I read this article after I've received about four email messages with strange macro-virus-looking attachments in the past 24 hours (is there some epidemic again)? O sancta simplicitas!
Actually according to a result poll on/. most readers are using Windows!
But wouldn't it be more accurate to use the access logs from the Slashdot's webserver, and look at the user agent authentication string in order to determine what the browsing platform is?
Now the question is, how many web admins would give away their access logs for such purposes?
Much has been said about Linux and Unix "desktop" applications in other threads today, and this topic invariably causes lengthy flamewars, but let me tell you about my experience with typesetting software under Linux. Please note that I'm not covering image processing and drawing, which I consider separate from publishing and typesetting.
Anyway, on to the subject: for the past two years, I have been using LaTeX, which is a package of macros for the typesetting engine TeX, orginally written by Donald Knuth of Stanford University. TeX was intended to be used as a typesetting tool for scientific papers, and to this date, the quality of typesetting mathematical and other formulae with TeX remains unsurpassed (the reason probably being that the commercial niche for that type of stuff is too small to be profitable).
TeX and LaTeX have become the de facto standard in the scientific world -- they are the official typesetting tools of the American Mathematical Society, the American Astronomical Society, the American Institute of Physics, many universities and scientific magazines produce all their official publications using a flavor of TeX.
As far as I am concerned, the main advantages of TeX and LaTeX are the (unsurpassed) beauty of the output, the tremendous flexibility, and wide availability of packages. Consider the following uses I have for LaTeX:
Typesetting of simple documents such as letters, papers, and others.
Typesetting of more complicated, composite documents, such as multi-chapter books (I do a little bit of translation in my spare time).
Typesetting of documents in other scripts -- Cyrillic, 18th century Church Slavic, European languages which use the Latin alphabet.
Typesetting of software documentation, which must obey strict rules in formatting.
Typesetting of diagrams -- be it high level block diagrams of electronic circuits, or any kind of UML diagrams, LaTeX does it all for me.
To me, the only application which can perform all these tasks, using a uniform description language, is TeX. The reason TeX is so flexible, is because it uses a mark-up language (HTML is another example of a mark-up language), which describes the text layout. It also allows for creating macros and extending the basic TeX capabilites.
One of the most widely used macro packages for TeX is LaTeX, which provides templates for many standard types of documents -- letters, articles, books, etc. But you are not constrained with that -- you can create your own templates and macros, or you can use one of the multitude of packages available, which can do almost anything, even typesetting musical scores.
In TeX, you also have the ability to include and manipulate images -- you are given the opportunity to include PostScript code, so almost everything you can do with PostScript, you can do in TeX (and there are high-level PostScript macros, so you even don't have to know PostScript).
But beware -- the greatest strength of TeX is probably why graphic designers may find it inappropriate for their needs -- due to its scientific roots, the description of the document is exact -- it's very difficult to force TeX to produce funky, non-standard text layout, like you can do in Word. And most TeX packages for creating complex graphic objects are geared towards diagrams and scientific graphing, so an artist would not be able to just freely draw something and place anywhere at will -- the design of the document is very carefully considered and calculated in TeX, and it is difficult to produce ugly or non-standard documents.
In recent years, people have been developing WYSIWYG environments for TeX, so now you have TeX front ends like LyX, which allow you to edit the layout in a more "visual" manner.
For more information about TeX, one can go to the Web site of the TeX Users Group. There are plenty of good resources in the Interesting URLs section, so that should be a good start.
I didn't mean to pull it out like that but it is important to always be concise and precise with your arguments otherwise people will find holes and discount your entire argument.
I agree. I just lose it all when I start bashing Microsoft.;-)
I'm saying that free software advocates, the purists, don't give a hoot about the superiority of free software or how many people can participate in the process. It's a secondary issue for them. Even if Linux sucked by every kind of measure---user interface, speed, stability, etc.---they would gladly use it because it's not proprietary.
I may be wrong -- I can't say I have closely followed and understood all philosophies and teachings of free software and open source. But I do believe that they are closely related, and here's why: a fundamental principle underlying modern science is the free, unrestricted sharing of information and ideas, as this is believed to be beneficial to scientific development, since you have more brains thinking on the same problem at the same time, and sharing observations and thoughts.
For example, it took the ancient Greek mathematicians less than a century to discover what the Egyptians couldn't discover in half a millennium. The reason? In ancient Egypt, only the priests were allowed to study mathematics. In ancient Greece, all free citizens could study whatever they felt like.
