Fallout From Munich

The newly elected mayor of Munich and some of his buddies have raised a furor by suggesting Munich migrate back to M$’s clutches.“The city’s IT department, the city council, as well as third mayor Christine Strobl, support the current IT strategy and thus distance themselves from the criticisms of the first and second mayor. Strobl emphasises that "upon careful checking" she still considers the switch to Free Software as the right thing to do.
She has a sound economic basis for this view: Due to reduced licensing expenditures alone, the city was able to save 11 million Euro. The hardware cost alone of switching to Windows 7 would amount to 3,15 million Euro, with Windows 8 being even more expensive according to the IT-administration. A switch would incur additional costs, and mean the loss of the achievements in the support of open standards.”
That’s not happening if they go by the numbers, many $millions wasted for no improvement in IT at all. Meanwhile the migration to GNU/Linux more or less broke even, costing about what one or two rounds of staying on the Wintel treadmill would have cost and saving a bundle with each step Munich doesn’t take on that instrument of torture. It’s interesting that cost was not a prime motivator for Munich to migrate but it may be the reason they stick with GNU/Linux.

see Munich sticks with Free Software.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology and tagged , , , , , , , . Bookmark the permalink.

224 Responses to Fallout From Munich

  1. DrLoser says:

    I can’t stop myself. TEG correctly points out that the following is gibberish:

    Posix applications don’t like this [threads being assigned a unique PID]

    To quote TEG, “POSIX applications” don’t need to “like” anything.

    TEG, being a rational human being with a decent understanding of POSIX, stops right there … which is to underestimate the sheer ignorance displayed by oiaohm’s comment. So, let’s think this through.

    POSIX applications do not like this.

    I assume this is not some weird anthropomorpisation of a simple CS standard. Ignoring the emotional state of a “POSIX application,” therefore, let us consider the purely bit-logical state.

    There are only two ways in which an application, of whichever genus one chooses, can demonstrate dislike of a particular standard … and I think we can all agree that POSIX is a standard.

    1) They can be starved of information relevant to them, which should be returned by each and every syscall in the POSIX standard. I’m hard-pressed to come up with an example here.
    2) They’re badly-written pieces of junk that just cannot cope with reality.

    Which means, oiaohm, that you have just accused either POSIX standards of being extremely badly written or Linux applications of being extremely badly written.

    Your choice, really.

    And it should be noted that I, the troll … and probably even TEG too … would defend POSIX-compliant Linux applications against this weird and unsubstantiated judgement.

  2. DrLoser says:

    Once you’ve passed that simple (indeed trival) test, oiaohm, we can move on to the following little conundrum:

    Why does POSIX have any need for sem_open(3)?

    Kindly soul that I am, I have even provided you with the link …

  3. DrLoser says:

    The fact the kernel does not care means you can use multi process to emulate multi thread without any major issues.

    Once more with this “kernel does not care” and “emulate” hand-waving.

    All right then, oiaohm, let’s try to bring this discussion to a satisfactory, yet basic, conclusion.

    Ignoring PID1 and other processes/kthreads/etc that actually get a Linux system up and running, let’s focus on the simplest possible illustrative case.

    I specify two applications, FOO and BAR. Each of them are launched in some way from PID1. I further specify that both FOO and BAR, in an MMU system, call pthread_create().

    We now have exactly four “processy-thready thingies,” and since your supposition is that the kernel treats them in an identical fashion, I’m not going to pass judgement on whether they are processes or threads. I’m just going to call them A, B, C and D, in order of creation.

    Now, suppose that FOO has a static variable (not thread-static, mind you) called “fooname.” And suppose that BAR has a static variable (ditto) called “barname.”

    Which of A, B, C and D has access to “fooname?” And which of A, B, C and D has access to “barname?”

    Bear in mind that this is just the simplest example. To take another, you’re going to find the behaviour of pthread_join() a little surprising, given that you fondly believe that the kernel treats A, B, C and D in identical fashion.

  4. That Exploit Guy says:

    That Exploit Guy uclinux page is one of the traps I warned you about. uclinux is referring what the kernel provides with limited options uclibc…
    No fork is not exactly true. uclibc has a stubs mode on nommu systems that redirect the fork implementation. Yes you can have fork its a hybrid vfork beast.

    Everyone having worked with uClinux knows there is no fork(). You must be friggin’ unique.

    “Memory mapping behaviour also involves the way fork(), vfork(), clone() and
    ptrace() work. Under uClinux there is no fork(), and clone() must be supplied
    the CLONE_VM flag.”

    No “traps”, no fancy clone() tricks. You simply live without fork(), and that’s for everyone working with uClinux, Period.

    Fork usage in Linux kernel space is older than kthread.c. sys_clone and Kthread appear in the same year as part of the same patch set.

    Idiot. sys_clone has been around as early as kernel version 2.0.40, whereas kthread.h and those what you call “[bla, bla]” shown by ps -ax are strictly version 2.6+ things.

    Yes you are trying to build a pre 2004 Linux driver you can find fork usage in kernel space. Not everything about early Linux was sane. [bla bla] stuff is history markers.

    kthread.h is just a convenient way to create threads in kernel code. Ultimately, what it does is simply calling the misleadingly named do_fork() in fork.c with the CLONE_VM flag, meaning that the parent and child share the same memory space.

    Again, there is not fork().

    That was only one example for fork on dos.

    Then you proceeded to provide no example of fork() in DOS.

    There are many examples prior where application ported to dos has fork without having.

    It either has something, or it doesn’t. There is no in-between.

    That Exploit Guy please note DrLoser said not at all. Fork on dos does exist in strange and bad ways.

    DOS itself does not provide a C API at all. What you think of as “fork” is a set of functions modelled after the VMS API and offered by the compiler.

    That Exploit Guy if we were talking about BSD or Hurd kernels a Process means something to the internals. So you cannot just go screw it.

    No, I do not want to “screw it”. Do you?

    The fact the kernel does not care means you can use multi process to emulate multi thread without any major issues.
    Again, despite the distinction between processes and threads being an easy-to-understand concept, you are showing a feeble grasp of it.

    Before Linux 2.4 each thread got its own unique PID. Posix applications don’t like this.

    “POSIX applications” don’t need to “like” anything. Thread groups are just a way to implement POSIX threads in the kernel. The alternative is to simply implement POSIX threads in the userspace.

  5. oiaohm says:

    That Exploit Guy uclinux page is one of the traps I warned you about. uclinux is referring what the kernel provides with limited options uclibc.

    No fork is not exactly true. uclibc has a stubs mode on nommu systems that redirect the fork implementation. Yes you can have fork its a hybrid vfork beast. Still passes posix standard requirements but its not behaving exactly how you expect. Of course for stability it recommend to build uclibc with stub mode redirecting functions to close approx off.

    Fork usage in Linux kernel space is older than kthread.c. sys_clone and Kthread appear in the same year as part of the same patch set. That Exploit Guy I said old stuff. Kthread/Clone patches remove all the horible fork stuff in Linux kernel space. kthread had to maintain the listing in the process table for backwards compatible. Yes you are trying to build a pre 2004 Linux driver you can find fork usage in kernel space. Not everything about early Linux was sane. [bla bla] stuff is history markers.

    This is one of these funny ones when people speak about Linux kernel not having API stability in kernel space. Since fork was exported to Userspace and also was exposed inside kernel space it has remained so even that nothing any more uses it. There is a long list of strange functions you can use from a driver in linux that make absolutely no sense that only remain for API compatibility reasons.

    Old Old drivers before clone exists
    That Exploit Guy I told you prior to sys_clone. That is prior to Kthread.c so why in hell quote that. Was that line not clear enough for you.

    That Exploit Guy
    That’s a DOS extender. In case you are too stupid to read up on the subject, then simply remember this explanation from Wikipedia:
    That was only one example for fork on dos.
    http://www.drdobbs.com/cpp/a-portable-library-for-executing-child-p/184408994
    There are many examples prior where application ported to dos has fork without having. spawn was similar. It was having to implement multi threading on top of Dos that was a true discouragement to having fork. Its not that libraries or Dos extenders did not provide fork on DOS it was just painful. The extender version was the most functional example of fork on dos.

    That Exploit Guy please note DrLoser said not at all. Fork on dos does exist in strange and bad ways.

    Basic CS concepts get you so far. That Exploit Guy. When it comes to implementation of concepts Linux kernel is very limited. The fact the kernel does not care means you can use multi process to emulate multi thread without any major issues. Concept vs Implementation. Concept can say you need to care about something. Real world Implementation can say screw lots of Basic CS concepts as they are not implemented as anything other than virtual.

    That Exploit Guy if we were talking about BSD or Hurd kernels a Process means something to the internals. So you cannot just go screw it.

    POSIX threads notion of a set of threads that share a single PID
    Before Linux 2.4 each thread got its own unique PID. Posix applications don’t like this.
    gettid() returns the ID the Linux kernel is using internally. gettid is not posix but Linux internal. So Linux old PID is now TID. Sharing memory between threads in different processes on Linux is not a problem. Processes are just virtual structures.

  6. That Exploit Guy says:

    Mostly because you are reading the man pages without understand. CLONE_THREAD as you read it you would have notice its going on that POSIX says we require this so-called thread group identifier. Does the Linux schedulers use this information. The answer is no.

    Sigh…
    1) POSIX doesn’t “require” any “thread group identifier”. Thread groups are simply a mechanism to make POSIX threads work. They are simply a mean to an end, in other words.
    2) Whether the scheduler uses “thread group identifiers” has no bearing on what constitutes a thread and what constitutes a process. Boy, that’s some exceptionally weak grasp of basic CS concepts you are showing, isn’t it?

