This is obviously a good thing to do for the Open Group. I wonder, though, why they say they grant the permission to the Linux Man Page Project. Their documents are freely redistributable anyway, so this in effect gives the content away.
Many Linux man pages were already direct or very close copies, any the BSD man pages are also very close. And why wouldn't they? The content is quite formulaic and there is not much space for exploring language creativity, it's technical documentation after all.
Java is in effect a way to shop a program with source code. The GPL does in no way force people to leave any comments in or leave the identifier names meaningful. So, shipping Java bytecode is technically the same as shipping the source code, you just need to run it through a decompiler. But what is the difference between extracting the source code with unzip and extracting it with a decompiler? The important point is that the GPL says the source code must be shipped, so it's OK to run the decompiler (which it otherwise, depending on your legislation, isn't).
You say they stripped information. That's not a GPL violation.
GPL enforcement is hard enough. It is not wise to make it harder by enforcing GPL on trivial code, where the other side may actually be able to successfully defend themselves in court. Losing a GPL enforcement lawsuit is much worse than some guy's trivial programming 101 "let's open a dialog box" code being misappropriated.
You ask why they stole the code if it was trivial. How should I know? In fact, how should I know whether it was really stolen? It may be a freak concidence that they used the same class, function and variable names (Java does have many conventions that make less not as unlikely as in other languages). In particular if you take the Base64 code, there is an RFC with example C code, so everyone recoding this in Java would probably reuse the identifiers.
I must congratulate you on your troll, though. The spelling could use some improvement, but other than that, it has ad hominem attacks, a straw man, appeals to higher authority, quoting stuff out of context. Not too bad for a beginner.
In the same sense that a school child playing with a water pistol is in violation of zero tolerance no-gun regulations.
Just because it would technically be your right to do something against this, does not mean you should actually do it.
In Java, it is normally possible to restore the source code almost completely, even down to variable names and type structures. That does not mean not giving the source code does not constitute a violation of the letter of the law, it just means that your case is not very strong.
In particular, if you complain to someone who took some trivial source code from you, they can just replace it. In the end, noone has any advantage from your complaint, and if you took them to court, you wasted public resources and made some lawyer weasel richer in the process. To them, it is just an annoyance to write some trivial code. Neither did the public gain anything from it, nor did you gain anything from it, nor does the company who took your code have a reason not to take your code again next time.
The question is not, why did they steal it, but how can you be sure they stole it. If it's a trivial function every Java programmer can do in his sleep, and everyone probably would do it large the same way.
Java, unlike perl, does not offer 20 ways to do every problem, so if ten people program a Base64 encoder, nine of them would probably do it the same way (they would read the RFC and translate the C code literally to Java). The tenth guy would probably do something else and break things in the process.;-)
I have all the sympathy for people suffering from GPL violation, but I find this case ridiculous for several reasons:
1. Java can be trivially decompiled, so I don't see how this can be regarded as "closed source" with a straight face.
2. Your library does not look like rocket science to me.
That does not lower the formal legal bar, but come on, how many ways are there to do Base64 encoding and circular buffers? I don't know what exactly you think someone took from you, but this looks to me like one of the junk patent cases where someone gets a patent on something blindingly obvious, like using names in urls, and the tries to sue others.
If some non-trivial code of yours was stolen, then, by all means, sue the bastards. First, I would talk to them, then I would talk to my lawyer, and then I would tell my lawyer to sue them. If you don't want to do that, you can sign over your copyright to the Free Software Foundation, and they will do the enforcement for you. But please, don't make the whole free software movement look like SCO by trying to enforce rights on trivial pieces of source code.
Your library offers a Swing dialog box for entering passwords, for crying out loud! That's like "my first programming project", it's the hello world of Swing programming. My opinion: come back when they took something worthwhile from you. This way you only make yourself look bad and give SCO and Microsoft ammunition on why free software people are communists and morally corrupt people.
They have now missed over a dozen opportunities to do something very stupid! Has someone removed the alien face huggers there or what is happening here?
This would have never happened with the old Novell we all loved to loathe.