So, this concept of free information sharing evolved in computer science, and was developed into the notion of free software. The way I see it, if a piece of free software sucks, as you say, it will be immediately fixed, because everybody has access to the source. In other words, the free software "purists" will not tolerate software of poor quality -- they'll fix it because they can.
You rejected a whole line of OS's because of problems you installing it on a particular computer with 3 other OS's alreay installed!
Well -- at one point or another I've installed the following operating systems, without losing any data: GNU/Linux, OpenBSD, NetBSD, FreeBSD, QNX, SCO OpenServer, Sun Solaris, OS/2. Somehow, they all could get it right. Windows couldn't.
I wouldn't say I reject Windows because of this single incident -- it's more like I was getting more and more aggravated, until this happened, which finally pushed me off the edge, and I decided not to run any Microsoft software on principle. If I wasn't getting frustrated with Windows, I wouldn't even attempt to install other operating systems. That's all.
Okay, so I was wrong about FAT32 -- it was a few years ago, I have forgotten. I was probably planning on changing this partition to FAT16, then.
However, when I said fdisk I meant a generic partitioning program, not any particular one -- let's not forget that there's a BSD partitioning program called fdisk, and a Linux partitioning program called fdisk (they might have common code base).
At any rate, I don't see how that invalidates my story. Have you tried installing NT 4.0 on a Packard Bell 4500 with 6 GB Seagate hard drive, and Win95, Linux and SCO OpenServer on it?
The free software purists out don't use free software because it's more user friendly or technically superior, they use it because it is free (as in liberty).
But on the other hand, the reason they insist on having that liberty, is because more people can participate in the development process, therefore the process is superior to that of non-free development, and when you have a superior process, eventually you end up with superior products.
So I do think that free software is implicitly superior in all technical aspects. The most obvious (and perhaps oldest) example is TeX/LaTeX -- as of today, there's no application that can match the wide range of abilities of TeX, simply because there are so many different packages, created by so many different people.
By the way, NT4 didn't support FAT32 (except maybe in a later service pack - never used FAT32 so I don't remember).
Why yes, it did (or at least it claimed it did). However, the installation program did not present me with a dialog, asking whether I want to use the FAT32 partition, did not start fdisk for me, or anything like that. It simply wiped out the partition table (as I discovered later), and showed some generic error message. Prior to that point, it did not allow any user interaction at all.
Firstly, I'm not talking about graphic design (the process of designing and drawing graphical elements), I'm talking about typesetting. Then, TeX automates the above mentioned text flow, so you don't have to "drag blocks of text around on screen," you just type it in, and (optionally) specify page layout dimensions. So everything is a lot less effort than toying with some useless GUIs.
Ho-ho-ho! Little you know, young one! I did try to switch to NT 4 in August of 1998. You know what happened? I had a hard drive with three partitions -- Windows FAT32, Linux ext2, SCO OpenServer whatever (Xenix fs?). The gloriously dumb Windows "installer" wiped out my entire partition table, then said it cannot continue, without giving any meaningful error description. That, my friend, is the reason why I've never installed Microsoft software on my computer since then.
If SUSE took 12 hours to install, you are doing something very seriously wrong.
He might have had some driver issues -- although it rarely takes me more than an hour to install any distribution an any PC or Mac, I once tried and failed to install Mandrake 8.0 on an old machine, which had a BusLogic SCSI adapter.
The problem was with the Linux driver for this controller; it was causing a lot of timeout errors, so writing to the hard drive was painfully slow and the install never finished. I thought that was weird, because OpenBSD installed perfectly fine on said machine.
Despite all the linux hype, it is still harder to install, maintain, and use than windows is.
Funny that, I find that I spent a lot less time maintaining Linux than I used to with Windows. Most Linux distributions come with a variety of maintenance scripts which are run automatically, so I only make minor tweaks in configuration files every once and a while, and upgrade the software packages I use every 4 - 6 months. I've been running basically the same configuration since November 1999, because Linux already has everything I need. The only major application I recently switched to is Mozilla (versus Netscape 4.7x).
Windows on the other hand, was a constant install, reinstall, defrag, tweak QEMM's memory manager to improve Win32's VM performance, and that took up to 15 hours a week -- an order of magnitude more than Linux.
This is one of the most insightful, intelligent analyses of college CS mentality, that I've ever read. I can only say: Amen, brother.