    Availability and form. Just because sys_clone is missing X feature does not mean you cannot use Y feature to achieve the same ends.
    The bottom-line remains that there is no fork() without an MMU, whether you like it or not.

    Linux is ridiculous in places. Do a ps aux and notice all the things in [bla bla] These are all kernel bits running and listed in the process table. Old Old drivers before clone exists in the Linux kernel use fork and listed themselves in the process table.

    Nonsense.

    Fork was desirable to the point that someone implemented for Dos.

    That’s a DOS extender. In case you are too stupid to read up on the subject, then simply remember this explanation from Wikipedia:

    “Essentially, a DOS extender is like a miniature operating system, handling much of the functionality of the underlying operating system itself.”

  7. oiaohm says:

    DrLoser glibc 2.3.3 is still 2003 or over 10 years ago.

    That Exploit Guy sys_clone man page is another trap you walked straight into. Mostly because you are reading the man pages without understand. CLONE_THREAD as you read it you would have notice its going on that POSIX says we require this so-called thread group identifier. Does the Linux schedulers use this information. The answer is no.

    1 word of fear to see in a Linux man page “so-called” this mainly means barely implemented.

    The difference between Processes, Threads to internal Linux kernel schedulers and security mean absolutely nothing. A cgroup entries on the other hand the schedulers and secuirty respond to. So Processes and Threads are just implemented enough to keep Posix coded applications happy. So data structures that do nothing most of the time. Most of the Linux kernel security does not give a rats but instead depends on other assigned information like capabilities and LSM ids.

    This also means that the availability of fork() to user-space programs is affected by how sys_clone is implemented for a given architecture.
    Availability and form. Just because sys_clone is missing X feature does not mean you cannot use Y feature to achieve the same ends.

    That Exploit Guy
    Having fork() available to kernel code is not only inherently useless as an idea but pretty ridiculous as well.
    Linux is ridiculous in places. Do a ps aux and notice all the things in [bla bla] These are all kernel bits running and listed in the process table. Old Old drivers before clone exists in the Linux kernel use fork and listed themselves in the process table. Linux a lot of ways has Micro-kernel strangeness. Hey minix use to fork these things off so will we even that we are now a monolithic kernel and it makes absolutely no sense todo so. That Exploit Guy welcome to strange Linux for a monolithic. Yes the fact Linus started with Minix becomes very clear with some of the oddness of Linux Kernel.

    That Exploit Guy really has no clue about Linux. That Exploit Guy is applying general Unix/Posix stuff. Linux in reality is not general Unix/Posix. Linux is its own beast.

    DrLoser
    It’s also true that DOS lacked these features, and was consequently unable to support forks/joins, even if such a thing had been remotely desirable. Which, at the time, it clearly was not.
    True and False at the same time.
    ftp://i44ftp.info.uni-karlsruhe.de/public_html/sather/sck-old/stud-sck/msdos/disk3/dpmigc3a.txt
    Fork was desirable to the point that someone implemented for Dos. Dos itself did not support fork but enough code on top you can make anything happen.

  8. DrLoser says:

    Fork requires particular features in the OS.

    To be fair, this is not quite #22. Fork does indeed require particular features in the OS.

    To whit, the ability to run more than one process at a time. I believe this ability was present on PDP-8s, back in the late 1960s. Thus, amongst many other state-of-the-art kernel implementations of the time, Unix.

    It’s also true that DOS lacked these features, and was consequently unable to support forks/joins, even if such a thing had been remotely desirable. Which, at the time, it clearly was not.

    Just trying to be helpful.

  9. DrLoser says:

    A diligent researcher reports:

    Arguement about the difference between a thread a process dies with the release of Linux 2.3.3 as well.

    The diligent reporter is not quite correct.

    For the twenty first time on this thread.

  10. DrLoser says:

    To further back Robert up on the principle of ad hominem, by the way, I should point out that the particular individual who is (correctly) being shielded from mere personal abuse has made posts that are almost entirely tangential to the thread of discussion.

    It seems a bit unfair to attack somebody personally when they’re not even talking about Munich, doesn’t it? Remember, this thread is all about Munich.

    With that in mind, Robert, I hope you will permit me (as a once-off) an ad civitatem attack, appropriately toned down and very much relevant to the topic at hand.

    I already know that Munich is going to fail miserably in this ludicrous attempt to botch together some form of unified calendaring and mail. Why? Because the civic IT representatives have proven themselves either completely uninterested or completely incapable in this domain over the last ten years.

    I therefore choose to laugh at their total ineptitude.

    Come December 2015, at which point it will be obvious even to you that I am right, I fondly imagine that you too, Robert, will join me in laughter and ad civitatem abuse.

  11. DrLoser says:

    [AD HOMINEM ATTACK REMOVED – rp]

    I think this is an excellent policy, Robert. I trust you will apply it with Solomonic fairness.

    Now, given these new rules for the site.

    There are at least twenty utterly cretinous comments by the same poster on this thread. Don’t take my word for it, kiddies: go search them out!

    It’s like an Easter Egg Hunt for Goose Eggs, isn’t it?

  12. That Exploit Guy says:

    Since Linux kernel 2.3.3 fork and vfork has not be kernel side except for drivers.

    This is patently false.

    For one thing, kernel driver code is supposed to be part of the kernel. Having fork() available to kernel code is not only inherently useless as an idea but pretty ridiculous as well.

    Also, have you not mistaken glibc 2.3.3 for Linux kernel 2.3.3? 😉

    Clone from Linux kernel is not portable. Yes this is the reality fork is not a Linux kernel function used by most applications.

    Again, sys_clone is just a generalised case for sys_fork and sys_vfork. In fact, you can find all three of them in the include/linux/syscalls.h header.

    LOLOLOLOL to the Linux kernel there is absolutely no difference That Exploit Guy.

    There is. It’s kind of a shame you don’t seem to have a clue about how threads are implemented in the Linux kernel, or, indeed, the fundamental difference between a thread and a process, isn’t it?

    Again, “LOLOLOLOL” perhaps?

    The reason for requiring cgroups in the init system is the fact there is absolutely no difference between a process and a thread to Linux kernel internal logic.

    There is. In case you haven’t paid any attention to the man page for clone(2), then perhaps its time for you to read up on the section regarding the clone_thread flag and the concept of “thread groups”.

    Also, of course, there is the fundamental difference between threads and processes, which till now you still don’t understand. Trust me – it has stuff all to do with the Linux kernel.

    Arguement about the difference between a thread a process dies with the release of Linux 2.3.3 as well. Please note Linux 2.3.3 was release in 1999 so That Exploit Guy is way more than 10 years out of date.

    Again, are you sure you haven’t confused glibc 2.3.3 with Linux kernel 2.3.3?

    PAX set of patches from grsecuirty for Linux do in fact implement NX bit for fork guard.

    Again, are you banking on the fact that no one will try and write a simple fork() bomb and run it on a PAX-patched Linux system?

    Here’s an idea: instead of taking my word for it, why not post a question to the grsecurity forum and let Brad Spengler and co. tell you just how ridiculous your above statement is?

    I bet the outcome will be nothing short of spectacular.

  13. That Exploit Guy says:

    posix_spawn is 2003 posix. Notice that That Exploit Guy claimed it did not exist. So must not have any pthread experience post 2003. So is 10 years out of date.

    This is posix_spawn(). Notice that there is still no such thing as “pthread_spawn”. Notice also that posix_spawn() has fundamentally stuff all to do with POSIX threads.

    What can I say? “LOLOLOLOLOL”?

    posix_spawn and sys_clone and a few other methods are used to implement fork and vfork in the different libc that Linux has.

    This is clone(). It is indeed true (and my mistake) that under glibc, fork() and vfork() are just wrappers around the Linux-specific sys_clone function. This is because sys_clone is the kernel function that does everything related to creating processes. In other words, it’s a generalised function of fork() and vfork(). This also means that the availability of fork() to user-space programs is affected by how sys_clone is implemented for a given architecture.

  14. That Exploit Guy says:

    Nope.

    I like how you have completely ignore me pointing out the fact you are never above launching ad hominem attacks at others.

    Please, save this sort of BS hypocrisy for someone gullible, Robert Pogson.

  15. oiaohm says:

    posix_spawn is 2003 posix. Notice that That Exploit Guy claimed it did not exist. So must not have any pthread experience post 2003. So is 10 years out of date.

    The introduction of posix_spawn was to allow about 50 percent of all use case of fork to be replaced with it alone. 100 percent coverage is possible with some handling..

    DrLoser glibc uses sys_clone() to imperilment fork and vfork this is why tracing programs on Linux can get confusing. sys_fork() and sys_vfork() in Linux is rarely called. In fact glibc implements vfork using sys_clone as well. Basically Linux kernel could remove sys_fork and sys_vfork from kernel space with 99.9 percent of applications would not even notice.

    This is the big trap. fork and vfork as syscalls are fairly much not called by applications but instead resolved by the libc in use. Linux kernel redirection back to the libc is a bug fix. posix_spawn and sys_clone and a few other methods are used to implement fork and vfork in the different libc that Linux has. By the way posix_spawn has wrapper functions.

    Fork requires particular features in the OS. Fork is simple to make work in a MMU designed kernel.

    And you will be “shock horror” to find that no libc under Linux provides fork(). Again, the reason of this lies in the difference between a thread and a process, which you don’t understand.

    LOLOLOLOL That Exploit Guy
    http://blog.andrewvc.com/fork2-is-dead-long-live-fork-i-mean-clone2
    Since Linux kernel 2.3.3 fork and vfork has not be kernel side except for drivers. Its the minimum libc that provides fork and vfork to user-space applications. Sorry libc under Linux provides fork and vfork.

