such as reproducing the terms of the original license) must be made.
What the author is missing is that the resentment from the BSD community comes not from the success of GPL license itself, but from the slanting GPL community makes of FreeBSD.
GNU org itself has been part of this. The effect is so widespread that even the author, who seems to be quite unbiased, shows signs of it!
You can see that the way he casually references the claim-credit clause as "advertising clause", an expression intended to reinforce the false notion that advertisement of a software containing BSD-licensed code must include the credits of that code. Not only this is false, but GPL advocacy sites have use this argument to downplay the BSD license even though not all of them even contain such clause!
The strong reaction seen nowadays from the BSD community toward the licensing issue is trying to dispell this and others false claims.
The author, otherwise, is very objective and precise in it's arguments. One thing, though, caught my attention and needs to be corrected.
Though it's true that there are people in the BSD community that associate the GPL with the "communist" ideas expressed in the GNU Manifesto, there is a couple of problems GPL imposes on commercial software which BSD doesn't have.
One problem is that GPL code not only insures that the GPLed code will remain available, but also calls for all other code added to that to be licensed as GPL. For business which rely on proprietary code as their raison d'etre, this prevents the use of GPL code. Notice that many on the BSD community do work on such companies, and while they actively contribute to BSD-centered efforts, they do need to keep some of the code to themselves.
Another problem is that while GPL code can easily be used on commercial "package" softwares, that is not true of turn-key and rommable systems, such as net appliances. This is not the same thing as distributing documentation... you have to ensure that the product is made available with the correct version of the source code, which can be quite expensive.
I hope this helps further clarify the issue. I'd like to thank the author for what is, otherwise, an unbiased, concise and very clear description of this "conflict".
Linux VM system doesn't scale. You can tune it to a given load, but it just won't be able to adapt to different loads.
Though I suspect this issue is not relevant for wcarchive. For one thing, since it's sole purpose is serving files through ftp, it doesn't need to adapt to different load usages. For another thing, I doubt it ever swaps... Not that Linux can handle 4 Gb of RAM on ia32, but...
Nevertheless, there are still gains in the handling of memory on the FreeBSD. David Greeman _is_ one of the responsibles for the vm subsystem in FreeBSD, and has been so for a long time.
The second you have to access two disks simultaneously, IDE looses. If you make each one a master, then you are really putting them on two different controllers.
If you have lots of disk access, lots of disk, there is no way IDE can compete. Even more if you want your CPU time for other things. Ever heard of tag queues?
Logs are there for anyone to see.
"Unscientific"? What is it in your definition of "scientific" that makes this claim "unscientific"?
This is *NOT* like a golfer alone at midnight. WCArchive is accessible 24 hours a day, and you can see it's logs at any time. Moreover, there are a lot of people out there partially responsible for this record.
FreeBSD 3.2 supports nfs3, and it's pretty stable as long as you keep to UDP. NFS/TCP still have it's share of problems.
As for performance, you may have to tune nfs sysctl settings. On a recent thread on one of the FreeBSD lists, for instance, it became clear that OS/2 clients interacted badly with FreeBSD's NFS write-gathering.
Also, Linux (by default, at least), violates NFS2 specs by immediatly acking write requests, instead of waiting until they are committed to disk. FreeBSD has this tunable, with the default being not violating the specification. But if you are going to use nfs3, that's a non-issue.
You forgot to mention that after the Microsoft flooded the market with the "proprietary fork", and all Linux users started using it for unknown reasons, we would end up in a world where the dominant OS didn't actually s*ck.
There are two things *I* would like to know, though:
1) Why is it such a bad thing if MS sold a decent OS? and
2) Why is it that Linux users seem absolutely convinced they would adopt the Microsoft version if Microsoft released a proprietary fork of it?