Slashdot Mirror
What Needs Fixing In Linux
An anonymous reader writes "Infoweek's Fixing Linux: What's Broken And What To Do About It argues that the 17-year-old open-source operating system still has problems. Leading the list is author Serdar Yegulap's complaint that the kernel application binary interfaces are a moving target. He writes: 'The sheer breadth of kernel interfaces means it's entirely possible for something to break in a way that might not even show up in a fairly rigorous code review.' Also on his list of needed fixes are: a consistent configuration system, to enable distribution; native file versioning; audio APIs; and the integration of X11 with apps. Finally, he argues that Linux needs a committee to insure that all GUIs work consistently and integrate better on the back-end with the kernel."
and the integration of X11 with apps. Finally, he argues that Linux needs a committee to insure that all GUIs work consistently and integrate better on the back-end with the kernel.
Call me old fashioned or whatever the cute term is now. But fuck that! If I ever see programs like cp become bloated with X library calls because some news reporter needs to see a GUI progress bar, I'm going to be very angry.
1. Why is Linux blamed for configuration files that are written by application developers? Linux is a kernel and is not responsible for Sendmail. Further I fail to see how the point and click method of configuration is better than editing a text file than can be searched, backed up and version controlled.
2. Why is it the responsibility of Linux distribution maintainers to provide a means for commericial vendors to package their product? Vendors had to spend money to get certified for other operating systems. How about putting a little work into understanding and using a Linux distribution.
3. X freezing? Umm...
Perhaps I've just feed the troll but, I'm sure the pointy hairs will read the artical and think it's all true.
UNIX/Linux Consulting
In the sense that there is little originality, and it seems anything added to linux has to have occurred in another operating system.
Linux/Unix has plenty shortcomings, but its evangelists believe it's so perfect it cannot be improved. Here is my short list of major peeves.
1. Filesystem metadata/permissions. Why do files still have to have rudimentary metadata? Drives are massive and a few bytes would not harm. MacOS has added metadata. An example would be that a file should be able to keep a list of all the dates it was accessed. Why can a file only have one owner/group?
2. Root is God. This must really be fixed. There should be a way for root to irrevocably divest its powers, and root does not need to access users file. A user should explicitly grant root permission to read his files. It will always be a major security issue because all one has to do is become root. Plan9 managed to do that.
3. They lie about everything is a file. Why not extend this to networking resources ('cd http://www.gnu.org/ would be cool ). Plan9 also succeeded there.
I am sure linux evangelists are going to propose (hack-filled) workarounds or reasons it can't work, but I don't buy it. That is why I left linux.
I use Linux and FOSS almost exclusively for my photography workflow.
Almost.
See, when I work in film, I need to have a Mac around to handle the flatbed scanner. Because, unfortunately, Linux support for flatbed scanners really sucks rocks.
gimp has some shortcomings as well but I understand they are being actively addressed so I won't bitch about that.
I disagree... it IS about choice, but more importantly, the competition to be top dog keeps everybody on their toes, and they are all doing some great work.
I disagree that Linux is a mess, it's no more of a mess than Windows, and in many ways it's a lot better. We may have different distributions competing with each other, but MS OSs still have to compete with each other... XP, XP Pro, XP Media Center, Vista, Vista Ultimate.... all with at least slightly different capabilities.
Stupid sexy Flanders.
I'm with you. If my laptop Broadcom wireless worked out of the box on Ubuntu, I'd be using that instead of Windows. I used to know plenty about setting up a Linux system (back in the 90s), but I have better things to do with my time than figuring out how to make something work that should "just work" after installation.
I don't see it getting better. I don't really see a lot of money for vendors in better supporting Linux. Personally I don't care enough about the OS to buy hardware based on Linux support. Hardware shopping for me is about comparing price vs capabilities. It should be a given that the machine will work. I'm at the point now where if I ever buy another "Unix" system, it will almost certainly be a Mac.
