Announcement

Collapse
No announcement yet.

cuz linux is teh s uck

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • cuz linux is teh s uck

    This was a really great article that put into words my thoughts on the true problems surrounding Linux, namely that there's so much fragmentation that no one group ever manages to get everything right:

    http://enterprise.linux.com/article..../03/01/2248250

    I get asked, by more junior Linux users, and people just looking to try it out, which distribution of Linux I use or recommend. It occurred to me that I never actually published an answer to this question, even though it is, by far, the question I am asked most often. I think my stock answer is maybe slightly unusual only because, unlike most of the rest of the Linux-using world, I hate every distro I've ever tried.

    That's right: every distribution of Linux sucks in its own special way. Some just suck less. However, eventually, no matter what distribution you use, something about it will drive you insane, and you'll try another distro to see if it's any better. What you'll then realize is that, while it may handle what drove you crazy before much better, it handles something else in a way that drives you even more crazy.

    Package management is really the best example I can think of. I'm not even going to talk about the package format, because any self-respecting distro will make it so that the end user doesn't really have to care about that. What I'm talking about are the frontends. The interfaces that enable users to install software. All of these pretty much Suck (note capital "S"). YaST, URPMI, apt-get, apt4rpm, Synaptic, up2date, yum, autorpm, portage (yes, even portage), pkgtool -- they all have features that'll drive you up a wall sooner or later (some sooner than others I guess).

    My favorite? Not that it'll be relevant in about a year, but today, for a desktop system, I think YaST is pretty good as of about SUSE 9.1. However, it can be quite slow, it can be a little weird to navigate at first (because YaST doesn't just do package installs), and some of the default configurations are quirky. I still think SUSE is a very slick desktop, and yet I don't really recommend it much because if a user needs help, and lives in the states, my experience has been that the best help threads are on German mailing lists, and Babelfish does a terrible job with German geekspeak. You can't get help from SUSE unless you've paid, and even then the help you get is extremely limited, and until recently you couldn't even get a free downloadable ISO for installation (they've recently allowed downloads of the DVD ISO).

    But it's not just package management: what about window environments? If you've used a couple of distributions and have come to like KDE, I'd recommend you steer clear of Redhat/Fedora. KDE has always been broken in Redhat, and Fedora appears to be carrying that torch into the community-based distro on Redhat's behalf. Weird things abound in KDE on Fedora. The most recent one I've found is that if you've configured a default browser in KDE control center, and then decide you want to try Konqueror, and you open it up and type in a URL and hit "enter", your default browser will open! Konq won't load the page! These things become funny after 5 or 6 years.

    Maybe you like GNOME? Well, then you have to stay away from Slackware, because Patrick (the maintainer of Slackware) doesn't seem to have a clear direction on GNOME. From what I can tell now, GNOME, going forward, will not be a part of the core slackware distribution, but rather will be handled by an outside third party. I'm not sure how this will pan out, but it doesn't give me a case of the warm-fuzzies. If someone can shed more light on this, leave a comment. ;-)

    Maybe you don't like KDE *or* GNOME! Maybe you like Waiamea or blackbox or fluxbox or windowmaker or enlightenment or twm or fvwm. For older, more stable desktops like Windowmaker and twm, you can find these as installable options in many major distros. For things like blackbox and fluxbox, you don't want the distro-supplied version even if they offer it half of the time, because by the time they release the distro, there are already updates that add major feature enhancements. These younger desktops, while cool, are often moving targets. One I forgot and is very nice for sysadmins is xfce. Give it a shot if you haven't. You can get it in most distros these days.

    What about X and the kernel? These two items are in a state of flux in most distros right now. If you're on the bleeding edge, hardware-wise, you'll likely want the Xorg version of X and the 2.6 kernel. If you're on a laptop, you almost definitely want the 2.6 kernel. However, if you build your own machines, you have to be extremely careful -- I upgraded a mobo to one that supports SATA drives. I'm not using them. My mobo doesn't seem to understand that and reports the drives in the wrong order to the OS, and the newest linux distros choke on it and fail to boot. 2.4-based distros work fine. Earlier, NVidia wasn't so quick in distributing drivers for the 2.6 kernel. All seems ok there now, but fact remains that, while I think these are things users should never have to think about, you do.

    What a choice of distro comes down to, in the long run, is answering the question "which distribution best fits my brain". Asking someone else which distro is best for you is like asking them how you should write a Perl script: there's infinitely more ways than one to do it, and the solution you come up with will undoubtedly look unlike anyone else's.