    Clone from Linux kernel is not portable. Yes this is the reality fork is not a Linux kernel function used by most applications. Thread and a process if you are talking about Linux fork is not the words to use. Clone is the syscall. Clone of all things is used to implement posix_spawn as well.

    Again, the reason of this lies in the difference between a thread and a process, which you don’t understand.
    LOLOLOLOL to the Linux kernel there is absolutely no difference That Exploit Guy.
    The reason for requiring cgroups in the init system is the fact there is absolutely no difference between a process and a thread to Linux kernel internal logic. Arguement about the difference between a thread a process dies with the release of Linux 2.3.3 as well. Please note Linux 2.3.3 was release in 1999 so That Exploit Guy is way more than 10 years out of date.

    The idiot here was That Exploit Guy raising fork as so called idea.

    That Exploit Guy of course is a idiot I did not say that Linux uses NX bit to prevent fork bomb. Instead has to half quote then say test it on Linux. It was VMS usage of the predecessor to the NX bit to prevent fork bombs. PAX set of patches from grsecuirty for Linux do in fact implement NX bit for fork guard. After so many forks flag the memory. For someone who claims to be a exploit guy not to be aware that custom patched Linux kernels for security do in fact include this feature is a laugh.

  16. TEG wrote, “It is obvious that the sudden censorship is simply the result of him being on the losing side of the argument rather than any moral objection.”

    Nope. It just occurred to me that the ad hominem crap was getting fast and furious around here and I coded the sequence into AutoKey so I just select the crap and replace it. Maybe folks would stick to the topic rather than criticizing people for who they are.

  17. That Exploit Guy says:

    Not unexpected when you are on the losing end of an argument.

    “Ad hominem” is what Robert Pogson does pretty much every time before he quotes one of the “trolls”. It is obvious that the sudden censorship is simply the result of him being on the losing side of the argument rather than any moral objection.

  18. That Exploit Guy says:

    Yes they were. (Robert Pogson then proceeded to babble about his personal “experience” that had nothing to do with Extremadura)

    This is all very typical, isn’t it. Someone point out the misleading nature of your claim, and then you either ignore the criticism or simply blather about something completely irrelevant in order to distract readers from it.

    Again, as olderman has already pointed out, you have done nothing near the scale of deploying 10k machines. PXE boot is just technical minutiae when it comes to a project this big. Your “experience” simply means stuff all to this discussion.

    Leave your fanciful imagination to fooling some clueless school principles. Nobody here is buying any of it.

  19. Olderman says:

    Censorship Robert pogson?

    Not unexpected when you are on the losing end of an argument.

  20. DrLoser says:

    These lead the developers of 9x to think it was possible todo stacks of things without using the MMU code because coding examples without MMU was in univerity binds. Yes the gates historic university bin raids.

    I counted too soon. [AD HOMINEM ATTACK REMOVED – rp]

    This one is fabulous. I very definitely regard it as the [AD HOMINEM ATTACK REMOVED – rp] . (And that involves one heck of a lot of competition.)

    Let us, as TEG points out, pause for a moment and drink in what this [AD HOMINEM ATTACK REMOVED – rp] implies.

    1) Bill Gates was personally responsible for scads of memory management code in Windows 9x.
    2) Bill Gates, what with this monumental responsibility (and what with having that other Day Job being a Monopolist Conspirator, (c) Pogson), was a bit pushed for time.
    3) … and this is actually the least preposterous insinuation … Bill Gates therefore went back to the piles of line-printer output that he had dumpster-dived from Harvard in Nineteen Seventy Four and lovingly preserved, just in case, and thoughtlessly cargo-culted every last idea on memory management.
    4) And nobody else at Microsoft noticed.
    5) And despite that, Windows 9x had MMU baked in from the very start.

    OK, fine. Enough of oiaohm. I need to drop several large tabs of LSD-25.

    They might help stop me hallucinating.

  21. DrLoser says:

    posix_spawn, btw. Wouldn’t like to deny oiaohm his customary cheeky little diet of nutritious seeds, would we?

  22. DrLoser says:

    Did I say thread? I meant process. No wait, I meant thread.

    I’m hopelessly confused. Perhaps HIMEM.SYS can help me out?

  23. DrLoser says:

    So the libc is more than able to implement fork …

    That is Idiocy Number Nineteen on this thread, oiaohm.

    You can either give up now, or go for a new personal record. I’m absolutely certain you have it within you to take the second course of action.

  24. DrLoser says:

    There is no such thing called “pthread_spawn” in POSIX threads.

    Oh dear, that was really naughty of me, wasn’t it? I set a trap, and oiaohm plunged into it. Like a rock. Or a moose-gobbling bullet. Or an absurd parabolic analogy. Hey, this is Robert’s site: whatever tom-fool analogy Robert is currently enamoured of is good enough for the rest of us.

    You’d sorta think that an expert witness like oiaohm would (a) either know ahead of time or (b) take the minimal trouble to look up posix_spawn(), wouldn’t you?

    What with it being immediately relevant to both vfork() and POSIX and all.

    [AD HOMINEM ATTACK REMOVED – rp]

  25. TEG wrote, “none of those machines were set up over a weekend or in one place.”

    Yes they were. Typically, a BIOS can be tweaked to boot PXE in a minute or so. I remember opening many cartons and doing that to them in a couple of hours. The biggest problem was what to do with the pile of fluffy debris, a fire-hazard… The guys at Extremadura had machines already working with that other OS in place so they had to walk around their buildings once doing that. It can be done in a weekend. It took me two days with a helper to walk around and distribute and plug in and test my units. The work I planned to do in a month (build terminal server cluster, multi-seat clusters, wire 4 racks, and set up thin and thick clients and 700 accounts) was done in 10 days of long hours because the construction was not complete on time.

    At another school, with a lab that had 17 machines running XP on a good day, I was able to walk around collecting broken (ie that other OS would not boot) machines from storage, installing a terminal server and getting 24 machines working in a lab in 3 hours each with a new bootloader (some machines had no on-board PXE). There were 4 eras of machines all working well with the same boot-image of GNU/Linux except a few needed custom video set-ups. I used to do hours of work every week in that lab to get XP working again, but with GNU/Linux, that lab hummed for months with not a single request for service from teachers. I felt lonely…

    So, I’ve several times been involved with projects per-building that went more or less like the Extremadura migration. All it takes is a few people in each building to make such quick migrations happen, less if the machines already boot PXE and/or can be remotely provisioned. In my case I used LTSP with an OpenSSH server in the image. I used that capability to wipe drives of XP and to install boot-loaders with a single command.
    for f in list-of-PCs;do ssh $f “dd if=/dev/zero of=/dev/sda bs=1024k;dd if=boot-loader of=/dev/sda”&done
    Then, all I have to do is boot the machines once any way I can, BIOS tweak, boot-menu or floppy, and I can set diverse machines to boot from the terminal server and give a bunch of users at once the power of a new machine with new software, huge shared resources, and minimal network-lag.

  26. That Exploit Guy says:

    DrLoser your suppose here is kinda correct. pthread_spawn…

    There is no such thing called “pthread_spawn” in POSIX threads.

    … plus wrappers on a stack of c functions is required to make the fork behave as per posix standards.

    Everyone with the most basic knowledge in computer science understands the difference between a thread and a process. You on the other hand, have posted comments after comments indicating rather clearly that you have absolutely no idea how to differentiate between the two.

    Yes using pthread_spawn in the libc is possible.

    Again, there is no such thing called “pthread_spawn” in POSIX threads.

    So the libc is more than able to implement fork and libc is able to trace particular operations that need memory cloning.

    fork() is quite simply not a function provided by libc. Knowing what provides a given function is the prerequisite of being able to use it in the first place.

    So there is a huge reason why MMU less fork is in the libc not the kernel as it the only place posix fork can be implemented in a MMU less environment is in the libc.

    And you will be “shock horror” to find that no libc under Linux provides fork(). Again, the reason of this lies in the difference between a thread and a process, which you don’t understand.

    Yes there are historic Posix systems without a MMU.

    There have been licensed Unix systems targeting platforms that come without MMU. Of course, this means fork() needs to be implemented in a way that is workable without the concept of a page table.

    The answer is not threads.

    Rather, there is an obvious, fool-proof way to achieve fork() that is guaranteed to work with or without an MMU. In fact, there is at least one famous Unix(-like) system implementing fork() in such a fashion. I’ll let you figure out what that Unix(-like) system is and how it implements fork().

    These lead the developers of 9x to think it was possible todo stacks of things without using the MMU code because coding examples without MMU was in univerity binds. Yes the gates historic university bin raids.

    Yes in an alternative timeline where Gates is a dumpster-diver rather than a billionaire. 😉
    I don’t know how you have managed to come up with such ridiculous lies, but I bet you will be “shock horror” by just how long the concept of MMU predates Windows 95 and how long it has been taught in universities around the world.

    DrLoser vfork cannot be substituted for fork. There posix requirements on fork and vfork that in fact make them completely incompatible functions.

    The difference between fork() and vfork() simply does not matter when the function are used in conjunction with execve(). In fact, fork/exec is pretty much the only reason that vfork() exists.

    It was to prevent fork in fork. You fork once you mark the fork location no execuite in the child and the child cannot fork that way again. Fork bomb limitation.

    I won’t attempt to debunk this utterly ridiculous statement. Let’s just say it’s obvious that you have not even at least tried and implemented a fork() bomb on a Linux system that has NX bit enabled.

    Those who have tried simply know from experience that those above words of yours do not even have a leg to stand on.

    DrLoser just to be really strange 9x will run without it vxd memory manager.

    We mere mortals simply call that “MS-DOS”.
    Trust me – there is nothing strange about it.