How can Linux win me back? Whatever machine I bring home from Best Buy has to "just work" at the end of the install/config program. Is that too much to ask for?
A few years ago you might have had a point, but now the standard "Granny" answer is quite clearly something like "Yes, there's all these different versions with various (effectively minor) technical differences but Ubuntu is just fine for what you want."
If Granny later needs a new PC she can even have a cheap Dell desktop/laptop/netbook with Ubuntu pre-installed. She doesn't have to think about Gnome vs. KDE - she'll get the default (Gnome) and not even know about it.
Ubuntu: The best choice for those who don't want to have to choose.
Oh, and re the "elitist attitude" you won't find that on the Ubuntu forums.
What I find in *most* of these sorts of pieces is that they are either cynically or subconsciously pushing for the winozification of Linux. He makes some good points along with the bad.
(1) Package Management
This is a good point if the debian people and redhat people could work toward a solution, it could be fixed as both systems have a great deal in common.
(2) Configuration Files /etc are, IHO, better and can be backed up and diff-ed.
Bzzzt. Wrong. The foolish part of this subject is that while the Windows registry provides a standardized access to the data store, it only defines types and not what they are supposed to be. Lunux configuration files under
(3) Kernel Application Binary Interfaces
I would like to see a stabilized and standardized device interface API for standard devices, something exposing a limited subset of the kernel that would simplify simple devices like block, serial, and network types of devices.
(4)Native File Versioning
Bzzt. Its called automatic backup people. This is a relatively new feature in Macs and barely working in Windows. Would be nice, but can't characterize it as something that's broken.
(5)Audio Application Programming Interfaces
This I 100% agree with. Choice is nice, but the geometric product of "choice" in system services means that rich multimedia applications are much harder to develop.
(6)Graphical User Interface
He sort of has a point about this and it has often been a problem.
(7)Integration Of X11 With Apps
Bzzt Wrong. X11 is a HUGELY powerful system and if you encounter a bug that crashes your session, that's a bug. Fortunately I haven't seen one of these in about 6 years.
(8)Commercially Hosted Backup And Restore
Bzzt Wrong. This is not "Linux" being broken, it is 3rd party vendors being stupid.
Video drivers. Video drivers. Video drivers. Or possibly X11. Or possibly the whole multi-layered graphics approach.
I'm not sure whether it's the fault of the Linux kernel or the graphics card companies that won't release their hardware specs / open-source drivers. To the "newbie" Linux user, that's more-or-less irrelevant. I just know that I installed Linux for the first time two months ago on a brand new computer (AMD Athlon X2 6400+, Asus M3N78-VM with onboard GeForce 8200 graphics, 2 x 2 GB DDR2-800 [PC2-6400] memory) and gave up last week and installed Windows XP after spending eight weeks dicking around with video drivers / KDE vs. GNOME / xorg.conf / etc. trying to get the desktop performance up to the level where it doesn't take a good half-second for a bloody Firefox window to stutter its way up to full-screen from the minimized state.
Karma: pi (Mostly due to circular reasoning in posts).
Agree very much with your comments. Really the article should have been entitled "Clueless journo thinks success=approval by crappy prorpietary vendors" or "Clueless journo thinks becoming Windows XP is the be all and end all of free software development" or "Clueless journo used some distro, ignored 99% of tools and capabilities and then sounded off like an asshole after a big dinner". You can always spot a doofus when he/she starts saying "integrated" in a completely meaningless way. The only valid points he has is that there is a mess of audio APIs, the rest is really junk. Why on earth would anyone want to pay for some commercial remote back up service like the ones he mentions? What exactly is supposed to be broken or wrong with cron, rsync and ssh? If you need a gui for that stuff it's also available (kde kio slaves and kcron come to mind). If the Linux kernel and the Linux/GNU/Busybox operating systems were as broken as this journalist's understanding then we would indeed have a problem.