    In the shorter term, rather than tell newbies to use distro x or y, I tell them what I think was my most important lesson starting out: pick a distro, and stick with it come hell or high water. When you start learning to fix problems instead of reinstalling, only then should you consider trying something else. Different distros treat things differently enough that jumping from distro to distro until you find one that "just works" will only cause you to pull your hair out. Pick *one*, and stick with it. Find the support forums for it, or find a guru who doesn't care what distro you're using. Get it to do what you want, and see if you can get it to do what you want the way you want it done. If nothing else, you'll learn about yourself. You'll learn how you like things to be. Then you can ask in the forums "hey, is there a tool that does what x does, but instead of doing it like y, does it like z?".

    Someone way smarter than me once said: Know thyself, for it is all there is of you.
    45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B0
    45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B1
    [ redacted ]

  • #2
    Sounds like a rant from a inept sysadmin if he even is one to be complaining about more than 10 package managers.

    Notice he likes YaST, what a cool GUI proggy.

    Maybe he should go back to win32 and use Windows Update.

    Every single package manager will give you problems but that is not the fault of the actual software, but developers who make the packages did not follow directions, and yes contrary to what he says, the format DOES matter. You cannot compare for example a .tgz to an .rpm, two completely different technologies which are packaged very differently.
    Delicious Poison:

    The difference between a nerd and a geek? Well a nerd does not wear Spider Man butt huggers.

    Comment


    • #3
      Originally posted by klepto
      Every single package manager will give you problems but that is not the fault of the actual software, but developers who make the packages did not follow directions
      rpm/librpm are awful, namely their use of db3 and the inherent locking problems that creates. There's no detection of stale flock()s in librpm, meaning that if a librpm utility crashes the lock on the package database is still held and rpm can't detect stale locks, so it simply hangs on the stale lock. That doesn't exactly scream 'competent engineering' to me.

      I think Debian has a wonderful format in that it's composed of formats which can be read by standard utilities. All packages are ar archives that contain .tgzs of control files and the actual contents of the package, whereas RPM is a proprietary format that third party utilities can convert to a CPIO archive.

      and yes contrary to what he says, the format DOES matter.
      I think what he was saying was that the popular package managers abstract the intrinsics of the format from the end user.

      You cannot compare for example a .tgz to an .rpm, two completely different technologies which are packaged very differently.
      Which he certainly didn't do, he compared package managers that use those formats noting that the intrinsics of the formats have been abstracted from the end user.
      45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B0
      45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B1
      [ redacted ]

      Comment


      • #4
        After fucking around with NFS all day on Redhat (actually White Box EL) on a 64-bit opteron system where a simple ls would lock up the terminal, I finally dug a couple newsgroup postings saying that it was broken on 64-bit postings. Strace confirmed this.

        Never had this problem on solaris.

        Comment


        • #5
          white box EL should hardly be considered as a real distro to "benchmark" linux. If you want a better RHEL clone, go for CentOS or Tao Linux.

          Getting back on topic, the article writer gives no substantial reason why the mentioned package managers "suck". This is a hit 'n' run article as far as bashing.

          I'm waiting for him to write another article about how a BSD flavor or UNIX flavor 'rocks'.

          Read the section on window managers, he says gnome and kde suck, but "mentions" at the end,

          "One I forgot and is very nice for sysadmins is xfce. Give it a shot if you haven't. You can get it in most distros these days."

          This is a rant article, nothing rock solid to prove linux "sucks".

          Brian Jones is a troll.
          Last edited by klepto; March 4, 2005, 04:06.
          Delicious Poison:

          The difference between a nerd and a geek? Well a nerd does not wear Spider Man butt huggers.

          Comment


          • #6
            Originally posted by klepto
            white box EL should hardly be considered as a real distro to "benchmark" linux. If you want a better RHEL clone, go for CentOS or Tao Linux.
            Its also occuring on my official RHEL distro, my point is that NFS is so freaking old this kinda crap should not be happening.

            Comment


            • #7
              Part of the Point

              I think part of his point is being proven here with the feedback. We all have our favorite distro's and it's for different reasons. How many times have you answered that same question "what distro should I use". It's hard to answer that question to anyone based on anything other than personal preference.

              gg

              Comment


              • #8
                Originally posted by murakami
                Its also occuring on my official RHEL distro, my point is that NFS is so freaking old this kinda crap should not be happening.
                There have been significant changes to NFS implementations on Linux over the years on Linux from original to v2 to v3, and then support for NFS over TCP (instead of the original UDP) as well as 64-bit kludges and then closer 64-bit support. As a result, the implementation keeps being done over or patched.

                This is not an excuse.