I find this deeply disturbing. Stupidity does not simply go away just like that. Where is my tinfoil hat again?
How do you help society or mankind if you bring together 200 dot-com millionaires and let them talk about how to make more money by relaying emails?
Yeah I am oversimplifying this, but here in Europe, we have been doing geek camping for years. We have a geek summer camp every two years, alternating between Holland and Germany. And you know what? We let everyone in, not just the rich and famous.
And it pays off. This kind of elitism is bad for everyone. Where do you expect the next generation of good hackers to come from if you don't let them come to your hacker camps? Elitism leads to losing the ground under your feet.
And it makes the whole thing less fun. The harder you have to fight or the more you have to have achieved to be perceived as the "brightest and most intelligent" people who are then allowed in, the less you can just be yourself, the more this becomes an ego show where everyone is concerned about how to look good so he will be invited again next year. It's style over substance.
And frankly, who cares about the Google founders? Who cares about some egomanic bloggers who write up profound sounding essays on their blogs to keep their name in circulation? The people who are really important and interesting are people like Brewster Kahle (archive.org) and John Gilmore (eff).
a) they firewalled ICMP echo (WTF?!?) b) cvs pserver is not available and apparently never will be again. So I went through my checked out gcc source tree and changed all the CVS/Root files to their new scheme, but it didn't work, "directory not found". c) I would have double checked with the webcvs, but that's also not operational. d) The other option would have been to download a snapshot from the download area, but the download areas are also not available. OK ok, for gcc the download area is somewhere else, but for all the other projects?!
This begs the question: what _is_ back online? The web server with the note that they are back online?
So they discovered that pserver has security bugs. No, really? The solution is to provide pserver cvs in a chroot with a uid that can't write anything and maybe use systrace to disallow nasty operations.
Sorry, folks, but I don't like people who discontinue all the important features and then say it's for security reasons. That's bullshit. I would help, but I didn't see them asking for help anywhere.
How many companies do you know that actually create their own GUI applications? And I'm not talking about some internal admin interface here that some guy wrote for himself. Obviously, web interfaces don't count either, and neither does Java stuff.
And if an in-house application gets written, the planning stage until even the decision whether something will be written is done, even that phase easily costs a hundred fold of the Qt fee. Then you count in the time and productivity lost to internal training sessions, and you end up with numbers where the $1300 don't even register in the 0.x percentage range.
Also, since in-house applications are used internally and not given away, you don't need a license that allows selling the code, the GPL is perfectly fine for that.
To make this very clear: if there is anyone who is actually making graphical applications for Linux, and this argument has to carry any weight, it's neither a big company nor is it an in-house application. And if you expect to make money on an application, labor cost by far dwarfs any fixed hardware or licensing cost. Heck, MS Visual C++ alone probably costs more than the Qt license, and do you know any Windows development company who ever went under because of the MSVC license cost?
If Bruce Perens honestly wants this to be a Linux for business people (instead of the unwashed masses of normal users), he should not call it UserLinux but BusinessLinux or whatever.
I'm a user and I want KDE. Most people agree that KDE is more mature and robust than GNOME anyway, so from a business point of view it is obviously better suited. KDE also has more stability from other points of view, for example it doesn't change the default window manager for each major release, the groupware and the kiosk mode are very important as well. I'm not talking down on GNOME here, but KDE is more mature and all the major business wins Linux has had so far were with (and because of) KDE.
I think the maintainability argument is a fallacy. Admins already are completely unable to contain the complexities of different applications. Each major application and framework calls for its own class of admins. In large companies you have a Cisco admin for the networking infrastructure, you have an Oracle DBA, you have the Apache guy, you have the SuSE/RedHat/whatever admin, and the 5000 Windows reboot monkeys. Nobody expects all of this to go away if they switch to Linux. There will still be complexity. Deciding to standardize on GNOME will not make OpenOffice any less daunting to install and maintain in a multi-user environment. Or Mozilla. Or Apache.