Indeed, I don't understand why students (not only CS students, mind you) are just so unwilling to think outside of the box. They're just afraid of anything new, of any change. I'm glad there are people like you who are well aware of these issues and put a lot of effort into helping their fellow students.
Yes, most of that code is Fortran, but so what?
First of all, the compiled Fortran code will be
much better optimized because there are no
annoyances such as pointer aliasing. Second,
it is easy today to compile the Fortran code
as a black box and link the library in with your
C/C++ code.
Those are perfectly valid points, and I agree that in 99% of the cases, you can either link with the Fortran code, or use F2C.
However, I recently had to figure how to get differential equation solver code to work in a real-time thread under Real-Time Linux.
I don't know whether you are familiar with RT Linux, but its essential premise is that all real time tasks are lightweight kernel threads (i.e. you have a POSIX thread interface, with some extensions for the real time stuff). That means all your real-time code runs in kernel space, and of course you don't have the standard C and C++ libraries, and all the goodies you are used to in user space.
So anyway, to make a long story short, it turned out (for me at least) to be very difficuly to make the Fortran code work in kernel space -- it's just that it depends on too many user-space libraries, and porting them is a non-trivial and probably rather lengthy process.
It may be "better" than Runge-Kutta, but the nice thing about techniques such as Runge-Kutta is that it is fast as well. I would guess a large number of the integration techniques in the ODEPACK package are designed not to be fast, but to be accurate.
Well, okay, I was just thinking of accuracy, without realizing that some people may not have a bajillion MHz at their disposal... Of course, you have a tradeoff between speed and accuracy.
You can probably try Netlib -- http://www.netlib.org.
It's a collection of mathematical software, including of course many differential equation solvers. The only problem is that a lot of these implementations are written in Fortran, so you may have to use the F2C translator.
I believe one of the most accurate numerical integration methods is the famous LSODE (Livermore Solver for Ordinary Differential Equations), an implementation of which is available at http://www.netlib.org/odepack/.
Slashdot Mirror
I read this article after I've received about four email messages with strange macro-virus-looking attachments in the past 24 hours (is there some epidemic again)? O sancta simplicitas!
Actually according to a result poll on /. most readers are using Windows!
But wouldn't it be more accurate to use the access logs from the Slashdot's webserver, and look at the user agent authentication string in order to determine what the browsing platform is?
Now the question is, how many web admins would give away their access logs for such purposes?
Anyway, on to the subject: for the past two years, I have been using LaTeX, which is a package of macros for the typesetting engine TeX, orginally written by Donald Knuth of Stanford University. TeX was intended to be used as a typesetting tool for scientific papers, and to this date, the quality of typesetting mathematical and other formulae with TeX remains unsurpassed (the reason probably being that the commercial niche for that type of stuff is too small to be profitable).
TeX and LaTeX have become the de facto standard in the scientific world -- they are the official typesetting tools of the American Mathematical Society, the American Astronomical Society, the American Institute of Physics, many universities and scientific magazines produce all their official publications using a flavor of TeX.
As far as I am concerned, the main advantages of TeX and LaTeX are the (unsurpassed) beauty of the output, the tremendous flexibility, and wide availability of packages. Consider the following uses I have for LaTeX:
To me, the only application which can perform all these tasks, using a uniform description language, is TeX. The reason TeX is so flexible, is because it uses a mark-up language (HTML is another example of a mark-up language), which describes the text layout. It also allows for creating macros and extending the basic TeX capabilites.
One of the most widely used macro packages for TeX is LaTeX, which provides templates for many standard types of documents -- letters, articles, books, etc. But you are not constrained with that -- you can create your own templates and macros, or you can use one of the multitude of packages available, which can do almost anything, even typesetting musical scores.
In TeX, you also have the ability to include and manipulate images -- you are given the opportunity to include PostScript code, so almost everything you can do with PostScript, you can do in TeX (and there are high-level PostScript macros, so you even don't have to know PostScript).
But beware -- the greatest strength of TeX is probably why graphic designers may find it inappropriate for their needs -- due to its scientific roots, the description of the document is exact -- it's very difficult to force TeX to produce funky, non-standard text layout, like you can do in Word. And most TeX packages for creating complex graphic objects are geared towards diagrams and scientific graphing, so an artist would not be able to just freely draw something and place anywhere at will -- the design of the document is very carefully considered and calculated in TeX, and it is difficult to produce ugly or non-standard documents.