  27. That Exploit Guy says:

    The beauty of FLOSS is that all the problems are solvable, so if you really want to hurry there is a way to do it over the weekend. Extremadura did that.

    And none of those machines were set up over a weekend or in one place.

    Misleading much?

  28. ram says:

    The movie industry routinely sets up hundreds of Linux workstations for animators, editors, and the like when a new movie project gets the go ahead (i.e. gets fully funded). Usually most of it is up an running in under a week. Some extreme high end workstations take longer due to hardware parts availability, but they are usually done a week later.

  29. DrLoser, thinking inside the box, wrote, “It takes a minimum of nine months, even under commercial pressure, to deliver anything like this.”

    The beauty of FLOSS is that all the problems are solvable, so if you really want to hurry there is a way to do it over the weekend. Extremadura did that. 40K seats on a weekend.
    “We changed the desktop systems from Windows to Linux during the weekend, when the civil servants came back next Monday morning they found Linux running on their desktop machines”

    It’s not that hard. The software works. They have some inventory. They have networks, storage and servers. It’s not that hard. Because it’s client-server stuff they can play around on the server without disrupting workflow. They can distribute clients in a small test. Develop training/information. Roll out all clients. If you take more than a few months to do that you are doing something wrong. I built and deployed a whole IT-system for >100 seats, including many web-applications in about 1 man-month of work, including building the damned servers, wiring the racks and connecting it all up in a huge building. Clone me in every building and it’s done. In this case, they already have a working IT system and they only need to add processes to the servers and clients. Trivial. Done by lunchtime.

  30. DrLoser wrote, “And Kolab since 2006 has been successfully deployed by which large German governmental organisations”.

    The schools in Basel (not exactly German but German-speaking and governmental…):“the installation is home to almost 40,000 users. Entire school classes, teachers and administrative staff work together on Kolab. Any student who is currently enrolled in Basel receives, directly at enrollment a Kolab account and can begin to work with the system.”

    The folks who originally commissioned Kolab still use it. See Revamping the first open source groupware solution

    “Living up to the principle of practicing what they helped define and preach, they continue to use Kolab on 500 Linux desktops as they have for years. And they continue to do important work in open source that will continue to benefit the IT security of the German government as well as users around the world.”

    In, 2008, Software migration guide 1st edition Berlin, April 2008 mentions Kolab as if it were widely used and gives this example:“Brandenburgischer Landesbetrieb BLB uses the Kolab groupware
    system for more than 600 computer workplaces.”

    They spend a lot of time going on about Kolab in the 2012 Migration guide. It hasn’t gone away…

  31. DrLoser says:

    Number 18 in a continuing series of idiocies:

    NX bit is critical to fork security. The reason why NX bit history ancestor was added to Alpha MMU for VMS is fork issue prevention. Please note the OS VMS the direct OS the lead developer of NT comes from. It was to prevent fork in fork.

    “It was to prevent fork in fork????”

    I’d like a second opinion from TEG or DeafSpy, but I believe this one Wins The Thread so far.

  32. DrLoser says:

    Kolab was in fact designed to meet requirements of the German Government IT security policy.

    And Kolab since 2006 has been successfully deployed by which large German governmental organisations, oiaohm?

    Do your own work for a refreshing change. I’m guessing none over the level of, say, 500 seats.

  33. DrLoser says:

    Do you really think that Kolab will not see that a working product is delivered on time and on budget to such a high-profile customer?

    Do you really think that they are capable?

    If you look at their release pattern (big spike in 2009, small blip in 2013), and you consider the fact that they had ten whole years from 2004 to prepare for this momentous leap into the as-yet unknown, the smart money says not a chance in Hell.

    Platitudes and hopes and wishes don’t work in this domain, Robert. Hard-Nosed Large Project Experience does.

    I fully respect your astonishing achievements in kitting out several schools with a thin-client Linux solution over CAT-3. That is something that nobody should take away from you.

    But you’ve never been anywhere near a project of this size with this sort of pressurized deadline.

    I’ve worked in Airline Res and in Credit Card Authorization and in both cases we’re talking about a minimum of the Munich established inventories of 10,000 seats.

    It takes a minimum of nine months, even under commercial pressure, to deliver anything like this.

    And that would be starting with an established base of software — a base established over ten or twenty years. Not starting with an established dung-heap of rubbish that nobody much else thinks is worth trying to use, which is the case with Kolab.

    If you start with an established base (res, auth, etc) and you need one small incremental change, let’s say a single protocol that slipped the requirements net …

    … You are Tits Up.

    Not a chance, Robert. Not a chance in Hell.

    See you on the other side of 2015.

  34. oiaohm says:

    DrLoser Microsoft exchange did not pass German government requirements for security until 2006. Kolab was in fact designed to meet requirements of the German Government IT security policy.

    The reality here is if Munich was running exchange they would have been committing an breach of regulation. This is the shock horror its quite common for government parts not to be running share-point or exchange due to failures to be certified.

    Alfresco and Kolab have quite a home in Germany. DrLoser the complete globe is not flat.

    DrLoser
    Anyhow … that corresponds exactly to my point that vfork() is the MMU-less equivalent of fork() for one very important use-case: that is, spawning a new process. IMHO, all of the other possible use-cases for fork() are antediluvian and futile in any case, although I supposed one might substitute pthread_spawn if one felt like being “Posix 2008 compliant.”
    DrLoser your suppose here is kinda correct. pthread_spawn plus wrappers on a stack of c functions is required to make the fork behave as per posix standards. So this cannot be implemented in the kernel.

    Little issue uclinux kernel provides Pthread features in its threading. Yes using pthread_spawn in the libc is possible. So the libc is more than able to implement fork and libc is able to trace particular operations that need memory cloning. Kernel seeing the syscalls after libc cannot see the details. So there is a huge reason why MMU less fork is in the libc not the kernel as it the only place posix fork can be implemented in a MMU less environment is in the libc. Sections of 9x threading are also outside its kernel space. Remember Pthread userspace also tracks to really early posix systems without MMU. Yes there are historic Posix systems without a MMU. These lead the developers of 9x to think it was possible todo stacks of things without using the MMU code because coding examples without MMU was in univerity binds. Yes the gates historic university bin raids.

    DrLoser vfork cannot be substituted for fork. There posix requirements on fork and vfork that in fact make them completely incompatible functions.

    NX bit is critical to fork security. The reason why NX bit history ancestor was added to Alpha MMU for VMS is fork issue prevention. Please note the OS VMS the direct OS the lead developer of NT comes from. It was to prevent fork in fork. You fork once you mark the fork location no execuite in the child and the child cannot fork that way again. Fork bomb limitation. The name Execute Protection sound strange as that is what it was designed todo at first just happened to provide extra benefits if used on Data. Yes the history Unix world fork bomb disaster limitation. VMS had NX equal protection before even the first line of code was written on Windows NT.

    Also Alpha chips NX equal is inversed. The execute bit has to be on for it to run a section on a Alpha. Execute Protection is the slight different in operation. Yes when you are referring to a Alpha its name is not DEP.

    This is knowing the history where the stuff comes from.

    The reality when went security backwards with the selection to use x86 and we have suffered for it.

    DrLoser just to be really strange 9x will run without it vxd memory manager. People who attempt to run 9x in in the smallest amount of memory possible found this one out. Yes HIMEM is mandatory for 9x to operate but 9x own memory manager is only required if you wish to use more than 64Meg. Please note when Microsoft made HIMEM and Microsoft decided to cripple it. It also like 9x 700 something meg limit in ram. NT full 4G of ram support. If 9x was using the MMU properly its limitations should match those of the MMU or exceed. There are many things about 9x that are just wrong.

  35. DrLoser scares easily.

    “Kolab Enterprise combines email, calendar, contacts, tasks, file storage, data sharing and more, into one convenient package built atop a secure architecture designed to protect your customers’ data. This complete solution, not only integrates desktop and mobile apps, but also the web on top of Roundcube, the world’s favorite webmail client, used on more than 500,000 sites by millions of users daily.”

    Do you really think that Kolab will not see that a working product is delivered on time and on budget to such a high-profile customer? Kolab will see to the integration, migration and service of the product.

  36. DrLoser wrote, “Anybody care to explain?”

    A timeline might be useful:

    1. April 2012, Wollmux 11.10a was released.
    2. February 2013, LibreOffice 4.0 was released

    So, the “effort” lacking may just be an inability to see the future. I wouldn’t blame anyone for that. Wollmux works just fine for what they are using, LibreOffice 3.x.

    You can download Wollmux here. It’s Java, neither ELF nor .exe code.

  37. DrLoser says:

    The good news from the WollMux Monkeys is:

    27. März 2014

    Hinweis für Benutzer von LibreOffice 4.x: Die zuletzt veröffentlichte Version 11.10a funktioniert nicht zusammen mit LibreOffice ab Version 4.x.

    Yes, that’s the good news. Allow me to translate for you:

    “Note for users of LibreOffice 4.x: the most recently released version 11.10a will not work with LibreOffice version 4.x.”

    Well, no problem there; it’s always nice to have advance notice of these things. And the drones can always rely on LibreOffice version 3.x. Possibly even that nasty Apache thing.

    But here’s the bad news:

    Both 11.10a and 11.11a have Windows downloads from this official site.

    Which means that these monkeys are actually putting more work in to ensure that WollMux works with Windows (of all things) than they are ensuring that it works with LiMux.

    Go on, I’m fascinated. Anybody care to explain?

  38. DrLoser says:

    Kontact is scary to install on windows as you watch 2G of harddrive space disappear.

    Robert will be overjoyed: in 2009, they reduced this “bloat” to a mere 995MB. (As evidenced on my machine.) The scary thing is, it’s still worthless garbage.

    I’ll take a step further than that.