And if we accept the argument, we would clearly choose the platform with the more robust administration interface, which clearly is KDE. kcontrol is integrated and pretty much all-encompassing, while GNOME is constantly shifting from CORBA over XML to a binary registry and back. GNOME has become so bad that they actually added a regedit style "config editor" and apparently really expect users to use it to configure applications. Hint: This is the kind of nightmare people want to get rid of when they switch from Windows to Linux.
Anyway, I don't see why we need to standardize on a GUI, and if we do, we standardize on KDE, of course, as it fulfils more of the requirements businesses have, hands down.
Actually, if you use "innovation" in the new-age business lingo of those hipster PR suit bearers, then it may actually be correct. These days it apparently counts as innovation if your software puts the menu choices in a different order than the previous version.
Innovation used to actually mean something a few years ago. *sigh*
So Solaris gets capabilies. Only, what, five years after Linux? And they finally abandon that obsolete slow-as-molasses file system of theirs? The level of technology leadership they are displaying is nothing short if breathtaking!1!!
On a lighter node, the article says their current partitioning scheme is software based. Good to know. Fits well in the general impression I got from them, with their shell script based "high availability" solution, and their industry leading "backup" solutions. There really is no need to know more than this about Sun and their software. Long live admintool!
I need to read mails from unknown people, because those are... my new customers!
How about remove Outlook and Internet Explorer instead and installing a secure email infrastructure. I have never ever, not even once, felt the need to not open an email because it might be insecure.
Advocating not opening emails is even worse than running exploitware from Microsoft in the first place.
They look like this because I used gnuplot, which can make great graphs, without really mastering it. I had trouble selecting colors that stood apart at all. That's why the colors for each OS chance from graph to graph, by the way, because the fifth color would have to be some ugly brown, and I didn't want to prejudice against one OS by giving it an ugly color.
I can give out the data on a case by case basis, though. Please email me if you really want it. Or you could open the graphs in Gimp, Photoshop or ImageMagick or so and change the colors yourself.
But since none of the graphs actually measures disk I/O, that does not matter in the least. You did read the web page before commenting on it, didn't you?
I did a benchmark about HTTP throughput that did not make it into the PDF or the web page: I downloaded one large ISO over Fast Ethernet. To make sure the test was fair, I mounted the same Linux ext2 file system on all the BSDs read-only. My disk has ~25 MB/sec throughput, so any OS should be able to saturate the Fast Ethernet link with ~11 MB/sec, right? OpenBSD couldn't.
But since the ext2 implementation in Linux is obviously at an advantage here, I did not publish these graphs, that would have been unfair. And I wanted to not feed the trolls.
I am not pretending to produce the be-all end-all benchmark here. That would at least include having an SMP server for the operating systems that support it, and it would have used Gigabit Ethernet. Frankly, I don't have the hardware to do that.
But I do have these benchmark programs, and I was hoping to spark enough interest in doing some real comparative benchmarks by posting the results here. The tools are now there, and they do use the platform specific API hacks to get the best performance on each operating system, so the biggest area of concern for comparative benchmarks should be out of the way -- that the benchmark program is skewed in one direction or the other.
If you are in possession of a real benchmarking lab with enough free disks and spare time on your hand to complete these benchmarks, please do so!
I'm not from the Linux mafia. If my goal was to discredit BSD, I would not have added support for kqueue in the first place. If my goal was to pat myself on the back, I would not have used the BSD sysctl to increase the number of files in the system (on Linux 2.6, you basically don't have to do that, just use ulimit), and that would have been fair, if only in the Netcraft/Microsoft sense.;-)
But neither am I from the Intel scalability labs with 1000 identical PCs and multiple gigabit ethernet connections to the server. I'm just a guy with a notebook who happened to have a little spare space on his hard disk and decided to include the BSDs in the benchmarks.
Nowhere in the commentary is any negative conclusion drawn from the fact that the FreeBSD fork graph stops early.
