I just read the LJ article on the performance comparison between Linux and Windows 95/98/NT. In the article, gcc and pgcc on Linux were compared to VC and CodeWarrior on Windows. The results showed that gcc and pgcc were faster, and the article made the claim that this means Linux is faster.
I'm curious—would Linux really be faster if gcc were run on Windows?
How does the program perform when compiled with DJGPP, Mingw32, or Cygwin? Also, how would lccwin32 (the lightweight c compiler) perform?
—Paul Brannan
Thanks for your note.
You are definitely correct that performance is a function of many things—the compiler, the memory, the motherboard, the CPU, the operating system, etc. I think we were able to isolate performance to primarily the OS and the compiler in our study (although I think we also were able to do some comparison of CPUs in the study). We've tried to emphasize in the article that it's the Linux/gcc or Linux/pgcc combination that is faster than the Win/VC or Win/CW combination.
Here's one datapoint which we didn't present in the article. We have benchmarked the application on some Sun Solaris machines which had gcc and Sun's C compiler. On the Suns, the Sun C compiler produced binaries that were about 30-40% faster than gcc.
Also, gcc—on Sun/Solaris and on Intel(or AMD)/Linux—takes very little advantage of advanced architectural features. For example, gcc typically does not generate MMX instructions, and it never generates 3DNow or SSE instructions. On Sun, gcc does little to take advantage of Sun's Visual Instruction Set. Under Windows, Code Warrior CAN issue MMX and 3DNow instructions, and Visual C can issue MMX instructions. Moreover, both Code Warrior and Visual C have customization settings to take advantage of Pentium and Pentium II (and now, probably Pentium III) architectural features. Code Warrior can also generate code that takes advantage of K6-2 architectural features (features besides only 3DNow!). Therefore, we believe that it is unlikely that gcc is generating code which is closely targeted to a specific architecture—at least, the code would not be expected to be targeted as well as Code Warrior and Visual C since those two can optimize for the specific architecture. Therefore, we believe that Linux is indeed likely to offer a more efficient execution environment than Windows.
I'd also like to note that we have considered comparing assembly code generated by gcc and the commercial products, but it's been suggested to us that the terms of licensure for most of the commercial compilers imply that “reverse-engineering” activities are not acceptable uses of the licenses. We felt that comparing assembly code could be viewed as a reverse engineering activity, and so we have made no attempt to compare the assembly code being generated by the compilers. Instead, we've done what we could do, which is just focus on the bottom line and observe which complete environment produced the best execution performance.
However, again let me state that you are correct that it's not completely possible to decouple the compiler and the OS. That was a point that we tried to make in the article, although possibly we could have made even more explicit and strong statements on that issue in the article. That's one reason that we tried to point out that at least the gcc/LInux (or pgcc/Linux) _combination_ was more efficient than VC/Win or CW/Win. We find that to be fairly remarkable - a completely freeware solution can actually operate more efficiently than a commercial solution (especially given the enormous amount of effort on the commercial products by some very talented individuals) and we thought that the computing community needed to be presented with our study's results.
One last word. We did not test this particular application using gcc under Windows. However, in the past we have compiled other (smaller) applications using gcc under Windows. We have compared those applications' performance with the performance of code compiled with commercial compilers and observed that the commercial compilers produced faster code. Our experiments were not rigorous, however, and were on a limited set of examples. Due to those experiments, we thought that it was unlikely for our application to run as efficiently if it were compiled using gcc under Windows than if we used something like Code Warrior under Windows, which we know produces good-performing code. Also, I believe that our version of DOS gcc is a 16 bit version, and we thought it would be unlikely to be able to produce reasonable execution times for our application.
But, I like your suggestion of trying additional compilers, which could make for a strong case that Linux's computational performance is superior to Windows. I will try to benchmark the application using gcc and a few other Windows world
compilers. I'll get back with you with results. If you don't hear from me within a week, could you drop me an email. I'm pretty busy and sometimes I put things off for a while and then don't get back to them for some time. A reminder would help me
get around to it quicker.