How can you blame anyone for coming to this conclusion? When you "face the facts" and clearly come to the conclusion that the pros do infact weigh heavier than the cons you have to ask yourself, "And why aren't we all using OSS yet?". Thus beating Microsoft is not relevant, but beating proprietary code is, and thus indirectly also Microsoft, Apple etc.
I am the lawn!
USB barely works. It's OK for mass-storage devices, but sucks hugely for high-bandwidth devices, or anything that's removable - and gets removed.
I'm sorry but as someone who spends a lot of time countering the Linux is good, Windows is shit brigade, that you're wrong. USB works a treat and has done for many a year. I have however had issues with an iPod on Windows and the usbstor.sys bug.
Conor "You're not married,you haven't got a girlfriend and you've never seen Star Trek? Good Lord!" - Patrick Stewart
"When talking to grandma about trying Linux since all she wants to do is check e-mail, look at pictures of the grand kids and keep her MySpace page updated, you get the question thrown back..."why so many different ones? Are they all different?"
Why the hell is your grandma worried about which Linux is she going to use? Does she knows about the differences between distros? Just grab her a Fedora or an Ubuntu and get over it.
"Stop being so darn elitist! You want people to use it, then be friendly about it.". No. They will come when they are willing to learn. GNU/Linux is not a shrink-wrapped product - it's a very versatile and configurable tool. And no. I don't want them just to use it. I want them realizing it is a better choice and, if for them it's not, I want them to use whatever makes sense for them. There is Windows and there is OSX. There is a lot of choice around.
"We should be able to point the average Windows user to "Linux", a single cohesive product."
No. Because GNU/Linux is not a single product. It's not even a product. It's, like I said before, an immensely versatile tool backed by a community of fairly bright people. It's made by them and for themselves. If you want to try it, fine. But don't start making suggestions or bossing them around. Did you pay for it? No? Then don't be surprised if they don't feel like they are your employees.
http://www.dieblinkenlights.com
Articles that point out what's wrong with Linux (or other OSes for that matter) invariably point out flaws in the interfaces or in the code. This one only falls partway into this trap, but still missed some stuff that can't be fixed by more programming.
1) Documentation. Linux documentation is STILL an unholy mess of inconsistent, incomplete, uneven snippets. They need to make the man pages complete and consistent. The docs HAVE to be updated consistently! There has to be quality control!
2) Compatibility. Both backwards and cross-platform, that is. The number of programs that require a specific version of package 'x' for no good reason is astounding. Similarly, the number of things written in a Linux environment that compile on the author's computer and NOTHING ELSE is ridiculous. One classic example was mplayer, which would (painfully) compile on other platforms, but the man pages would only format properly when using gnu-troff. Standard troff would have done fine, if the authors actually had a clue.
Another example: Burn bash into the ground, and hang the developers who made it incompatible with /bin/sh. ksh, zsh, ash, and others have managed to be feature-complete supersets of /bin/sh, but bash? No, it does not properly execute some sh scripts. (And before you start, legacy is at least as important as POSIX.)
3) attitude. It's not the fault of Linux, but the community that surrounds it. I can't be bothered with an OS if I have to wade through religious zealotry every time I want something done.
Linux is 17, going on 18. It's time for it to come of age and grow up into a mature OS.
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
If Linux beat Microsoft, it would probably mean Adobe programs and most or all commercial games would be released for Linux, which would mean I could stop dual-booting.
Consequently, the single greatest "feature" that Linux could work toward in terms of improving my experience as a user would be to "beat" Microsoft. Really, everything else works fine for me; I can't think of anything else I want. It's better for me than Microsoft products in every other way except those. Oh, wireless drivers I guess. More and better wireless drivers. And you know what? That's another problem that would solve itself if Linux took over as the market leader.
Games. Adobe programs. Working 100% at release. That's what I want, and it likely means beating Microsoft. Or, beating Adobe to become the industry standard in its arena(s), then still beating Microsoft to bring in gaming. Either way.