In recent years, people have been developing WYSIWYG environments for TeX, so now you have TeX front ends like LyX, which allow you to edit the layout in a more "visual" manner.
For more information about TeX, one can go to the Web site of the TeX Users Group. There are plenty of good resources in the Interesting URLs section, so that should be a good start.
Of course, I know about Stallman and the GPL. My notion of the quality of free software is probably similar to this.
I didn't mean to pull it out like that but it is important to always be concise and precise with your arguments otherwise people will find holes and discount your entire argument.
;-)
I agree. I just lose it all when I start bashing Microsoft.
I'm saying that free software advocates, the purists, don't give a hoot about the superiority of free software or how many people can participate in the process. It's a secondary issue for them. Even if Linux sucked by every kind of measure---user interface, speed, stability, etc.---they would gladly use it because it's not proprietary.
I may be wrong -- I can't say I have closely followed and understood all philosophies and teachings of free software and open source. But I do believe that they are closely related, and here's why: a fundamental principle underlying modern science is the free, unrestricted sharing of information and ideas, as this is believed to be beneficial to scientific development, since you have more brains thinking on the same problem at the same time, and sharing observations and thoughts.
For example, it took the ancient Greek mathematicians less than a century to discover what the Egyptians couldn't discover in half a millennium. The reason? In ancient Egypt, only the priests were allowed to study mathematics. In ancient Greece, all free citizens could study whatever they felt like.
So, this concept of free information sharing evolved in computer science, and was developed into the notion of free software. The way I see it, if a piece of free software sucks, as you say, it will be immediately fixed, because everybody has access to the source. In other words, the free software "purists" will not tolerate software of poor quality -- they'll fix it because they can.
You rejected a whole line of OS's because of problems you installing it on a particular computer with 3 other OS's alreay installed!
Well -- at one point or another I've installed the following operating systems, without losing any data: GNU/Linux, OpenBSD, NetBSD, FreeBSD, QNX, SCO OpenServer, Sun Solaris, OS/2. Somehow, they all could get it right. Windows couldn't.
I wouldn't say I reject Windows because of this single incident -- it's more like I was getting more and more aggravated, until this happened, which finally pushed me off the edge, and I decided not to run any Microsoft software on principle. If I wasn't getting frustrated with Windows, I wouldn't even attempt to install other operating systems. That's all.
Okay, so I was wrong about FAT32 -- it was a few years ago, I have forgotten. I was probably planning on changing this partition to FAT16, then.
However, when I said fdisk I meant a generic partitioning program, not any particular one -- let's not forget that there's a BSD partitioning program called fdisk, and a Linux partitioning program called fdisk (they might have common code base).
At any rate, I don't see how that invalidates my story. Have you tried installing NT 4.0 on a Packard Bell 4500 with 6 GB Seagate hard drive, and Win95, Linux and SCO OpenServer on it?
The free software purists out don't use free software because it's more user friendly or technically superior, they use it because it is free (as in liberty).
But on the other hand, the reason they insist on having that liberty, is because more people can participate in the development process, therefore the process is superior to that of non-free development, and when you have a superior process, eventually you end up with superior products.
So I do think that free software is implicitly superior in all technical aspects. The most obvious (and perhaps oldest) example is TeX/LaTeX -- as of today, there's no application that can match the wide range of abilities of TeX, simply because there are so many different packages, created by so many different people.
By the way, NT4 didn't support FAT32 (except maybe in a later service pack - never used FAT32 so I don't remember).
Why yes, it did (or at least it claimed it did). However, the installation program did not present me with a dialog, asking whether I want to use the FAT32 partition, did not start fdisk for me, or anything like that. It simply wiped out the partition table (as I discovered later), and showed some generic error message. Prior to that point, it did not allow any user interaction at all.
Firstly, I'm not talking about graphic design (the process of designing and drawing graphical elements), I'm talking about typesetting. Then, TeX automates the above mentioned text flow, so you don't have to "drag blocks of text around on screen," you just type it in, and (optionally) specify page layout dimensions. So everything is a lot less effort than toying with some useless GUIs.
Why is this a troll? Because I disagree with the previous poster?
Wouldn't that be akin to saying, "Bin Laden is not our enemy. Hitler was not our enemy. Hatred and violence are our enemies."
And exactly why is this "Overrated?"
You should have switch to NT4 back in the day.