    The really scary thing is that the latest Windows-huge-debug version is a broken link.

    (Even scarier is the hilarious “installation documentation.”)

    Let me point this out: this piece of garbage is dated 2013-06-12.

    And the WollMux guys, whose main “technical” expertise (they’ve proven themselves adept at ten whole years of lying to the procurement department) is, ahem, fiddling around with ODT macros …

    … have promised the incoming Mayoral organisation that they will deliver a “unified mail and calendaring system” based upon Kolab and Kontact …

    some time in 2015.

    This just ain’t gonna happen, is it?

  39. DrLoser says:

    As always, oiaohm charges to the rescue. I forgot this little beauty:

    DrLoser unified mail and calendar application when it comes to kolab is the application called Kontact. Kontact predated outlook. Kontact is scary to install on windows as you watch 2G of harddrive space disappear. Kolab was used in Munich when they were still running NT 4.0 by its Web interface. Munich was never an exchange environment.

    Naturally, an enormous pile of bull. But what’s interesting to me here is

    1) Kolab was used in Munich when they were still running NT 4.0 by its Web interface.

    Really? OK, well, no cite, because oiaohm only links to things he understands in the first three sentences, but interesting nonetheless. Although not necessarily in a positive way, vis-à-vis the Great Migration.

    2) Munich was never an exchange environment.

    Less interesting, and still with no cite as evidence. I take that back: more interesting.

    If this is actually true (I am of the opinion that it is standard oiaohm Walter Mitty fare), then it’s very interesting, because Munich was probably the only Municipality over 1,000,000 in the Western World that didn’t use Exchange in 2004.

    Which would obviously make it easier for a bunch of con artists like the WollMux Kollektiv to move in and spend ten years painting the municipal walls with faeces.

    And if it’s false, you have to wonder why said con artists spent ten years not replicating existing, essential, functionality.

    Ah, questions, questions.

    And no answers at all from you lot.

  40. DrLoser says:

    DrLoser wasted lots of bytes getting to “3. Important in the highest degree; indispensable to the attainment of an object; indispensably necessary. “

    I’ve shown that Exchange is not indispensible in the running of organizations large and small by any definition.

    In that your specious quote entirely avoided mentioning this definition, it’s hard to see how you’ve shown anything at all, Robert. Other than a capacity for refusing to accept other people’s arguments even as arguments, let alone as winning arguments.

  41. DrLoser says:

    Unfortunately, Robert Pogson, Given your limited experience with modern IT as it is used in large institutions outside the Arctic Circle, that is not a statement that you are qualified to make.

    Despite Robert’s obvious falterings when confronted with polysyllabic terminology that causes him to reach for a 1913 dictionary, he is, actually, qualified to make this statement.

    Robert is correct in stating that a substantial portion of Windows desktop users do not require, nor benefit from, Exchange. If you spend your life as an ATM, for example, you are more likely to use the several other amenities of the Operating System of Choice for Billions … but you won’t use Exchange.

    If you’re a Mom’n’Pop store, you probably don’t use Exchange. You might. But it’s not an absolute essential.

    I’ll back Robert even further. I have Exchange on my machine. I use the mail system; my main interaction with the calendaring mode is to hit “Enter” every time some idiot meeting reminder comes up.

    But then I’m a programmer, which basically puts me in the “Pogson category.” Exchange is not essential to Robert or to me.

    It’s pretty bloody essential to my management, otoh. And to management everywhere. And not just to people managers, but to logistics managers, appointment managers, all sorts of people. Managers of a Municipal Authority such as Munich to take a notable and relevant example.

    Why? Precisely because it obviates the need for a secretary with a Mercedes and a bullhorn to drive around town shouting at people.

    I made this point quite forcefully to Robert lower down, when Robert was all for the dolly-bird-in-a-Merc option. Robert sailed serenely by without answering.

    I suspect that Robert is fundamentally incapable of putting himself in the shoes of other professionals, which would be a problem in this case.

  42. DrLoser wasted lots of bytes getting to “3. Important in the highest degree; indispensable to the attainment of an object; indispensably necessary. “

    I’ve shown that Exchange is not indispensible in the running of organizations large and small by any definition. How can half the businesses on Earth dispense with something that’s indispensible?

  43. DrLoser says:

    Define essential: “Belonging to the essence, or that which makes an object, or class of objects, what it is. [1913 Webster]”

    Back to “refutation by Webster’s 1913,” I see, Robert. Apparently my spate of sourcing this somewhat antiquated reference, and I did so with intentional irony, doesn’t seem to have put you off.

    Nor has your insistence upon academic rigour in citing a source prevented you from a sophomore error in this case, has it? You don’t get to pick the single definition that helps you deflect what might otherwise be valid criticism, you know. When citing a an authority, you should always be mindful of the whole, not merely a convenient part.

    So here we go again, only this time with academic rigour:

    Es*sen”tial (?), a. [Cf. F. essentiel. See Essence.]

    1. Belonging to the essence, or that which makes an object, or class of objects, what it is.

    Majestic as the voice sometimes became, there was forever in it an essential character of plaintiveness. Hawthorne.

    2. Hence, really existing; existent.

    Is it true, that thou art but a a name, And no essential thing? Webster (1623).

    3. Important in the highest degree; indispensable to the attainment of an object; indispensably necessary.

    Judgment’s more essential to a general Than courage. Denham.

    How to live? — that is the essential question for us. H. Spencer.

    4. Containing the essence or characteristic portion of a substance, as of a plant; highly rectified; pure; hence, unmixed; as, an essential oil. Mine own essential horror.” Ford. [511]

    5. (Mus.) Necessary; indispensable; — said of those tones which constitute a chord, in distinction from ornamental or passing tones.

    6. (Med.) Idiopathic; independent of other diseases.

    Definitions 1, 2, 4 and possibly 6 are based upon the root meaning, “of the essence.” olderman is clearly using the term under the third definition (and, being a musical chap, may be alluding to the — minor, ho ho — fifth as a metaphor).

    But you knew this, didn’t you? So what was your intent in using Webster’s to deflect olderman’s meaning?

    Because, whatever it was, it was neither academically rigorous nor, indeed, successful.

  44. olderman wrote, ““Even today, when some call Exchange essential it is not.”
    Unfortunately, Robert Pogson, Given your limited experience with modern IT as it is used in large institutions outside the Arctic Circle, that is not a statement that you are qualified to make.”

    If olderman doesn’t respect my experience/knowledge/wisdom, perhaps he should read the market analysis:
    “In 2011, Microsoft Exchange Server will have a worldwide installed base of 360 million mailboxes. On-Premises Microsoft Exchange Server deployments currently account for the majority of Microsoft Exchange Server mailboxes worldwide, representing 76% of the worldwide installed base in 2011. We expect this to gradually decrease as the popularity of hosted, cloud-based solutions increases.”

    That when there were ~1000 million users of PC IT and probably ~500million in business. That means 140million were not using something called “essential”.

    Define essential: “Belonging to the essence, or that which makes an object, or class of objects, what it is. [1913 Webster]”

    Today’s analysis:“Combined, Microsoft Office 365 and Exchange Server currently account for 51% of worldwide mailboxes in the Business Email and Collaboration market. Cloud-based Business mailboxes offered through Microsoft Office 365 and Microsoft Hosting Partners currently account for 15% of worldwide Microsoft Exchange Server mailboxes, while On-Premises mailboxes still represent an 85% market share.” So, Exchange is not essential for 49% of business-use…

    Businesses don’t exist to buy M$’s products/services. They exist to make money, usually, and they do it the best way they know how.

    Now, e-mail, collaboration and scheduling may well be essential to most businesses but nothing M$ supplies is essential to doing that. I do that and I don’t use any of M$’s stuff, except, perhaps, a few random fonts.

    Another example, Largo FL:“”Within budget” is perhaps an understatement. When Largo looked into the conventional solution employing Microsoft’s Exchange server and Outlook clients, total hardware and software costs were “going to be somewhere between 400 and 500 thousand dollars.” By comparison, Schomaker explained, “the total cost to implement the Bynari solution we calculated to be about 95 thousand dollars.” “ They now use Zimbra.

    Exchange wasn’t essential for them, either, although collaboration and e-mail were.

    Oh, since when is the experience of a guy who has been working with IT since 1968 “limited”? I’d bet I’ve worked on more different types of hardware and software than the typical consumer or many younger IT-guys. I’ve worked on computers with germanium and silicon transistors, magnetic, static and dynamic RAM, hard drives that shook the room, mainframes, mini-computers, micro-computers, just about every kind of PC and I’ve programmed most of them in a variety of languages for many purposes. My last career-path before retirement happened to be in northern schools but many of them had more IT per student than many schools in the south thanks to GNU/Linux and FLOSS. I’ve also taught in schools in the city with 1400 students and no IT outside the office/computer lab.

  45. That Exploit Guy says:

    I don’t think I ever owned a 386.

    I did. So did Linus Torvalds. In fact, 386DX was his CPU of choice for developing the OS that you now preach like it’s a religion.

  46. olderman says:

    “Even today, when some call Exchange essential it is not.”

    Unfortunately, Robert Pogson, Given your limited experience with modern IT as it is used in large institutions outside the Arctic Circle, that is not a statement that you are qualified to make.

  47. That Exploit Guy says:

    I don’t think I ever owned a 386. I jumped directly to 486. If either had MMUs they were extremely crude.

    Really, care to enlighten me on the difference between a “crude” MMU and a non-“crude” MMU?

    After all, Gates thought 640K was enough for M$ and all its partners.

    You must be referring to a famous quote that Bill Gates has been known to not have said.

    Boy, you really are a sucker for BS, aren’t you?

  48. That Exploit Guy says:

    Even today, when some call Exchange essential it is not.