(I understand and respect your sentiment, just giving one perspective on why it can be reasonable to want Linux to "win" for reasons other than "M$ is teh suxor!")
Why the hell should we have to? I prefer having a choice. Windows users have plenty of choices when it comes to desktops -- Litestep, emerge, talisman, Aston and many more -- so why shouldn't Linux users? Simple answer: NO.
Clearly what you want is Mac OS X, not Linux. So use it instead of bitching that Linux isn't what you want.
I dream of a better world... one in which chickens can cross roads without their motives being questioned.
Depends on how you define "different". Different in default installed software? Certainly.
Different in how the developers want their distro (and in turn, their community) to evolve? Perhaps.
Different in how the system behaves and looks? Maybe so.
The question has IMO always been: Are the high-profile distro's (Ubuntu, SuSE and Fedora, for instance) enough for people to get the job done? If so, then why look further.
If the "big ones" don't give you what you need (be it software choice, level of difficulty or something else), _that's_ when you look further to the less obvious candidates.
I partially agree here: it would be easier if both DE's would quit quibbling and simply merge, taking the strengths of both and the weaknesses of neither. The problem with this is however, who defines what is a strength or a weakness? KDE users for instance like the high customizability of their DE of choice, while some GNOME users see that as overcomplication and bloat and refuse to use it.
If all users were in agreement of what the best course of action was, we wouldn't have these two DE's fighting over the market share now.
Fanboyism is a bad thing, and I have to confess that I myself have at times succumbed to it, but the point here kinda hinges on "You want people to use it". I don't.
I do not want nor need every Fred the Baker (Joe Plumber is out on his ear, sorry;))and his mother use Linux. Quite frankly there is no reason for them to switch from the OS of their choice. If they prefer Windows of any flavour, let them use that. It doesn't bother me in the slightest. (And yes I know about the botnets).
People should stick with what they know and prefer.
I like Linux, I've used Gentoo exclusively for 5 years now and will never switch to Windows myself, but my GF is free to change to and from any OS she so chooses to run.
Zealots exist in any discussion, it's nearly impossible to eliminate them. The best we can do is let them be and ignore them, they'll eventually have to come around when no-one listens to them anymore. And if they don't... Well, no-one's listening anyway. ;)
Posting obviously for anonymous reasons.
So the common user can install Ubuntu, and someone else can use Gentoo. What's the problem again?
As a Linux old-timer who was doing multi-homed networking on my laptops since the dark ages, I find no solution other than disabling NetworkManager entirely.
The underlying wpa_supplicant is actually good enough for my purposes as I do not have your need to avoid DHCP. But you could even script that I suppose.
I just have a wireless.sh script to start/stop wireless (including adding/removing driver module) and manually do "ifup eth0" when I want wired. The battery sucking way that NetworkManager tries to keep all links up and poll their status is not only really unstable for me but sucks battery.
The problem isn't NetworkManager, but that distros have the mistaken idea that NetworkManager will ever be general enough to support the majority of use cases.
I don't think choice is the problem, I see the "common user" as the problem. Years ago a computer was something you needed a degree in just to understand the concept of how it worked, let alone use it. The idea that computers would be sitting on everyone's desk at work manifested and blew up to the point of having many computers around the "common user" at all times. A computer for work, computer at home, and now netbooks to carry around in our pockets. What happened to researching the system that's being used? We can throw words around like "zealot" and "elitist", but its a pride thing. We shouldn't feel guilty for going the extra mile to understand and use a system that may not fit the mainstream mold. The "common user" doesn't fit that category, because they lack the drive to understand anything they can't quickly click a few times to grasp the concept of. This is a problem because they're more prone to being suckered into pop-up adds claiming their system needs fixing by the ad in the pop-up. The attitude changes when they want someone who knows the system to come in and fix it for them. All it takes is a little thinking, will power and the want to understand the system that they're using. Thats what seperated us "geeks" or "nerds" from Joe Average with computing. Why has this suddenly become the opposite. That the computer should just be simplistic, and it's a bad thing to use our brains to comprehend the system we're using. This isn't just a linux problem, this is a problem for all computing. Apathy, and expecting the system to do something you weren't telling it to do correctly. Linux supporters shouldn't feel bad for having a system that takes thinking to use. We should be encouraging more thought and brain power being used in everyday life, it's what makes us smarter people.