—Dr. Tim Newman, tnewman@cs.uah.edu
Two observations:
1. PPP:
I suspect that the point of Gerry George's question was missed in the answer to his PPP question. For purely local access, one can assign a private IP address to the interface by default. A private IP address is one selected from the ranges listed
in RFC1918 and guaranteed to be unroutable on the public Internet, such as 192.168.x.x.
If security is not a huge concern and users can generally be trusted once they provide their name and password to satisfy the ordinary login process, then one can insert “ipcp-accept-remote” into the “options” file which will be read by “pppd”, usually in “/etc/ppp”. This tells “pppd” to accept whatever the remote peer wants to use as its IP address.
If the answerer wishes to force the caller to supply IP addresses for the answerer's interface, then it should set its own address to 0.0.0.0. Note that some PPP software, such as Windows Dial-Up Networking, may be unable to cope with such an answerer.
One should also include the “lock” keyword in the “options” file so that a dialout does not occur while someone is dialed in. Also, both the dialout script and “getty” should reference the same interfaces using the same names, such as “/dev/ttyS0”. The use of the “/dev/cua0” form of interface name should be avoided for this purpose.
2. Network Time Synchronization:
In response to John Morley's question, it is important to emphasize that there are different types of network synchronization. First, in some cases, it is more important that colocated machines be precisely locked together than that they be
accurate relative to some external reference. Many protocols, such as NFS and Kerberos, benefit from tight local time synchronization among cooperating machines.
Since real NTP is relatively straightforward to configure and totally free, there are really no good reasons to rely on “rdate” or “timed” methods. The issue is not so much absolute accuracy as fault tolerance. All programs which I know about that use the RFC868 TIME (port 37) protocol simply query one server and believe whatever they are told. Personally, I have seen widely used public time servers at major universities give out junk answers on occasion; I once had a machine set its clock to February 2037 when such as server began sending all zeroes.
With NTP, attempts are made to measure network delays and apply some degree of sanity checking to the results returned. In addition, if a web of NTP servers are configured together, then they will tend to stay closely locked together around a consensus mean time. NTP also allows using MD5 and other cryptographic authentication so that only the machines configured together will be trusted to nudge each other's time.
There are excellent freeware Windows clients which implement SNTP, a simplified client-only version designed to be used to query a real NTP server. I recommend Dimension4, a freeware SNTP client for Windows 9x/NT, available from Thinking Man Software:
http://www.thinkman.com/~thinkman/dimension4/
For extremely casual Unix use, such as might be appropriate for a desktop workstation, the “ntpdate” SNTP client program included with the NTP distribution is a much better option than “rdate” or “netdate”
methods.
—Mike Bilow, mikebw@colossus.bilow.com
Anonymous replied to your post 'my editorial in LJ' at the site: Linux Journal. The reply was:
Well, for one thing your editorial made me glad I have decided not to renew my LinuxJournal subscription. There are several very important reasons why GNOME should be the standard linux desktop, and most of the have nothing to do with GNOME itself. a) The licensing issues. I know some people have tried to call this one of after the QPL, but it is definetly not resolved. Libraries should be LGPL anything more restrictive is a mistake. The toolkit put under the LGPL is important because it, for all practical resons, places the ownership of the toolkit in the hands of the community which means that all who helps develop from can potentially benefit equally. Another aspect of this is that Linux success is and will always be about the OS beeing a free platform for all, even for people who in my opinion makes the wrong choice and release the software under proprietary licenses (which most of these “open-source” licenses qualify as.) John “Maddog” Hall, the leader of LI, made a statement in a interview some time back when asked if he thought all software should be free, where he answered that the most important thing is that the OS itself is free. If Qt was to be accepted as the primary toolkit this would no longer be the possible. (Note, I don't know wether Maddog shares this interpretation.)