There are several reasons why I didn't pursue this. First of all, I was running out of time, I finished the benchmarks the night before driving to the conference. Secondly, I didn't fly in Linus to hand tune my Linux setup as well. Third, FreeBSD does not look bad because of this, on the contrary. The graph looks O(1), it just stops early. It is true that this test should be redone now that I know how to change this setting, but honestly, I don't think the FreeBSD crowd has anything to complain about here. After all, FreeBSD brings home the trophies for the BSD camp.
The article basically quotes some guy (who is actually selling Samba and thus has a vested interest) saying that Samba is 2.5 times faster than Windows 2003.
Now I have no reason not to believe him, but I was expecting a little more. And I'd wager the suits considering switching to Samba also expect more.
Forget about xlib compatibility for a moment. This is not just some software wrapper implementing xlib on top of some crackpot graphics interface (while this is tedious and time consuming work, it can be done by any programmer).
This is someone fusing the microkernel and exokernel concepts, which is way more exciting than fusing some implementation stuff together. This is actually computer science research what this man is doing, not just code hacking. I wish the man all the best for his work, for new directions in operating systems research are sorely needed, and exokernels are a particularly interesting concept (the exokernel inventors at MIT showed some really incredible performance gains, but their exokernel implementation was never aspiring to be a general purpose OS like this project).
I will try to monitor this project and maybe even help a little.
Yes, the Linux init process is also slow because of the artificial serialization.
But it is also slow because of the shells, many processes and glibc. I invite you to measure the difference yourself if you don't believe me. minit will do both, by the way, start stuff in parallel, and allow you to get rid of glibc and bash in your start up scripts.
Slashdot Mirror
This is obviously a good thing to do for the Open Group. I wonder, though, why they say they grant the permission to the Linux Man Page Project. Their documents are freely redistributable anyway, so this in effect gives the content away.
Many Linux man pages were already direct or very close copies, any the BSD man pages are also very close. And why wouldn't they? The content is quite formulaic and there is not much space for exploring language creativity, it's technical documentation after all.
Which part of "That does not lower the formal legal bar" did you not understand?
I didn't say it was not a violation. I said it he has a weak case.
I said distributing Java bytecode is like distributing the source code.
I did not say it is enough to satisfy the GPL requirements.
But please, don't let the facts stand in the way of your pathetic trolling.
Java is in effect a way to shop a program with source code. The GPL does in no way force people to leave any comments in or leave the identifier names meaningful. So, shipping Java bytecode is technically the same as shipping the source code, you just need to run it through a decompiler. But what is the difference between extracting the source code with unzip and extracting it with a decompiler? The important point is that the GPL says the source code must be shipped, so it's OK to run the decompiler (which it otherwise, depending on your legislation, isn't).
You say they stripped information. That's not a GPL violation.
GPL enforcement is hard enough. It is not wise to make it harder by enforcing GPL on trivial code, where the other side may actually be able to successfully defend themselves in court. Losing a GPL enforcement lawsuit is much worse than some guy's trivial programming 101 "let's open a dialog box" code being misappropriated.
You ask why they stole the code if it was trivial. How should I know? In fact, how should I know whether it was really stolen? It may be a freak concidence that they used the same class, function and variable names (Java does have many conventions that make less not as unlikely as in other languages). In particular if you take the Base64 code, there is an RFC with example C code, so everyone recoding this in Java would probably reuse the identifiers.
I must congratulate you on your troll, though. The spelling could use some improvement, but other than that, it has ad hominem attacks, a straw man, appeals to higher authority, quoting stuff out of context. Not too bad for a beginner.
As I wrote, legally it still is a GPL violation.
In the same sense that a school child playing with a water pistol is in violation of zero tolerance no-gun regulations.
Just because it would technically be your right to do something against this, does not mean you should actually do it.
In Java, it is normally possible to restore the source code almost completely, even down to variable names and type structures. That does not mean not giving the source code does not constitute a violation of the letter of the law, it just means that your case is not very strong.
In particular, if you complain to someone who took some trivial source code from you, they can just replace it. In the end, noone has any advantage from your complaint, and if you took them to court, you wasted public resources and made some lawyer weasel richer in the process.