    Of course not. Just party like you are in 1995 and you are golden.

    While you are at it, why not bring back the brick phone?

  49. Deaf Spy wrote, “I still see absolutely no excuse why the requirements analysis missed that very important requirement: unified e-mail and calendar application.”

    They were running NT4 back in the day. That was released in 1996 before Exchange became “essential”. So, it was not essential when they decided to migrate back in 2003. M$, itself, did not convert all employees to Exchange until 1996. They were eating their own dogfood but Munich was not. Even today, when some call Exchange essential it is not.

  50. TEG wrote, “one can easily see the reason why staying on the “Wintel treadmill” rather than forgoing essential productivity software would have been a much better idea for the City of Munich.”

    They did not forgo any essential software. They even retained a few machines running that other OS to do so, but no application required every seat to be running that other OS. Munich found they had applications they were not using, applications that were duplicated, and whittled them down. That was sound management, not a loss of productivity.

  51. That Exploit Guy says:

    Assuming that were true, and they wasted $millions less than they would have with staying on the Wintel treadmill…

    They could have also “wasted” an entire IT budget less by going from computers to stone tablets and chisels, by the way.
    You don’t need to be a genius to figure out why this would not have been a good idea to save money.
    Likewise, one can easily see the reason why staying on the “Wintel treadmill” rather than forgoing essential productivity software would have been a much better idea for the City of Munich.

  52. Deaf Spy says:

    Back to Munich.

    I still see absolutely no excuse why the requirements analysis missed that very important requirement: unified e-mail and calendar application. This is an application, which stays open throughout the whole working day. Very hard to miss.

    I suspect the miss is just because some of the little minds, doing the analysis, decided “Hm, we don’t have it, but that will not stop us now! You don’t need that!”.

    Pathetic, don’t you think?

  53. Deaf Spy says:

    After all, Gates thought 640K was enough for M$ and all its partners.
    Rather ignorant, Pogson. You should have known already that this is an architectural decision, taken by IBM when designing the PC. The remaining 384K (not the memory, mind you – there addresses) were necessary for the software to able to communicate with hardware (BIOS shadow copy, video memory area, DMA channels (basically all devices). MS and Gates had absolutely nothing to do with this decision.

    Btw, this is the reason why a 32-bit OS on x86 cannot use more than 3 – 3.5 GB of RAM, even if you PC is equipped with 4GB (of course, without technologies like PAE / AWE). Again, the upper addresses are reserved for hardware, and the amount you are going to lose depends on the hardware itself.

    And since I am an academic person (I do hold a PhD), here are two references for you:
    http://en.wikipedia.org/wiki/Conventional_memory#640_KB_barrier
    http://en.wikipedia.org/wiki/3_GB_barrier

  54. DrLoser wrote, of Munich, “Everything else, for the last ten years or so, was a complete waste of money on a basket full of gibbering useless monkeys who quite frankly couldn’t code their way out of a wet Liverpool Echo.”

    Assuming that were true, and they wasted $millions less than they would have with staying on the Wintel treadmill, DrLoser would really have been irate if they had stayed on the Wintel treadmill.

  55. DrLoser says:

    It’s really, really hard to understand any merit to comments beating on someone for the way they communicate. That won’t change oiaohm. It won’t do DrLoser any good. It won’t do anyone else any good either. It’s just DrLoser being a bore.

    Merely summarising seventeen separate instances of oiaohm being spectacularly technically ignorant, Robert. The information is all there, further down on this interminable thread, should anybody choose to read it. I see no reason why a summary should be objectionable.

    It’s not like a single one of those comments had any relevance to a defence of Munich, is it?

    And, were you to be even-handed, you would treat oiaohm, who routinely misrepresents *nix, in precisely the way that you would treat another commenter who routinely misrepresents M$, wouldn’t you?

    Perhaps not, Robert.

    Perhaps not. I wonder why?

  56. DrLoser wrote, “It’s really, really hard to pick the most stupid and comprehensively uninformed comment out of the above.”

    It’s really, really hard to understand any merit to comments beating on someone for the way they communicate. That won’t change oiaohm. It won’t do DrLoser any good. It won’t do anyone else any good either. It’s just DrLoser being a bore. Besides, many of oiaohm’s comments are not stupid but quite rational. It’s DrLoser who keeps going off on weird tangents trying to deflect the topic of the day from its natural flow.

    As everyone knows software/hardware are quite blurred. Intel’s CPUs are microcoded. So are many hardware devices. That’s what a lot of “firmware” is about. It doesn’t matter whether memory-mapping is done in some C-structure or a chip on the motherboard. Really, it doesn’t matter to any of us apart from the inefficiency involved in the Intel system over the years, wasting silicon and board-space.

    I don’t think I ever owned a 386. I jumped directly to 486. If either had MMUs they were extremely crude. After all, Gates thought 640K was enough for M$ and all its partners.

    Why waste time on such ancient history long since done and only clogging up M$’s backwards-compatibility? Can’t we move forward, instead, to a world where people can use their hardware to its maximum capability unrestricted by a EULA from Hell?

  57. DrLoser says:

    Basically, Robert, the “Big Win” that Munich achieved between 2004 and 2014 was WollMux, wasn’t it? Ironically it’s the only bit that anybody can download, although I have to admit that it’s a quantum leap forward in technology.

    If you define “technology” as “we used to have a complete mess of Word and Excel templates in 2004, but now we don’t, and what’s more we use OpenOffice, no, wait, LibreOffice to do it.”

    I cannot begin to express my admiration of this astonishing achievement. Only ten years!*

    Of course, a proper analysis of requirements would have taken about a couple of months, plus say a year and a half (we’re talking government bureaucracy here) to get to the same point. Quite frankly, the same point even if you adopt Open, oops LibreOffice on Windows.

    Everything else, for the last ten years or so, was a complete waste of money on a basket full of gibbering useless monkeys who quite frankly couldn’t code their way out of a wet Liverpool Echo.

    * The download site is at pains to mention that 11.10a works for Windows.

    Luckily, Linux is at least option number 3. A vote of (relative) confidence, I think.

  58. DrLoser says:

    And while I wait for moderation on a post that contains a very large number of links (ironically to Mr Pogson’s own site — Wikipedia would approve of this degree of due diligence), a final comment from oiaohm:

    By the way its super tricky to at times understand what works.

    Evidently so, oiaohm: evidently so.

  59. DrLoser says:

    Dr. Loser, are you sure this is not better than the statement about HIMEM.SYS allocating the first page?

    I don’t know; we could take a vote on it, I suppose. From the bottom:

    1) large section of the computers at Munich are over 10 years old

    2) LiMux and Goobuntu are not different to a internally Windows Image for Company usage

    3) Old machines can have failing hard drives without being a major bother to a Linux deployment

    4) you will have a shorter operational life until you find yourself with no drivers for the latest Windows OS sooner than you will find yourself without drivers for Linux. (re Apache, Libre etc on Windows)

    5)by this [POODLE] should we not presume that Microsoft own proved SSL is in worse disrepair than OpenSSL.

    6)If you are lucky you get a Myob Error 9004 that it has detected damage to the system files it depends on. (As TEG points out, this is a JRE error).

    7)NT leaks were a little more creative I wish I could remember one of the number. (This whole post is a mountain of bullshit from beginning to end.)

    8)Shared Flat memory mode is just a method of setting up virtual memory. Maybe this actually means something? If so, I have no idea what.

    9)Very simply by dos being HIMEM.SYS and this is what creates the first page table structures. In answer to the question, “how is the first process started?”

    10)This due to nice little design flaw fails at 64 MB. Design flaw?

    11)Fork under NT based like 2000 and XP worked how you described. (This is where the fool starts confusing Cygwin with Windows)

    12) Its starting a new process from a independent binary in the shared memory space. Relocating the start of the executable. (This actually gets more brain-dead a little further on where oiaohm claims that this “standard” behaviour of NT is ” just not a very secure way to run.”)

    13)without a page table real mode x86 can only access 1 MB. Welcome to the stupid world of x86

    14) Threads in 9x function like threads in uclinux not general Linux either. No copy on write. I think this one is still my favourite bit of gibberish.

    15)Because this complete line in the man page is only true until linux kernel 2.5.46 where uclinux was merged providing the other option.

    16)There is only such thing as fork if you are writing by current POSIX standards.

    17)386DX does not in fact include a fully functional MMU. Items like execute protection is missing. 386DX only contains a limited MMU (Gotta admit, this one is a diamond, too)

    And, of course, the NX-bit nonsense. Sofware, hardware, what’s the diff? Threads, processes, what’s the diff? Memory mapping via MMU, memory mapping without, what’s the diff?

    It’s really, really hard to pick the most stupid and comprehensively uninformed comment out of the above.

  60. Deaf Spy says:

    . “NX (No eXecute) support for x86″ is the name of the patch in Linux to implement it. http://en.wikipedia.org/wiki/NX_bit history is of processes that had it before x86 chips got is is interesting.

    Dr. Loser, are you sure this is not better than the statement about HIMEM.SYS allocating the first page?

    The poor maroon can’t tell between hardware and software feature. This is getting sad.

  61. DrLoser says:

    DrLoser wrote, “catastrophic Munich Linux train-wreck?”

    Stand in line, please, Robert. I’ve tried to bring this thread back to its original purpose at least three times. You’ve not only allowed oiaohm to subvert it with his bizarre incomprehension of a) how fork() works and (b) whether or not it was ever relevant to Win9x — you’ve even encouraged him to do so.

    Let’s get the Posix compliancy thing wrapped up, shall we? After that … Munich!