This is the single most important reason to reject Qt, and therefore unfortuantly also KDE.
b) The language choice. The fact that Qt and KDE is so C++ based, makes it very difficult to make wrappers for other languages. Which would lessen the attraction to Linux for many developers. AFAIK the only wrapper currently available is Perl althought there might be a few others. Where Gtk and Gnome on the other hand uses stanard C and therefore already have gotten wrappers for C++, Objective C, Python, Pearl, Ada and many more. Even more wrappers are under development to support languages such as Java and Pascal. The use of C as the base language should be of little consequense for C++ developers with great C++ and Gtk based projects like Mozilla and AbiWord to show the way.
c) The last reason I will give at this point is simply that GNOME is to great to be discontinued, even though the project was started around a year later than KDE is has already caught up, and in many areas surpassed KDE. The last few areas, such as some rough edges caused by product immaturity, is being taken care of with the 1.0.50 release being readied these days.
Okay an extra bonus reason: GNOME is the official GNU desktop and without GNU there would be no Linux. Sincerely,
—Christian Schaller, frostking@linuxrising.com
To: Jason Kroll
Keep up the great work in Linux Journal. It was because of your cool articles that I bought and subscribed to the mag!!! Sincerely,
—Richard Tricoche, richard.tricoche@rpawireless.com
John C. Burgess wrote, in Letters to the Editor, in regards to being able to format floppies as a non-root user:
chmod 777 /dev chmod o+w /dev/fd* chmod 777 /usr/bin/fdformat rebootAnyone following this advice can compromise the security of their system. An explanation follows.
First, there is usually a group called floppy in the group file. the /dev/fd* files are also usually members of this group. A user can be added to this group (usually in /etc/group,) and they will have access to the files that groups owns the next time they log in. This means they will have write access to the floppy block devices the next time they log in.
So instead of giving global access to the floppy devices, add specific users to the group which has write access to the floppy devices. Giving global access to the floppy device allows anyone logged in to change any of the files or blocks on the floppy at whim.
Don't give global write access to the fdformat program (anyone who logs in can change the nature of the program and insert bootloader viruses on the floppies! - a rather hysterical example, that is.) I.e. don't ever type chmod 777 on any system binaries, libraries, configurations or any other files.
As for rebooting, your login shell will have to be restarted, but nothing else should need to be restarted, esp. the kernel. I typically add a user to the group file, and then I have to restart the login shell before the new permissions are recognized.
Another piece of errata: I've heard most recently people referring to root access as `from any directory except root.' This is a misconception. The root directory, /, is the first mount point in the Linux system, and is the start of the filesystem path (hence the chroot command.) Root access, as the user, is merely just another login, only the root user has access to every aspect of the system, including the running kernel! Don't use root!
For John C. Burgess: whatever `gurus' you asked this about were flat wrong, but the solution you came up with, while testimonial of the flexibility and power of Linux, was probably the worst solution you could ever force on yourself in the long
term.
—Christopher Rhodes, christohper@evergreen.net
In your reply to “Memory Error, Parallel Computing” in the October issue, one possibility for the first question was overlooked. Some systems (like Compaq's 9546) limit the amount of memory reported to the kernel to slightly less than 16M. A simple fix is to use the boot option:
mem='XXM'
where XX is the amount of memory installed in the system.
—Kevin, kevin@obone.net
I just downloaded and installed Word Perfect 8 for Linux on a RedHat 6.0 system. This product is not ready for prime-time.
The download and installation instructions were, on some points, vague, and even misdirecting. This is not good.
More importantly, all application graphics show as garbled checkerboards, making the Word Perfect 8 application unusable. Even more importantly, according to tech support, this is the behavior of the commercial product, and they say they have no idea what the cause is or how to fix the problem.
I was hoping to find a familiar word processing application for my Linux oriented clients, but I certainly can't recommend WP8.
I'm sorry your company is presenting applications for the Linux operating system in this light, since I experience a superior level of reliability with most Linux applications. You can see what this says about Corel, of course.
—James Feeney, james@topaz.nurealm.net
Well, the San Francisco School of Design has a great reputation in the business, as does the Rhode Island School of Design. Other than that, most universities will have some kind of design program, and there are smaller specialty schools. We have a junior designer here in the office who just graduated from the Art Institute of Seattle.
Your planned majors sound good, you'll have a broad base there. The business classes will help if you decide to work for yourself.