Mod WAY up. Although, I must correct you: the original Unix parallel model was multiprocessing. Threading was popularised by Windows (which has a heavy process model) and Java, more than anything else - yes, I'm well aware that threads existed long before these guys got in the act. Linus himself always argued against threading, and he was right.
Anyway, a move away from multithreading back to a multiprocessing model can only be a good thing.
Exactly.
This is "The Free Market" in action. People see a need not already covered by existing Distros (or covered, but with a pile of other crap they don't need) and they go ahead and roll their own distro, and then throw it out there in case anyone else has a similar need.
What this guy is suggesting: Central Planning for Linux, is essentially Communism. Nobody gets to make changes without going through the central planning group. In theory, it can sound alluring. In practice, it would mean the death of Linux. Does this guy work for Microsoft? What horrible ideas!
I particularly get a kick out of one of his major ideas: A COMMITTEE! Yeah. that's JUST what Linux needs, a "Designed by Committee" label on it! /sarc
Leave Linux alone. It doesn't need central planning, it works just fine as it is.
Official Heretic from the "Church of Global Warming". Proven right thanks to whistle blowers. AGW = Flat Earth Theory
OK, so over the years I have always heard this gripe about Linux not being consistent in this area or that area. X11 doesn't have standard GUI? Well, if I recall correctly and I'm not an X11 hacker, but it is open source and Window Maker has always ran well on all of them well.
So, maybe everyone is talking about perhaps the NeXT Step-ish style of Window Maker as opposed to GNOME, KDE, Enlightenment, xfce4? Well, while it might be true, going from Blackbox to Afterstep is a bit of a culture shock, it's hardly a bane to the Linux community.
Non consistent configuration means.... Non consistent UI and non-consistent interfaces through out.
All of this makes PERFECT sense... IF DEALING WITH PROPRIETARY SOFTWARE. But since it's open source, my argument is that all the interfaces are easily accessible via source if nothing else.
There is a operating system which has much of these things... consistent interfaces, UI and configuration. It's called Windows. Oh, and might I add, that even Apple had the good mind to realize that a centralized-consistent-unified configuration utility is often enough a bane where they have a clear and obvious ability to turn it off in MacOS X! Windows, you can't so easily turn off the registry... if you do... good luck.
Linux has a clear and consistent informal trend in regards to configuration utilities. It's so simple it's ridiculous. It also contributes to 98% of the "power" of Linux and other unices. Flat file, plain text key-value pair configuration files. Someone in the Unix history, can't remember who, said plaintext is the universal API. Flat file, configuration is an application of a plaintext API. There are opensource applications that make configuration into binary files and such, and they are a pain to deal with.
So Linux does have a consistent, generally accepted means for a configuration architecture, Plain Text Files!
It really bothers me, these yahoo's, who have the power of the pen and thus perception, these so-called journalists, columnists, lingual devils and verbal vampires who spew their assertions onto the web in hopes of change; who write in total disregard to the essence of Linux's existence. These people, obviously do not realize, that what they are complaining about is what makes Linux so blatantly powerful. They view the surface, but ignore the mechanics. They see that sendmail configuration is different than apaches, but fail to see that both are plaintext and editable from program, script or human just the same.
And some of use actually listens to them!
They are lazy. They don't like to think. They want to be told, once and for all what to press, how and when. To launch an app, you may only use a recognizable cursor, across the same blue background, over an identical 32x32 icon which has been blessed as the only icon for that application regardless if you are on FreeBSD and KDE, or Gentoo Linux with enlightenment. They want a unified means of configuration, even if it means everything must be binary, and uneditable and totally obfuscated to the human.