                A non-functional NFS Server/Client in a product that is supposed to be a server is awful. You would expect a little testing might have shown this as an issue before the product was shipped.

                Comment


                • #9
                  Originally posted by TheCotMan
                  There have been significant changes to NFS implementations on Linux over the years on Linux from original to v2 to v3, and then support for NFS over TCP (instead of the original UDP) as well as 64-bit kludges and then closer 64-bit support. As a result, the implementation keeps being done over or patched.

                  This is not an excuse.

                  A non-functional NFS Server/Client in a product that is supposed to be a server is awful. You would expect a little testing might have shown this as an issue before the product was shipped.
                  THANK YOU!!

                  This is exactly the point. I have tried several other "Enterprise" distros and they all suffer from this bug in varying degrees.

                  NFS has always worked properly in unixes, and I've probably cross-mounted any number of systems ranging from Solaris, Irix, Clix, and HPUX.

                  Comment


                  • #10
                    Originally posted by murakami
                    THANK YOU!!

                    This is exactly the point. I have tried several other "Enterprise" distros and they all suffer from this bug in varying degrees.

                    NFS has always worked properly in unixes, and I've probably cross-mounted any number of systems ranging from Solaris, Irix, Clix, and HPUX.
                    I run a mixed environement, Linux (RH 7.3, Debian Sarge and FC3), SGI, SunOS 4.1.3 (read that right!) and OS X.

                    For my 64-bit opteron servers, I find that Fedora Core 3 with the latest 2.6.9 kernel (the 2.6.10's appear to have a borked qla2300 fc hba driver) give me acceptable and stable NFS. This is over a bunch of 600GB shares that live on a 5.6TB disk array. There were still weird things (I seem to be able to lock up the server if I do network backups, not sure if the scsi (aha7899), fc or tigon3 driver is to blame... Stale NFS filehandles were an issue, but the latest 2.6.9 kernel along with "no_subtree_check" solved it.)

                    Comment


                    • #11
                      Originally posted by AlexCV
                      I run a mixed environement, Linux (RH 7.3, Debian Sarge and FC3), SGI, SunOS 4.1.3 (read that right!) and OS X.

                      For my 64-bit opteron servers, I find that Fedora Core 3 with the latest 2.6.9 kernel (the 2.6.10's appear to have a borked qla2300 fc hba driver) give me acceptable and stable NFS. This is over a bunch of 600GB shares that live on a 5.6TB disk array. There were still weird things (I seem to be able to lock up the server if I do network backups, not sure if the scsi (aha7899), fc or tigon3 driver is to blame... Stale NFS filehandles were an issue, but the latest 2.6.9 kernel along with "no_subtree_check" solved it.)
                      Thanks, I'll give that a try.

                      Back to the point of the discussion, I can do this to my system with no problem; however, this system replicates a customers environment. They have real issues when it comes to recompiling kernel's or even patch managment. Their configuration management is very strict, so changes like this are show stoppers. Changes like this can generate days and days of meetings.

                      It's stuff like this that makes me hesitate deploying Linux, even well established "enterprise" distros, in corporate environments.

                      Comment


                      • #12
                        http://www.linuxense.com/challenge/


                        If linux is such 'teh suck' then try this. Could be fun.

                        Comment


                        • #13
                          Originally posted by Windblows User
                          http://www.linuxense.com/challenge/


                          If linux is such 'teh suck' then try this. Could be fun.
                          I've never understood the point of these challenges.

                          The puzzle of it is the fun part, but what does it exactly prove? That a bunch of morons have tried and failed or that the sysadmin is too lazy or stupid to properly configure a machine?

                          Comment


                          • #14
                            Originally posted by murakami
                            I've never understood the point of these challenges.

                            The puzzle of it is the fun part, but what does it exactly prove? That a bunch of morons have tried and failed or that the sysadmin is too lazy or stupid to properly configure a machine?
                            I think it's really just for fun, I don't think there's some hidden meaning about how Windows is better than Linux or how blue is the best color in the world, I think they're just letting people have some fun.
                            - Programmer -

                            Comment


                            • #15
                              Originally posted by d3ad1ysp0rk
                              I think it's really just for fun, I don't think there's some hidden meaning about how Windows is better than Linux or how blue is the best color in the world, I think they're just letting people have some fun.
                              I don't know if this is the case for this contest, but contests in the past have used such things (usually with prizes) as a marketing gimmick to say things like, "Hacker Tested, Admin Approved," or whatnot in their glossy advertisements for their product(s).

                              Not offering a prize is a point in their favor to not being a publicity stunt with some commercial application, but that is only based on history of other contests.

                              Comment

                              Working...
                              X