My education includes 2 years of Design from the now-defunct Factory of Visual Arts in Seattle. I can't say I learned anything relevant to doing magazine layout there, other than general design concepts. My computer design and layout skills are entirely self-taught, and I'm learning all the time. I *can* say that you should plan to become an expert at Adobe PhotoShop, Illustrator and InDesign, as well as at Quark Xpress. Beyond that, some CAD or 3-D Rendering software would be really helpful. Hopefully some or all of these applications will be available for Linux soon.
And, most importantly: a lot of design schools teach nothing or next to nothing about real-world design for print. Learn how to scan art, set up graphics for print, determine the correct file format to use, trap, and do color separations. You'll make mistakes as you gain professional experience, but try to avoid the “oops, that won't print” or the “oops, I don't know how to plan for output resolution” mistakes that happen to new designers. Also, learn to make your printer your friend (not your Epson or HP, rather the press man at the print house). You can learn more from a good pressman than from almost anywhere else, and a lot of designers don't realize how much time and money can be saved by talking to the pressman before beginning a design.
Well, that's it. Good luck!
—Lydia Kinata, info@linuxjournal.com
I read the question posted by Mark Foucht in regards to mounting CDs in your October issue. The answers given in regards to how to share the mounted images were correct, but one detail that was left out was HOW to get those images so that they could be shared. I wasn't sure if Mr. Foucht's question was more about how to actually make/mount the images, or how to share them.
To make/mount the cd images:
mkisofs -r -o cdrom0.iso /mnt/cdromwill create the file cdrom0.iso with Rock-Ridge extentions in the current directory from the filesystem mounted at /mnt/cdrom.
losetup /dev/loop0 cdrom0.iso mount -t iso9660 /dev/loop0 /mnt/cdimage0
umount /mnt/cdimage0 losetup -d /dev/loop0
I hope this, along with the answers supplied will give Mr. Foucht a more complete answer to his question.
—Steve Fuller, sfuller@fuller.des-moines.ia.us
hello,
there are two corrections to the picture on page 14 of your october magazine.
—engelbert gruber, engelbert.gruber@ssg.co.at
Hello,
I just read your article on LJ issue 66, and I would like to give my opinion on it. I realize you will get plenty of “opinions” about it, so I'll try to be short, even if it's a complex matter.
The first thing I thought, trying a KDE beta some time ago, was “oh! Finally a decent window manager!” but I'm sure this can be only a so personal point of view, maybe because I've been using Windows more than X-Windows; after all this can be the thought that so many ones have had. One of the major killers to Linux diffusion, in my humble opinion, is after all it's appeal: you can do wonderful things with it, but users will have more troubles configuring /etc/printcap than a GUI equivalent. The point you raise, about not skilled users to think they gained “computer literacy” is after all sadly true. I found so many small ISP, and sometimes even small informatics departments in schools, having a SysAdmin who didn't even protect TCP ports like fingerd to outer world, giving away full users' names. From my point of view these guys are Windows NT administrators that just discovered Linux, and when they succeeded with their first internet connection started thinking they could “master the internet”. But this example addresses something that's not mentioned in your article: the level of literacy required by a computer user.
Your words appear more a crusade toward the pureness of computing, than pointing to the users' lack of skill. I've heard and read similar things before, and it's so sad to see that what's presented as an “open movement” after all wants just to be a close community of experts: unless you are dreaming of a world where everyone knows the most about operating systems.
And that close minding on open software happens, ironically, even in teams like the KDE one, let me tell a short story. When the KDE was out as a Debian package I installed it, and found that the FHS adopted was giving major troubles trying to configure and compile older software, even if there was a 'configure' script made with autoconf. I contacted Stephan Kulow, the maintainer of the Debian packages, and he (kindly?) answered that he wasn't not susrprised an Italian like me didn't understand about autoconf, and he declared the argumentation over.
So I consider such comments as a lack of reality, because the daily users of a computer, like people doing just word processing, or charts and invoices, can't be expert users of an operating system. And they will never be, considering how fast software products are evolving. Then of my questions is: are we cutting out 90% of computer users just to feel the ones who hold the real power of computing? If this is the purpose of the open source movement then I don't even want to be in it, no thanks. I don't want to be using an operating system that discriminates what users are good or bad ones: wasn't that a prerogative of the already estabilished commercial OSes?