To hell with these reports of Linux should actually be Windows. That is essentially what they are.
I was citing one office as an example, not as the definitive sole issue holding back linux. Be realistic. No one piece of software will make or break an OS, but my use of Solidworks, Pro-E, and Lightwave are, in this case, definite examples that keep this particular office from looking at linux. Denying the existence of the elephant in the room does not make its actual presence a fact.
You can extrapolate from that to include Photoshop (no, GIMP is not good enough for serious work), accounting software, 3DS max, and much more.
The point was that by presenting a real moving target, that the lack of software is rather self-inflicted.
Windows presents much less of a moving target that linux ever has. Games that ran on my 2000 desktop over 8 years ago work just fine on my Vista desktop. Old Photoshop works fine too.
I'd love to see real graphics packages ported to linux. Having Photoshop and a real 3D package (3DS or Lightwave) available would go a LONG way to making it my primary desktop and not just my development box.
UNIX, and Linux doubly-so, aims for source compatibility and improved architectural simplicity at the cost of some administrative complexity, aka 'Worse is Better'
Which "UNIX" would that be ? Of all the remotely mainstream "UNIXes", Linux is the only one that has serious problems with binary compatibility.
sorry to burst your bubble, but open source development is essentially communism. there's no free market at work here. it's the collective efforts of the open source community that drives development. it's not the invisible hand of the free market that's writing code and submitting patches. most open source developers are not motivated by profit, but instead donate their time to open source projects either out of altruism, a sense of community, or simply the love of writing code.
the free market has more relevance to closed/proprietary software, which users actually have to pay money for, and is supported entirely by commercial profits. how does Linux fit a free market model when Linus Torvalds and Richard Stallman both intended Linux and GNU, respectively, as collaborative development efforts that anyone could freely use, modify, and distribute?
GNU/Linux is absolutely the Free Market. Not a free market of MONEY, but a free market of IDEAS. The very same principles that make Capitalism work are the ones that make GNU/Linux work. Only here people are doing it for Merit or personal satisfaction/need rather than money (Although there are plenty of people that make good money off of Open Source work as well.)
The key difference is the mechanic. In Communism, everything is directed by the Central Planning group. Whether that group be a committee, or a single "strong man" everything runs through that central group, and there can be NO WORK DONE without prior central approval. This means that communistic or centralized economies (and organizations) are slow to react to change, prone to error, and monolithic in approach to any issue. It's a poor way of getting things done.
In Free Market Capitalism, people make their own decisions on where to work, what work to do, and whom (if anyone) to work for. They work to fulfill their OWN NEEDS FIRST. Whether that need be for money, or merit. By each individual fulfilling their own needs, the "invisible hand" of the Market brings together all those disparate desires into a cogent whole. This is the very definition of Open Source. GNU/Linux falls quite neatly into this arena.
In other words: Proprietary Software is Communism, Open Source is the Free Market.
The sooner we all understand this, the better off GNU/OSS will be.
Official Heretic from the "Church of Global Warming". Proven right thanks to whistle blowers. AGW = Flat Earth Theory
MS Windows and the Mac also have hierarchical file systems, but they don't use the notion of current working directory the way Unix does. My reference to the "file system" was sloppy, but what I said should have made it clear that this was what I was referring to.
No, these are not diametrically opposed views. You have to step outside the box to understand its dimensions, or lack thereof.
We've seen a Window-izing of Linux for a long time that amount to parallel development efforts undertaken under the aegis of different values and development animals, but they both still walk like a duck.
Fortunately, the need for ducks is high. But we have to see beyond what's currently here. The basic principals go back to Unix and PARC SmallTalk and iPCs/RPCs, and variations on that theme. Hardware gets cheaper, and coders get sloppier. This cycle's been going on for four decades now. I yearn for something that breaks the model, gives state machine computing a run for its money, and really challenges how we think.