  62. DrLoser wrote, “catastrophic Munich Linux train-wreck?”

    Saving $millions and getting the job done doesn’t sound like any train-wreck I’ve ever heard about. If Munich’s IT were a train-wreck, they wouldn’t have to do a study to review possible changes. The savings are credible. That the system works is true. That they did it differently than I would (heavy on thin clients and terminal servers) means they did spend somewhat more than they should have but compare that to any M$-shop… and it’s no contest. M$-shops are very wasteful right from the licences to the re-re-reboots to the malware, downtime, whatever measure you want.

  63. DrLoser says:

    After about half an hour’s idle contemplation, I’ve finally come up with a use-case for vfork()/_exit(). It’ll be interesting to see if oiaohm, who is a master of this particular subject, can come up with one.

    Clue: mine is to do with data communications.

  64. DrLoser says:

    I should point out that, in strict Posix 2.0 (I think it was) terms, it is possible for pthread_create() and other Pthread calls to be executed entirely in user space, and therefore not be syscalls.

    Completely irrelevant, but, I submit, a more interesting qualification than oiaohm’s ceaseless ignorant drivel on the subject.

  65. DrLoser says:

    Fork is one of these items no one who know how complex the Linux implementation really is would normally not bring intentionally into an arguement.

    Except, possibly, an argument on the question of whether the Linux implementation of fork() is dangerously and stupidly complicated. Luckily, we don’t have to, oiaohm, because you have performed that small service for us.

    In passing, I should say that this is the first and only time I have ever been invited to argue about fork(), a *nix syscall that I have been using on and off since roughly 1996. (Just, not on M$ platforms.) Professionals just use the damn thing. Only the sort of people who need tin-foil hats generally bring the subject up as an “argument.” (Even then, most tin-foil hat wearers can probably muster up a more relevant discussion.)

    Normally you would bring native threading in because at least that does not have all these messes.

    And here I note the pointless use of the qualifier “native.” Aboriginal, perhaps? But no, I suspect it’s a reference to Java … mostly because, outside the Java world, pretty much everybody refers to it as “threading.”

    Pop quiz, oiaohm: on a fully-functional MMU machine, you run two programs. One of them executes a fork() and one of them executes a pthread_create() (or, hem hem, “native” equivalent).

    Here’s the question: Do you end up with the same number of processes in each case?

    If not, is there any useful way at all that you can regard the two syscalls as “equivalent?”

  66. DrLoser says:

    By the way “POSIX.1-2008 removes the specification of vfork()”. There is only such thing as fork if you are writing by current POSIX standards.

    So what?

    I accept TEG’s distinction, because he knows what he’s talking about, but I don’t accept yours. Because yours is based on a Wikipedia article (I think; I tracked down something that was almost word-for-word what you are claiming).

    I am tolerably certain that most if not all *nix-like embedded systems without MMUs avail themselves of vfork(), whether or not it’s specified by POSIX 2008. If only because back in 2008 I was programming against a VxWorks platform, and I used vfork(). Nobody gave a stuff about narrow Posix compliance then. I imagine nobody gives a stuff about narrow Posix compliance now, either.

    There are only two things you can do after a vfork(): one is to call execve() and the other is to call _exit(). I never really worked out why _exit() is in that set, but I’m sure there’s a good reason.

    Anyhow … that corresponds exactly to my point that vfork() is the MMU-less equivalent of fork() for one very important use-case: that is, spawning a new process. IMHO, all of the other possible use-cases for fork() are antediluvian and futile in any case, although I supposed one might substitute pthread_spawn if one felt like being “Posix 2008 compliant.”

    And absolutely none of this has anything at all to do with your risible “analysis” of memory management in Windows 9x and beyond, which, let’s face it, was utterly and completely wrong.

    Down to and including your failure to recognise, as TEG pointed out, that Windows 95 required an i386 processor (complete with MMU) even to run in the first place.

    386DX only contains a limited MMU.

    I hate to burst your bubble, oiaohm, but even some 286 chips featured “a limited MMU.” And as DeafSpy points out, “DEP” has no bearing whatsoever on whether that MMU can efficiently execute a fork(). Even though, on a Win9x machine, it never even needed to do so.

    In that small context, DEP is a protection against buffer overruns and nothing much more than that.

    Now, other than admitting that you have spent the best part of 30-40 posts being thoroughly ignorant of your own subject matter, oiaohm, would you care to explain a single one of the following?

    1) How is any of this relevant to Win9x?
    2) Assuming that any of it is relevant to Win9x, how is it relevant to any IT requirement since, say, the year 2000?
    3) Assuming that it has been relevant to any IT requirement since the year 2000, how on earth is it relevant to the catastrophic Munich Linux train-wreck?

  67. oiaohm says:

    Deaf Spy
    Execute protection, which you should have called by its correct name, and that is DEP, is a very different feature than the whole discussion so far. Nothing to do with threads, either.
    LOL. DEP is not it correct name for the time of 9x. You have been breathing too much Microsoft marketing Deaf Spy. “NX (No eXecute) support for x86” is the name of the patch in Linux to implement it. http://en.wikipedia.org/wiki/NX_bit history is of processes that had it before x86 chips got is is interesting. Windows NT for Alpha had Execute Protection. DEP had a different name from Microsoft back in the time of 9x and that was Execute Protection. This is why PE binary format did not need any alterations at all when DEP was implemented in 2004 in XP. Reality this was just reimplementation of a historic feature from Windows NT for Alpha and a few other archs but this time for x86. 386 was not a full MMU by 1995 standards. It was a part implementation when you compare it to other chips of the time. NX was not the only feature missing either from the 386 MMU that other MMU’s of the time had.

    If Microsoft not renamed Execute Protection to DEP the fact a feature had been not provide and not advertised clearly would have been come more public knowledge.

    Deaf Spy
    Nothing to do with threads, either.
    In fact NX/Execute protection/DEP does in fact relate to threading implementation. To prevent out of control execution. You get to end of execute code block next block should be NX bitted so stopping program executing into no mans land. Missing the feature introduces a stack of particular exploits that are removed by its presence. 386 MMU is a hazard. Yes insecure design OS combined with insecure design processor equals headaches for users. If you are talking about OS stability the quality of the MMU is just critical as the OS usage of it. Best OS with the worse MMU is unstable the Worst OS with the best MMU is unstable.

  68. oiaohm says:

    By the way its super tricky to at times understand what works. If you read the Linux kernel or uclinux documentation alone this is another dead end trap for the incompetent because the uclinux documentation in kernel and from project will tell you that an MMU Less kernel does not implement fork. True the in-kernel does not implement fork. Interesting feature of Linux kernel syscalls like fork can in fact be forwarded to the libc. Only people who have worked with MMU Less are aware how much is implemented but not exactly normally but works anyhow. Yes to have posix conformance to modern the libc has to implement some form of fork and vfork is not suitable. This is why Exploit Guy is out of his depth this is an area were large percentage of the on-line documentation will be telling you all the wrong things. Fork exists MMU less because it has to since 2008. How it exists could be kernel or user space. MMU fork has a in kernel space implementation and a different implementation in glibc/most libc with MMU support and MMU Less fork exists only in the libc but the syscall to Linux kernel still works as a forwards back to the libc.

    Fork errors on MMU Less equals only having 1 possibility code areas of problems where MMU has 2 possibilities. Yes the syscall forwards to userspace works on MMU as well as MMU Less so a syscall fork issue is on a MMU based system is either the libc or the kernel.

    Fork is one of these items no one who know how complex the Linux implementation really is would normally not bring intentionally into an arguement. Normally you would bring native threading in because at least that does not have all these messes.

    Yes fork under Linux is documentation trap after documentation trap.

  69. Deaf Spy says:

    Ohio, you’re good as ever. Good, as relative to you. For common people that would be babbling dimwit.

    I decided to pick two things from your wall of text.

    386DX does not in fact include a fully functional MMU. Items like execute protection is missing. 386DX only contains a limited MMU.
    Execute protection, which you should have called by its correct name, and that is DEP, is a very different feature than the whole discussion so far. Nothing to do with threads, either.

    Just because Windows 95 min requirement is this does not mean Microsoft developers had rewritten the code for performing threading from the old MMU Less that Windows 3.x and before series used.
    Windows 3.x, my dear Ohio, or Win16, have no notion of threads whatsoever. Win32 does.

    Not only you speak utter your mumbo-jumbo, but it is completely off topic, too.

  70. oiaohm says:

    DrLoser
    Now, just today, I have admitted to TEG that I was sorely misled in my belief that fork() on an MMU Linux system was an exact equivalent of vfork() on a Linux system lacking MMU.

    By the way “POSIX.1-2008 removes the specification of vfork()”. There is only such thing as fork if you are writing by current POSIX standards. The Linux syscall fork exists on MMU Less ucLinux implemented as a copy fork. A fork has a different requirement list to a vfork. Posix 2001 that is the current define for fork does not state it has to be a copy-on-write solution. None of the requirements meet what Posix mandates requires a copy-on-write solution. Yes fork in MMU Less Linux and MMU linux both perfectly conform to Posix standards basically the more than 1 way to skin cat problem.

  71. oiaohm says:

    DrLoser please visit http://www.alfresco.com/ you will find Munich is one of it customers. Alfresco has all the features of sharepoint and extras including fully transactional version. By the way the fully transaction version is in the community edition download it and try it out.

    DrLoser
    I maintain that a Linux with MMU using a fork mechanism is, to all intents and purposes, the practical equivalent of a Linux without MMU doing the same.
    DrLoser so lazy you did not check out what uclinux is before commenting.

    They are in fact vastly different. To the point you cannot use gnu libc on Linux without MMU but instead have to use uclibc or other specialist for MMU Less uclinux mode. The native threading used in MMU less is also different in equal ways. If you use anything that expects a copy-on-write style fork or thread you are screwed on a MMU Less system.