I keep seeing software produced by wonderful programmers that has a hard to use interface, hard to configure, and in the end justified as “only for real computer people”. This is crazy, keeping things hard to use only for a small number of adepts. There's nothing you can call “open” in all this.
I'm developing embedded systems for more than ten years, designing them from hardware up to high-level system interface. With all this I mostly worked with C language, and a lot of assembly programming on so many microprocessors, and that lead me to keep liking more C to C++, because on small systems software has to be predictable with memory usage and real time response. Despite all this, another thing that I thought looking at how KDE is programmed made me think “Hey, this is how Windows should have been made”.
If you want to make a new world (since we already have one running), you have to get your hands dirty with something, and you can't ignore it.
I really don't understand what is making “internet safe again”: using KDE leads to making internet unsafe? Maybe it's because I'm european, so I'm not as close minded and bigot as many americans I see, after all we had a different history, but... I'm really not the kind to go out with a sign saying that the end of the world is soon to come, or asking for a holy war.
Ok, armed with console applications you can do more than with a GUI, maybe even I can do a little magic with it, but why the heck has _any_ Linux user to do so? Should it be one of those proof of adultness to achieve like in a primitive culture?
You address a real problem with things like paper size treated to be A4 instead of 'letter', that's something for developers to fix. I know very well that kind of problem: I've been fixing for many years software that was using 'letter' instead of A4, and inches instead of centimeters. So this just shows that can exist blind programming on both sides of the pond.
So you don't see any progress in '99 programming compared to '76 one? I'm not surprised. Can't see any way you could.
One final note. Linux Journal has now a better appeal, with more colored pages, better organized pages and sections: doesn't it make readers think they read it better, while this is totally wrong? I can't see how a publisher did
let this happen, as we know that in the end leads guys to think they know how to manage a magazine (after they badly drove their car all the way to come telling you, of course). Luckily the magazine has rare spots of this catastrophism in it,
otherwise I wouldn't have a good reason to keep being a subscriber. Open your mind before you open your software source, or you will give for free something that no one needs.
—Lou Cyphre, LouCypher@cheerful.com
I saw your comment about floppy formatting in the October issue of LJ.
In fact, the following is what I did (uh, back in 1995):
Regards,
—Ambrose Li, acli@acli.interlog.com
In your October 1999 editorial, you mention that KDE being configured for A4 paper size by default as a problem, even going as far as considering it a bug (though you _do_ point out it can be recompiled otherwise).
May I recommend that you do what's right and upgrade towards this world-wide standard format, instead of perpetuating that old, deprecated remnant of American industrialism know as US Letter?
The DIN paper formats may have started in Germany, but they have now become a world-wide standard, not only in Europe but also in Africa, Asia (except Japan) and South-America.
The standard itself is rather clear and logical and it includes just about every kind of stationary out there: paper, envelopes, business cards, invitation cards, passports, ID cards, etc. etc.
Like the metric system, the DIN stationary format is one of those standards America really ought to upgrade to. The question is not why but when, as both standards have become unavoidable.
Could Linux users lead the way, by implementing the DIN formats and adopting them from now on, even in America?
—Martin-Eric Racine, q-funk@intellitel.com
Dear Mr Hughes,
I am probably about the 999th person to protest to you about your editorial “Is KDE the Answer?” in the most recent LJ. But I hope I can get your ear nonetheless.
You start the editorial with two questions: “Is looking like MS Windows good or bad?” and “Is KDE a better answer than GNOME?”
You spend half the editorial arriving at an answer to the first question with which I agree: “It depends.” For certainly some people, fresh in out of the wilderness, will like to have a Windows-like interface, for a while anyway. Whereas others will want to put childhood behind them. Fortunately everyone can have what suits them.
But of course it is the second question that is bothersome. It reminds me of a similar question: “Is emacs better than vi?” Haven't we all learned that this kind of question is silly and irritating?
Don't you value competition? don't you see the benefits that ensue? to name just one example: would Troll Technologies have modified their license for qt if they, and the KDE community, hadn't felt GNOME's breath at their backs?