---- Teach Peace. It's Cheaper Than War.
true communism has nothing to do with a stalinist dictatorship. marxist philosophy proposes an egalitarian society without socioeconomic inequality. that clearly did not exist in stalinist Russia, as party members still had much more wealth and power than the average worker. and contrary to what many Americans believe, a communist economy does not preclude a democratic government. if you only have a single "strongman" or a non-democratically-elected "committee" making political decisions, then that's a dictatorship or oligarchy. a communist society by definition needs to be democratically run by the working class.
true communism has only really be achieved on a small scale in experimental communes. the basic idea of communism is of putting social cooperation ahead of competitive self-interests. this ideal is illustrated in the creation of local farming co-ops, in which everyone in the community works together to achieve a shared interest--providing food for the community. everyone contributes what they can to the farming efforts, and in this way everyone gets to eat for free, and no one starves. likewise, an open source project enlists the help of the community to develop the software. everyone contributes what they can, and their collective efforts result in free software than everyone can use/modify/distribute.
the difference between a free market economy and communist economy isn't freedom to do what you want. the difference is competition vs. cooperation. a socialist society doesn't have to be a dictatorship. just because health care and education are socialized doesn't mean people have any less freedom. arguably you have more freedom in a communist/socialist society because what you're free to do isn't limited by what you can "afford." that means that if i'm interested in computer science, and my grades are good enough, then i'm free to study CS. this stands in contrast with a purely free market education system, where if i'm not born into a privileged background, i can't afford to go college and pursue the career i want, just because i can't afford it.
Oh, and would someone do some work on documentation?
I wrote documentation for professional software for years, including Linux software. I had numerous colleagues ask me if there was a way they could write docs for OSS projects as a way of contributing and I tried really, really hard to help hook them up with projects... any projects. It was almost always a failure. Developers of OSS don't seem interested in the help and rarely take the time to explain parts of the software the writer can't figure out on their own, even though that explanation would then have made it into the hands of all the users. Good documenters are also user advocates who approach software from the end user perspective and thus tend to come across and point out a lot of usability issues in the course of their job. OSS developers tend to not only ignore such input, but often became rude and abusive when provided with such feedback (which companies usually have to pay big bucks to get). Finally, I've done graphics work professionally and can tell you the same is true for their attempts at contributions. One of the best graphics guys I know, who makes big bucks for his work, was so poorly treated when he tried to submit some free textures to an OSS game he played, he has vowed to never again try to work with "OSS nutcases" as he now calls them.
All that said, if you're developing an OSS project and would like documentation help and you're willing to commit to actually working with a professional documenter, I can probably hook you up with a recent college grad, or a professional looking to do some free work to expand their resume to include OSS.
Let's use apache as an example, threading MPM vs. forking MPM. Note that forking is the original heavyweight threading model.
Apache gets a request on port 80, forks a copy of itself, and starts processing the incoming request. Now if that forked copy throws an exception, it won't kill the master port 80, or the other 10's or 1000's of child processes that are still processing requests. Access to shared memory has the same problems as threads - it must be synchronized, and access must be explicity granted.
With threading, all this is cheaper and faster, but it's also far less robust. One thread deferencing NULL improperly and they all die.
To fix this issue you would first need to get rid of the dependencies, but to do that you would need a large enough and stable enough core system on which applications could depend instead, but given how Linux development works its not clear if that is ever going to happen or even desirable.
No, you don't. All you need is a stable libc. For everything else, if you *really* want to support a wide range of distributions and versions, you bundle all the dependencies into your package, which is either just a simple .tar.gz that the user can unpack whereever they please (with a launcher script that sets up the lib path, etc.), or a self-bootstrapping GUI installer that can install the app wherever the user wants, plus can take care of the other background install tasks like setting up mime-type associations, etc. (The mime-type stuff can use the freedesktop.org standards, and that will work fine for the vast majority of end-users who use GNOME, KDE, or Xfce.)