To them, it is just an annoyance to write some trivial code. Neither did the public gain anything from it, nor did you gain anything from it, nor does the company who took your code have a reason not to take your code again next time.
The question is not, why did they steal it, but how can you be sure they stole it. If it's a trivial function every Java programmer can do in his sleep, and everyone probably would do it large the same way.
;-)
Java, unlike perl, does not offer 20 ways to do every problem, so if ten people program a Base64 encoder, nine of them would probably do it the same way (they would read the RFC and translate the C code literally to Java). The tenth guy would probably do something else and break things in the process.
I have all the sympathy for people suffering from GPL violation, but I find this case ridiculous for several reasons:
1. Java can be trivially decompiled, so I don't see how this can be regarded as "closed source" with a straight face.
2. Your library does not look like rocket science to me.
That does not lower the formal legal bar, but come on, how many ways are there to do Base64 encoding and circular buffers? I don't know what exactly you think someone took from you, but this looks to me like one of the junk patent cases where someone gets a patent on something blindingly obvious, like using names in urls, and the tries to sue others.
If some non-trivial code of yours was stolen, then, by all means, sue the bastards. First, I would talk to them, then I would talk to my lawyer, and then I would tell my lawyer to sue them. If you don't want to do that, you can sign over your copyright to the Free Software Foundation, and they will do the enforcement for you. But please, don't make the whole free software movement look like SCO by trying to enforce rights on trivial pieces of source code.
Your library offers a Swing dialog box for entering passwords, for crying out loud! That's like "my first programming project", it's the hello world of Swing programming. My opinion: come back when they took something worthwhile from you. This way you only make yourself look bad and give SCO and Microsoft ammunition on why free software people are communists and morally corrupt people.
They have now missed over a dozen opportunities to do something very stupid! Has someone removed the alien face huggers there or what is happening here?
This would have never happened with the old Novell we all loved to loathe.
I find this deeply disturbing. Stupidity does not simply go away just like that. Where is my tinfoil hat again?
How do you help society or mankind if you bring together 200 dot-com millionaires and let them talk about how to make more money by relaying emails?
Yeah I am oversimplifying this, but here in Europe, we have been doing geek camping for years. We have a geek summer camp every two years, alternating between Holland and Germany. And you know what? We let everyone in, not just the rich and famous.
And it pays off. This kind of elitism is bad for everyone. Where do you expect the next generation of good hackers to come from if you don't let them come to your hacker camps? Elitism leads to losing the ground under your feet.
And it makes the whole thing less fun. The harder you have to fight or the more you have to have achieved to be perceived as the "brightest and most intelligent" people who are then allowed in, the less you can just be yourself, the more this becomes an ego show where everyone is concerned about how to look good so he will be invited again next year. It's style over substance.
And frankly, who cares about the Google founders?
Who cares about some egomanic bloggers who write up profound sounding essays on their blogs to keep their name in circulation? The people who are really important and interesting are people like Brewster Kahle (archive.org) and John Gilmore (eff).
a) they firewalled ICMP echo (WTF?!?)
b) cvs pserver is not available and apparently never will be again. So I went through my checked out gcc source tree and changed all the CVS/Root files to their new scheme, but it didn't work, "directory not found".
c) I would have double checked with the webcvs, but that's also not operational.
d) The other option would have been to download a snapshot from the download area, but the download areas are also not available. OK ok, for gcc the download area is somewhere else, but for all the other projects?!
This begs the question: what _is_ back online? The web server with the note that they are back online?
So they discovered that pserver has security bugs. No, really? The solution is to provide pserver cvs in a chroot with a uid that can't write anything and maybe use systrace to disallow nasty operations.
Sorry, folks, but I don't like people who discontinue all the important features and then say it's for security reasons. That's bullshit.
I would help, but I didn't see them asking for help anywhere.
How many companies do you know that actually create their own GUI applications? And I'm not talking about some internal admin interface here that some guy wrote for himself. Obviously, web interfaces don't count either, and neither does Java stuff.