To try to pick a “winner” between KDE and GNOME is like trying to pick a winner between emacs and vi. It is IMHO silly and, worse, unproductive. Let's have (at least) two wonderful editors, and let's have (at least) two wonderful windowing environments. That is the way for Linux and Open Source to make progress. That is what will benefit us consumers the most.
For the record, I am a charter subscriber to LJ, which I love, and: my window manager of choice is fvwm.
Best wishes,
—Alan McConnell, alan17@wizard.net
To: Phil Hughes
I just wante to write and let you know how much I agree with your review of Learning Python. I haven't actually read the book however, I just tried to learn python by it's predecessor. I'm glad you found the new book enjoyable and useful.
I started trying to learn Python from 'Programming Python' and quickly got bogged down. That was last Christmas. Recently, I picked the idea of python again, and came to a miraculously useful conclusion:
The OOP stuff doesn't matter until you want it.
When I started trying to write scripts as ordinary, straight through procedural stuff, just like my perl and ksh scripts, it just popped straight out! I was converting my perl scripts over and they became instantly more readable! I am now far more impressed with Python than I ever was reading the 'PP' book, even though the stuff in there is very interesting. Combined with Python's large library of Very Useful Stuff, you can just get on with writing useful scripts, which I suspect is exactly Guido vanRossum had in mind.
Of course, once you've spent a long time with it, you start to get a better feel for the OOP stuff. Right now, I'm just going over the 'PP' book again, and it's making more sense this time around. But it's not the best of introductions to a very lovely and gentle language.
Anyhow, I'm glad to see somebody else enjoying Python for what it is.
Many thanks for LJ; I've been subscribing since issue 9 and look forward to every issue.
—Dominic Mitchell, dom@myrddin.demon.co.uk
I'm not going to strain Phil's flame-proof suit :-)
Still, I would like to offer some thougths to his article in the October issue of Linux Journal. Well—is KDE the answer? Yes, it is. Is GNOME the answer?
sure. Is Python the answer? Of course! Is Perl the answer? How could you think otherwise. GCC? Egcs?. You name it.
I don't tend to see the diversity of solutions and approaches in the free software/open source world as a liability, but as an asset. A rather important one. In the metaphor of evolution (sometimes an overstrained metaphor, I admit), this diversity is the big genetic pool which ensures adaptability to changing environments. And environments do change in this field.
Besides, in this FS/OSS world, much of the development is fun-driven. The idea is that you are much more productive if you really enjoy what you are doing. There are those (I'm one of them) who shriek in horror at the idea of programming in C++, don't like particularly Python and are comfortable with the object models of Perl or Lisp. I'd rather go through the contortions of doing “more-or-less-kinda-OO” programming in C than “real-OO” programming in C++ for example. There are those who cry in disgust at this very perspective. A rich ground of tools and approaches will do much more to harness the energies of all of us.
I do hope that things get constantly reinvented to some extent. You can't afford to do radical changes to ext2, for example. It has to be a rock-solid basis for many production environments out there. But perspectives would be very dull without ext3 or reiserfs, for example.
The most important thing is to work on interoperability, and the Gnome and KDE folks seem to take this seriously (the last time I looked into their mailing lists, that is). It should be made as easy as possible to write applications which fit
as well as possible into *both* environments. Then we get something the closed-software world hasn't: cross-fertilization. This is a concept which has been very successful in natural evolution—it might work for us as well. The easier
it
is to “steal” software from each other the better.
So to offer my personal answer to your question: “Yes, definitely, all of them”. P.S. —again I enjoyed your magazine from cover to cover.
Thank you for the good work.
—tomas, tomas@aura.de (tomas zerolo)
I am 46 years old and stopped programming when Big Blue came along. I used to work with the then available 8 - bit machines and hexadecimal coding. Now I decided to jump back into the programming area and started with the installation of RedHat 6.0.
I must say it is like being reincarnated for the world seems to be dominated by C++ and Internet. While instaling RedHat I noticed that one should have benefits of an education in LAN or some kind of Networks.