Is this ugly and a redundant waste of space? Sure it is. But that's the de facto recommended practice on Windows, and it gets the job done. How many installers on Windows either expect you to have dependencies installed, or have some sort of dependency resolver built-in? Very few, possibly NONE. The only things that come close are gtk apps like Pidgin which offer to install a common copy of gtk for you, and apps like wireshark/ethereal that come bundled with a copy of libpcap that will get installed automatically if you need it.
None of this is rocket science, but the Linux crowd (myself included) is so hung up on code reuse, modularity, and shared libraries, that no one can target Linux because there's no such thing as "Linux the OS" from an ISV standpoint.
Apps targeting Linux should assume that libc is present, and bundle everything else.
Xfce: Lighter than some, heavier than others. Just right.
Most of TFA's criticisms of Linux assume that the lack of standardization is a problem. I say that this is the level of customization Linux users have come to expect.
One of the key features in Windows is that the graphical shell is basically hard-wired into the operating system. While this allows for the GUI consistency that the author seems to value most of all, the GNU/Linux architecture was never meant to do so. The result is great flexibility in how the machine's state is presented to the operator. The price, of course, is that any number of graphical libraries end up being used to write programs.
I like the idea that a distribution encompasses a large amount of software that can be installed from a single vendor. While this requires that software may need to be installed from many repositories, in practice it is not difficult to find a repository with the software that is needed. Package management is handled differently, because it achieves greater flexibility and security. It's the very idea that you must go to a vendor and download an executable that installs the package that Linux intends to avoid. Instead, the vendor provides a crypto signature that identifies their packages, and a repository to download them from. This way, the vendor's signature can be verified before installation, and no arbitrary privileged code is executed, as the package manager is invoked for installation. The package manager program can be monitored for changes to ensure that the installation procedure is always the same. This is not the case for most Windows installers (though MSI files are better than an EXE). Any software obtained through a repository will be automatically updated as the vendor provides new releases. While core Windows functionality may be checked regularly, each installed program must provide its own automatic update process rather than using the system updater. Having a single system updater is superior architecture since it minimizes the amount of privileged code that much execute.
The graphical/multimedia desktop is undoubtedly more complicated as there are competing design philosophies. While the general concept of one-click execution is present in even the most basic window managers (TWM, WindowMaker, BlackBox, etc.), other window managers attempt to provide a more unified interface to the system. On the other extreme, there is KDE, which provides many integrated services (such as audio extraction from CD) directly into the windowing system. Middle-of-the road desktop philosophies such as GNOME provide graphical and VFS consistency, but the same audio extraction task requires the installation of a separate program. In the end, however, every one of these environments can be quickly customized to provide efficient graphical shortcuts to any software that is installed. In most cases, the vendor's package will provide these shortcuts, and they can be placed in the "Start" menu, on the desktop, or in a quick-launch style configuration on the taskbar. There's really no difference in the end result, nor is there any appreciable difference in effort involved.
While there's some truth to TFA's criticism of the competing audio subsystems, in practice there's not really any issue. PulseAudio, included in Ubuntu 8, provides a unified interface to both ALSA and the legacy OSS driver, allowing for simultaneous use of the audio driver by both ALSA and OSS apps. Again, a simple one-time configuration, and the application is still launched with a single click.
TFA complains about the need to kill the X server should it crash. In practice, it's only buggy video drivers that will cause that to happen. Fair enough, I've seen Windows Vista successfully reload a crashed video driver. Still, anything prior to that version of Windows would require an entire system restart in this situation. Simply restarting X is preferable to the entire system restart.
Finally, there's some discussion about things that make binary-only development for GNU/Linux annoying. Of course, you could look at it the oth
"Please describe the scientific nature of the 'whammy'" - Agent Scully
Instead of all this dork gibberish, why not get Flash to work properly and the desktop icons to play nice with one another?
Walk before you can run.