And if an in-house application gets written, the planning stage until even the decision whether something will be written is done, even that phase easily costs a hundred fold of the Qt fee. Then you count in the time and productivity lost to internal training sessions, and you end up with numbers where the $1300 don't even register in the 0.x percentage range.
Also, since in-house applications are used internally and not given away, you don't need a license that allows selling the code, the GPL is perfectly fine for that.
To make this very clear: if there is anyone who is actually making graphical applications for Linux, and this argument has to carry any weight, it's neither a big company nor is it an in-house application. And if you expect to make money on an application, labor cost by far dwarfs any fixed hardware or licensing cost. Heck, MS Visual C++ alone probably costs more than the Qt license, and do you know any Windows development company who ever went under because of the MSVC license cost?
Neither do I.
If Bruce Perens honestly wants this to be a Linux for business people (instead of the unwashed masses of normal users), he should not call it UserLinux but BusinessLinux or whatever.
I'm a user and I want KDE. Most people agree that KDE is more mature and robust than GNOME anyway, so from a business point of view it is obviously better suited. KDE also has more stability from other points of view, for example it doesn't change the default window manager for each major release, the groupware and the kiosk mode are very important as well. I'm not talking down on GNOME here, but KDE is more mature and all the major business wins Linux has had so far were with (and because of) KDE.
I think the maintainability argument is a fallacy. Admins already are completely unable to contain the complexities of different applications. Each major application and framework calls for its own class of admins. In large companies you have a Cisco admin for the networking infrastructure, you have an Oracle DBA, you have the Apache guy, you have the SuSE/RedHat/whatever admin, and the 5000 Windows reboot monkeys. Nobody expects all of this to go away if they switch to Linux. There will still be complexity. Deciding to standardize on GNOME will not make OpenOffice any less daunting to install and maintain in a multi-user environment. Or Mozilla. Or Apache.
And if we accept the argument, we would clearly choose the platform with the more robust administration interface, which clearly is KDE. kcontrol is integrated and pretty much all-encompassing, while GNOME is constantly shifting from CORBA over XML to a binary registry and back. GNOME has become so bad that they actually added a regedit style "config editor" and apparently really expect users to use it to configure applications. Hint: This is the kind of nightmare people want to get rid of when they switch from Windows to Linux.
Anyway, I don't see why we need to standardize on a GUI, and if we do, we standardize on KDE, of course, as it fulfils more of the requirements businesses have, hands down.
Uh, so citizen of other countries are not entitled to a fair trial? Is that what you are trying to say here?
Simply shooting people is not a complete violation of "everything that this nation stands for" as long as they are not US citizen?!
Sounds vaguely familiar (I'm German). The US truly are the Land of the Free and Home of the Brave. A shining beacon of liberty and freedom!
Actually, if you use "innovation" in the new-age business lingo of those hipster PR suit bearers, then it may actually be correct. These days it apparently counts as innovation if your software puts the menu choices in a different order than the previous version.
Innovation used to actually mean something a few years ago. *sigh*
So Solaris gets capabilies. Only, what, five years after Linux? And they finally abandon that obsolete slow-as-molasses file system of theirs? The level of technology leadership they are displaying is nothing short if breathtaking!1!!
On a lighter node, the article says their current partitioning scheme is software based. Good to know. Fits well in the general impression I got from them, with their shell script based "high availability" solution, and their industry leading "backup" solutions. There really is no need to know more than this about Sun and their software.
Long live admintool!
I'm running a small business.
I need to read mails from unknown people, because those are... my new customers!
How about remove Outlook and Internet Explorer instead and installing a secure email infrastructure. I have never ever, not even once, felt the need to not open an email because it might be insecure.
Advocating not opening emails is even worse than running exploitware from Microsoft in the first place.
Cryptophone.de
It's actually a division of a privately held German company called GSMK.
I am sorry that the graphs have this problem.
They look like this because I used gnuplot, which can make great graphs, without really mastering it. I had trouble selecting colors that stood apart at all. That's why the colors for each OS chance from graph to graph, by the way, because the fifth color would have to be some ugly brown, and I didn't want to prejudice against one OS by giving it an ugly color.