It also seems to me that the package will sell but that a lot of buyers will knock off. The road to completion is hazy for one will never be at ease and done. I live in the Netherlands, my e-mail adres is
groetjes,
—Robert Vencent Racam, evergladenld@planet.nl
Two articles in October's issue of the Linux Journal concerned me a little. First of all, I have to find issue with the rather unfair article that Mr.Hughes' wrote in regards to the KDE/GNOME issue. First of all, QT is still not free. Nope. It's semi-free, it's sorta-free, but it remains a restrictive license, and the proliferation of these semi-free licences is potentially quite damaging.
We all appreciate the efforts Troll has made in order to loosen the license, but having the entire Linux desktop controlled by a for profit company is no different then Windows. If GNOME was the standard, we'd be able to have a wealth of applications, closed and open. However, with KDE the standard, commercial vendors will be less inclined to port if they have to fork over a large sum of money.
Mr.Hughes has condemned the GNOME project even though it has had far less time to reach stability. Recent releases, like the 1.0.53 builds have shown remarkable progress and I use it daily, under fairly strenuous conditions. By his thinking, we should go with whatever works, even if it doesn't satisfy us. If Linux believe that, we'd all still be running Windows, and the guys at Red Hat would have far less money.
While it's not crucial that all the applications are free, it it crucial that the underlying system be free. QT is not truly free, while GTK and GNOME are.
On another note, I'd like to mention to Mr. Kroll that Code Warrior works well on other distributions. I'm a Stampede developer, and I use Code Warrior frequently. It was a simple matter of running alien, and I had a working copy of Code
Warrior.
—Aubin Paul, outlyer@stampede.org
I think somebody should do a good article/tutorial on how to disable the many network services that users often don't need. I recently learned how to do this and by taking out nfs, mountd, sendmail, portmap, and probably others I can't recall, I have made my computer faster and more secure. My bootups are noticeably faster, the memory requirements of my system have been dropped by several Mb and I don't have to worry about security holes in all those daemons I don't understand.
I suspect that many people who, like me, use their linux boxes as workstations and not servers could benefit from an article telling them that this is OK to do. For people in my kind of situation, all that is really needed in terms of network servers is ftp and telnet. I've known for a long time that this is all I wanted to do, but I wasn't sure which daemons I could safely disable or the best way to do it, for that matter.
Just an idea,
—Colin Durocher, conan2@home.com
Ms. Richardson, Mr. Searls,
I have been a subscriber to Linux Journal for a few months now, and a reader for a long time. I have some general comments to make with regards to the direction of the magazine that I hope you will consider. Let me say that I am only forwarding these suggestions to you because I would like to think I can hold your magazine to a higher standard than other linux magainzes that have jumped on the bandwagon.
Firstly, I would appreciate it if the magazine moved away from such a position of 'hollow' advocacy. Obviously I don't mean you should stop promoting linux - its your business, but it seems that the magazine is always treating me as an NT user that isn't convinced of the usefulness of linux. This theme is so pervasive in your magazine that it even gives way to rather questionable quotes used to bolster your advocacy:
“...Since I have to work with NT for political reasons, I just cope with it. But I know if we could see the source, we could probably fix the problem pretty fast.” Windows NT is millions of lines long. Even if it was open source and well documented, it is *highly* unlikely that a single user, not deeply experienced with the architecture of the OS, could *ever* locate and repair the errors in question. Whether open sourcing NT would have prevented such bugs from appearing in the first place is another issue, but suffice to say, all of the open source operating systems have longstanding bugs, code availability nothwithstanding!
Secondly, I would like to see more meat in the interviews. This month's issue featured an interview with Linus Torvalds - a great opportunity to get the inside scoop on future developments, but instead we get a barrage of incredibly uninformative, almost insipid questions about his personal life. I know this last statement is rather rude sounding, but I cannot express to you how disappointed I was with the lack of content in this interview, which I was really looking forward to.
Thirdly, I would like to see more criticism of products and software. Some open source software is good. Some is terrible. I think users need to know about both kinds. As for hardware, I was disappointed that this month's review of VAResearch computers failed to mention cost as a criterion of purchase. VA boxes are historically overpriced by a significant margin, and your readers deserve to know that nearly identically equipped PCs can be had for much cheaper. I understand they advertise heavily in your magazine, but that should not preclude an honest appraisal of their products.