Ho-ho-ho! Little you know, young one! I did try to switch to NT 4 in August of 1998. You know what happened? I had a hard drive with three partitions -- Windows FAT32, Linux ext2, SCO OpenServer whatever (Xenix fs?). The gloriously dumb Windows "installer" wiped out my entire partition table, then said it cannot continue, without giving any meaningful error description. That, my friend, is the reason why I've never installed Microsoft software on my computer since then.
Find me three business executives who use nroff to format documents on a regular basis.
Find me one professional typesetter who uses Microsoft publishing software.
Find me one application that can do everything TeX/LaTeX can.
If SUSE took 12 hours to install, you are doing something very seriously wrong.
He might have had some driver issues -- although it rarely takes me more than an hour to install any distribution an any PC or Mac, I once tried and failed to install Mandrake 8.0 on an old machine, which had a BusLogic SCSI adapter.
The problem was with the Linux driver for this controller; it was causing a lot of timeout errors, so writing to the hard drive was painfully slow and the install never finished. I thought that was weird, because OpenBSD installed perfectly fine on said machine.
Despite all the linux hype, it is still harder to install, maintain, and use than windows is.
Funny that, I find that I spent a lot less time maintaining Linux than I used to with Windows. Most Linux distributions come with a variety of maintenance scripts which are run automatically, so I only make minor tweaks in configuration files every once and a while, and upgrade the software packages I use every 4 - 6 months. I've been running basically the same configuration since November 1999, because Linux already has everything I need. The only major application I recently switched to is Mozilla (versus Netscape 4.7x).
Windows on the other hand, was a constant install, reinstall, defrag, tweak QEMM's memory manager to improve Win32's VM performance, and that took up to 15 hours a week -- an order of magnitude more than Linux.
Wouldn't that be akin to saying, "Bin Laden is not our enemy. Hitler was not our enemy. Hatred and violence are our enemies."
This is one of the most insightful, intelligent analyses of college CS mentality, that I've ever read. I can only say: Amen, brother.
Indeed, I don't understand why students (not only CS students, mind you) are just so unwilling to think outside of the box. They're just afraid of anything new, of any change. I'm glad there are people like you who are well aware of these issues and put a lot of effort into helping their fellow students.
Yes, most of that code is Fortran, but so what? First of all, the compiled Fortran code will be much better optimized because there are no annoyances such as pointer aliasing. Second, it is easy today to compile the Fortran code as a black box and link the library in with your C/C++ code.
Those are perfectly valid points, and I agree that in 99% of the cases, you can either link with the Fortran code, or use F2C.
However, I recently had to figure how to get differential equation solver code to work in a real-time thread under Real-Time Linux.
I don't know whether you are familiar with RT Linux, but its essential premise is that all real time tasks are lightweight kernel threads (i.e. you have a POSIX thread interface, with some extensions for the real time stuff). That means all your real-time code runs in kernel space, and of course you don't have the standard C and C++ libraries, and all the goodies you are used to in user space.
So anyway, to make a long story short, it turned out (for me at least) to be very difficuly to make the Fortran code work in kernel space -- it's just that it depends on too many user-space libraries, and porting them is a non-trivial and probably rather lengthy process.
It may be "better" than Runge-Kutta, but the nice thing about techniques such as Runge-Kutta is that it is fast as well. I would guess a large number of the integration techniques in the ODEPACK package are designed not to be fast, but to be accurate.
Well, okay, I was just thinking of accuracy, without realizing that some people may not have a bajillion MHz at their disposal... Of course, you have a tradeoff between speed and accuracy.
So how long before a squadron of B-52 Stratofortress long range bombers is dispatched from Minot AFB to a certain location in the state of Washington?
I remember reading an article with a game developer a while back, who pointed out that the key for physics in a game wasn't realism, but consistency.
Exactly. Your physical constants may be different than those on Earth, but the principles, the laws of physics remain the same.
You can probably try Netlib -- http://www.netlib.org. It's a collection of mathematical software, including of course many differential equation solvers. The only problem is that a lot of these implementations are written in Fortran, so you may have to use the F2C translator.
I believe one of the most accurate numerical integration methods is the famous LSODE (Livermore Solver for Ordinary Differential Equations), an implementation of which is available at http://www.netlib.org/odepack/.
Another one is VODE, which is based on LSODE. http://citeseer.nj.nec.com/1230.html. I think both are more accurate than Runge-Kutta.