I can give out the data on a case by case basis, though. Please email me if you really want it. Or you could open the graphs in Gimp, Photoshop or ImageMagick or so and change the colors yourself.
Oi, you are right, I meant to write "FreeBSD" there.
I corrected the web page.
Every OS had its own partition on my notebook.
;-)
But since none of the graphs actually measures disk I/O, that does not matter in the least. You did read the web page before commenting on it, didn't you?
I did a benchmark about HTTP throughput that did not make it into the PDF or the web page: I downloaded one large ISO over Fast Ethernet. To make sure the test was fair, I mounted the same Linux ext2 file system on all the BSDs read-only. My disk has ~25 MB/sec throughput, so any OS should be able to saturate the Fast Ethernet link with ~11 MB/sec, right? OpenBSD couldn't.
But since the ext2 implementation in Linux is obviously at an advantage here, I did not publish these graphs, that would have been unfair. And I wanted to not feed the trolls.
It appears I wasn't entirely successful.
I am not pretending to produce the be-all end-all benchmark here. That would at least include having an SMP server for the operating systems that support it, and it would have used Gigabit Ethernet. Frankly, I don't have the hardware to do that.
;-)
But I do have these benchmark programs, and I was hoping to spark enough interest in doing some real comparative benchmarks by posting the results here. The tools are now there, and they do use the platform specific API hacks to get the best performance on each operating system, so the biggest area of concern for comparative benchmarks should be out of the way -- that the benchmark program is skewed in one direction or the other.
If you are in possession of a real benchmarking lab with enough free disks and spare time on your hand to complete these benchmarks, please do so!
I'm not from the Linux mafia. If my goal was to discredit BSD, I would not have added support for kqueue in the first place. If my goal was to pat myself on the back, I would not have used the BSD sysctl to increase the number of files in the system (on Linux 2.6, you basically don't have to do that, just use ulimit), and that would have been fair, if only in the Netcraft/Microsoft sense.
But neither am I from the Intel scalability labs with 1000 identical PCs and multiple gigabit ethernet connections to the server. I'm just a guy with a notebook who happened to have a little spare space on his hard disk and decided to include the BSDs in the benchmarks.
Why are you so angry?
Nowhere in the commentary is any negative conclusion drawn from the fact that the FreeBSD fork graph stops early.
There are several reasons why I didn't pursue this. First of all, I was running out of time, I finished the benchmarks the night before driving to the conference. Secondly, I didn't fly in Linus to hand tune my Linux setup as well. Third, FreeBSD does not look bad because of this, on the contrary. The graph looks O(1), it just stops early. It is true that this test should be redone now that I know how to change this setting, but honestly, I don't think the FreeBSD crowd has anything to complain about here. After all, FreeBSD brings home the trophies for the BSD camp.
Where are the numbers?
Where are the graphs?
The article basically quotes some guy (who is actually selling Samba and thus has a vested interest) saying that Samba is 2.5 times faster than Windows 2003.
Now I have no reason not to believe him, but I was expecting a little more. And I'd wager the suits considering switching to Samba also expect more.
Forget about xlib compatibility for a moment. This is not just some software wrapper implementing xlib on top of some crackpot graphics interface (while this is tedious and time consuming work, it can be done by any programmer).
This is someone fusing the microkernel and exokernel concepts, which is way more exciting than fusing some implementation stuff together. This is actually computer science research what this man is doing, not just code hacking. I wish the man all the best for his work, for new directions in operating systems research are sorely needed, and exokernels are a particularly interesting concept (the exokernel inventors at MIT showed some really incredible performance gains, but their exokernel implementation was never aspiring to be a general purpose OS like this project).
I will try to monitor this project and maybe even help a little.
Yes, the Linux init process is also slow because of the artificial serialization.
But it is also slow because of the shells, many processes and glibc. I invite you to measure the difference yourself if you don't believe me. minit will do both, by the way, start stuff in parallel, and allow you to get rid of glibc and bash in your start up scripts.