Other than that, “Best of User Support”, “New Products” and “At the Forge” continue to be excellent columns. I hope Mr. Lerner continues writing for your magazine for the forseeable future - his articles are extremely detailed and informative.
Thanks for your time
—Brad Clawsie, brad@yahoo-inc.com
“...Since I have to work with NT for political reasons, I just cope with
it. But I know if we could see the source, we could probably fix the problem
pretty fast.”
Funny, I didn't see it as Linux advocacy, hollow or otherwise. I saw it as a way to show how the software industry is changing from one we understand in terms of its suppliers to one we understand in terms of its builders. The builder I quoted was remarking on the prefab nature of NT. The world of software, like the world of building construction, has plenty of room for all kinds of construction materials and techniques, including the relatively prefabricated and closed offerings from Microsoft and other software suppliers.
But he didn't say that, and I didn't make it sufficiently clear. He did slam NT, however, with a statement you did well to criticize. And you're not the only one to make the same point. In fact, Craig Burton has been pounding on similar issues for some time. Witness this interview: http://www.linuxgazette.com/issue41/searls.html .
The other points you make are also helpful. And believe me: we do appreciate them. Regards,
—Doc Searls, doc@searls.com
In response to Microsoft release about 'Linux'.
We know that 'Salespeople' will spend more time lambasting a competitor with innuendoes instead of facts.
The best response, of a dignified nature, is to publish the results of comparisons from this months Linux Journal. This article list the test conditions and the results. You don't see Microsoft doing the same. Just muddy the waters with trash (verbal).
We in the Linux community should feel proud of the high priced advertisement and acknowledgment from such a large corporation.
I myself just laugh at the trash coming out of Redmond, but you don't see them doing a comparison of their system (Windows or NT) against Linux.
I believe, the best way to respond to this trash, is to not say a word (no flaming comments) and let the educated people of this world see how childish a large corporation can get. This corporation must think the people of the world are just a bunch of dumb people and cannot see through the trash.
Just think of all the free publicity Linux is getting. I say 'Thank you, Microsoft for your comments about Linux. Keep it up.' They are only making themselves look like a bunch of fools.
—Dr. Fred Lerssen, flerssen@worldnet.att.net
Marjorie;
I was hurt by your interview with Linus Torvalds in the Nov 99 issue of Linux Journal.
As linux user and person of faith, I am interested in Linus' spiritual health. I am not interested in having 4 questions, 3 with a decidedly anti-religious slant, give air to the same tired excuses people use to not participate in corporate worship.
The article was to give the reader a picture “of who exactly Linus is”. But the effort changed mid-stride from a personal glimpse to slamming those who are faithful.
“Religion is a personal matter”. Linus was correct. Some time ago you set a journalistic standard to avoid certain uses of language because you felt it important to remain sensitive to other's beliefs. I ask that you continue that stand. Report and comment on Linux, and the state of open source software in general. Please avoid offending readers with comments that have nothing to do with the magazine's topic or scope.
Sincerely;
—Leam Hall, leamhall@rcn.com
An LJ reader since Aug 95.
Sorry, you were hurt. I didn't think we spent much time on religion and didn't consider it offensive. Even if it was the “same tired excuses”, it is what Linus beieves. I don't think either of us thought we were slamming the
“faithful”. Why would I want to do that? I am one of the faithful.
—Marjorie Richardson
I would like to register my complaint concerning your recent interview with Linus Torvalds. It seems to me there was an inordinate amount of space devoted to Marjorie's and Linus' bashing of “organized religion”.
As a journalist, I thought this showed questionable ethics and poor editing on your part (at times, it almost seemed as if Marjorie was egging him on).
As a Christian, I was offended that a magazine to which I subscribe for news and information relating to a computer operating system should devote so much space to bashing Catholicism, religious practices (such as tithing) and beliefs that I
hold dear. So, Linus doesn't like organized religion. How about sticking to the topic in the future?
—Thomas Long
tlong@billsoft.com
Getting to know Linus and what his beliefs are in many areas was the topic.
—Editor