    A MMU copy-on-write fork vs a MMU-less copy fork are two major-ally different beasts. A copy-on-write fork you don’t have to perform pointer relocation as the page table running the copy-on-write fork is able to make the address location virtual. Why do you have to perform pointer relocation in a copy fork is because the complete fork code was copied to new addresses. So all pointers now have to pointer to the new addresses. Any location pointer missed in a copy fork will result in copy fork malfunctioning. Yes code converting pointer integers is not going to agree with a copy fork as the integer most likely will not be tagged as data to be relocated. Basically long list of conner cases come out the wood work.

    Threads designed for MMU-Less system have to have unique starting addresses and pointers of application relocated as per the starting address. Threads designed for MMU can used virtual addressing to avoid having todo pointer relocation. This is mandatory with MMU-Less. This relocation has been implemented as attack mitigation in MMU environments even that a lot of Windows applications flag and turn off this mitigation. Windows PE binaries include a flag to mark if an application supported relocation or not. If application does not support relocation 9x will not run it yet NT 3.5 to Windows 10 will.

    MMU Less equals no implemented memory read/write protection as there is no way todo this without a MMU. So it gets worse. Memory allocations in a MMU Less are straight up raw address locations from the point of view of the application and the OS kernel. Buffer overflows can take out near by applications. Yes this is another unique effect that is a give away that you are running on something using MMU Less methods. In a proper MMU based OS attempting to buffer overflow into a near by application should trigger a MMU error so terminating the problem applicaiton. Yes this is one of the things that don’t happen under 9x is MMU errors on lots of buffer overflows when they would throw errors under NT.

    The minimum CPU requirement for Windows 95 is 386DX. In other words, you simply cannot run Windows 95 unless your CPU has already got what an Intel 386DX has got, including a functional MMU.
    There is a problem with Exploit Guys statement here. 386DX does not in fact include a fully functional MMU. Items like execute protection is missing. 386DX only contains a limited MMU. Just because Windows 95 min requirement is this does not mean Microsoft developers had rewritten the code for performing threading from the old MMU Less that Windows 3.x and before series used. Its like I can run uclinux on a 386DX as well. Funny enough this has uclinux acting very much like Windows 9x. The uclinux on a 386DX runs swap and other things using the MMU but the MMU for the forking and threading of the uclinux the MMU is not being used. This shows how little Exploit Guy understands. The threading and process model of OS can be Independent to the swap/virtual memory of the OS. You learn this messing around with uclinux. You need to test if code will bust in a uclinux environment and you have something x86 at hands. Remember a MMU less environment might be direct addressing 4G+ of ram so how will you emulate that on a 386DX configure the MMU to suit this.

    DrLoser and Exploit Guy spend some time reading about uclinux how it works. Then learn the list of behaviors you expect on MMU Less then you will notice error after error from 9x is because of following MMU Less behaviors. We should be thankful that 9x is dead and gone. NT 3.5 and NT 4.0 in time of 9x were many times more stable than 9x because they were using MMU based methods for everything not like 9x that was using using MMU based methods swap and fairly much nothing else.

    Basically MMU Less equals crash in strange and nasty ways unless code is perfectly audited. You would not say that a desktop OS should expect its applications to be perfect.

  72. DrLoser says:

    Being “rational” is a two-way street, Robert. Being “rational” involves admitting it, when you are wrong.

    Even God is wrong occasionally.

    I cite the Axolotl as fairly conclusive proof. (No idea how popular it is on the desktop, however.)

  73. DrLoser says:

    And … you, Robert?

    Strike that. Don’t even bother to waste your time by answering my feeble riposte.

    Robert Pogson claims that he is “the most rational person I know.”

    Occasionally bordering on bigoted gibberish, though. From the same comment in 2011:

    That is not irrationality. I am the most rational human I know. Give me numbers and words describing reality, not personal impressions, any day.

    I’ve done my best to give you the relevant numbers and words, O Rational One. I think this is the beginning of a beautiful friendship.

    …hatred of M$ that causes their share of PC shipments to fall…

    Well, there goes the financial argument for being Miss Most Rational, doesn’t it?

    Apparently “hatred” is not quite enough. Have you tried “thinking,” Robert?

    It’s been all the rage since the sixteenth century, or so I’m told.

  74. DrLoser says:

    In other news, and back we go to Munich (you remember Munich? It’s the title of this song), apparently the fools in charge of a ten year migration still haven’t come up with the equivalent of M$ SharePoint.

    SharePoint can provide intranet portals, document and file management, collaboration, social networks, extranets, websites, enterprise search, and business intelligence. It also has system integration, process integration, and workflow automation capabilities.

    I’m guessing that the Munich “solution” has none of that. And certainly not a fully transactional version.

    But I’m not bigoted. I would be delighted to be proven wrong.

  75. DrLoser says:

    Here’s a completely open question, Robert, and it will help you prove that you are not a bigot.

    When was the last time you changed your mind on any subject of substance at all (not necessarily even IT related), based upon the advice and/or opinion of anybody else at all?

  76. DrLoser says:

    Webster’s, 1913:

    Big”ot*ry (?), n. [Cf. F. bigoterie.]

    1. The state of mind of a bigot; obstinate and unreasoning attachment of one’s own belief and opinions, with narrow-minded intolerance of beliefs opposed to them.

    2. The practice or tenets of a bigot.

    Certainly sounds like one of us, Robert.

    Now, just today, I have admitted to TEG that I was sorely misled in my belief that fork() on an MMU Linux system was an exact equivalent of vfork() on a Linux system lacking MMU.

    I’m humble that way. I’d love to be a bigot, but every now and again, somebody points out that I am completely out of my tiny little mind wrong, and you know what? I have to agree.

    And … you, Robert?

  77. DrLoser says:

    DrLoser wrote, were you to assume that a single one of them had any ideas or knowledge worth spit.

    That is the essence of bigotry, connecting a whole group of individuals with some common characteristic, some interest in FLOSS, with another unrelated characteristic, lack of intelligence.

    Not as such, Robert, and before I quote your beloved Webster’s 1913, may I point out the following?

    1) I am imputing an assumption to “you,” for some value of “you.” I am not denying that the conclusion (these guys are worth spit) is valid.
    2) I am deliberately using the qualified mode, ie “were.” I’m not insisting on anything at all.
    3) I’m not being rude to some anonymous collection of individuals. I’m being specifically rude to the FSFE, who I regard as a small bunch of worthless dingbats. Your mileage may vary. There are several other people whose opinion I value, given that they have “some interest in FLOSS.” You and ram, for example. But not the dingbats.
    4) You do me a grave disservice, Robert — I am appalled. With another unrelated characteristic, lack of intelligence…

    Did I mention “lack of intelligence?” For the purposes of this discussion, I will readily accept that every last member of the FSFE is a Nobel Laureate — I mean, I really don’t know, so why not?

    They’re still dingbats. Highly intelligent dingbats, no doubt.

    But, dingbats.

  78. DrLoser wrote, “were you to assume that a single one of them had any ideas or knowledge worth spit”.

    That is the essence of bigotry, connecting a whole group of individuals with some common characteristic, some interest in FLOSS, with another unrelated characteristic, lack of intelligence.

    “The state of mind of a bigot; obstinate and unreasoning attachment of one’s own belief and opinions, with narrow-minded intolerance of beliefs opposed to them. [1913 Webster]”

  79. DrLoser says:

    And getting back to the FSFE, since nobody here appears to be interested in what a pig’s ear Munich has made of itself. Oh well.

    Looking forward to your report on the T-Dose annual bean-fest, Robert. T-Dose is a totally respectable organisation with a website that looks like it came from the late 1990s.

    Luckily, it promises that “during this event Open Source projects, developers and visitors can exchange ideas and knowledge.”

    Which would be extremely useful, were you to assume that a single one of them had any ideas or knowledge worth spit.

  80. DrLoser says:

    As I have already explained, it’s more like the difference between “there is fork()” and “there is no fork()“.

    Needless to say, TEG, you are 100% correct and I am, on the face of it, wrong.

    I base my discussion of fork() and vfork() and so on not merely on CS theory but, in fact, on CS practice. In other words, use cases.

    Now, both you and I know what a (high level, anywhere from the birth of Unix in 1968 or whatever onwards) “fork” is intended to do. oiaohm obviously doesn’t. (Actually, I could print a T-shirt with that phrase.)

    I don’t want to slop too much over the sides of this discussion for oiaohm to pick upon, with his own unique brand of uninformed gibbering, but at a sufficiently high level, I maintain that a Linux with MMU using a fork mechanism is, to all intents and purposes, the practical equivalent of a Linux without MMU doing the same.

    I’ll agree that that equivalent is limited and subject to user error and basically very contorted and only useful in a very small set of use cases, though.

    May I introduce oiaohm to the exciting concepts of clone() and pthread_spawn?

    I sincerely hope that I may. The results should be spectacularly amusing.

  81. That Exploit Guy says:

    capybara asky asky writey writey

    Capybara? Isn’t that one of those huge guinea pig things?

    Do please explain the difference between a Linux fork on a MMU-based platform and a non-MMU-based platform. That would prove your assertion, wouldn’t it?

    As I have already explained, it’s more like the difference between “there is fork()” and “there is no fork()“.

    There is always a spoon, though.

  82. DrLoser says:

    Notice copy-on-write hypns guess where That Exploit Guy was coping his incompetence from.

    The word is spelled “hyphens,” oiaohm. “Hyphens. You weren’t even close, were you?

    There are probably about half a million graduates in Computer Science at this point. (You, oiaohm, are patently not one of them.)

    Terms like “copy on write” are so all-pervasive, or should I lose the hyphen and say all pervasive, that people will know what you are talking about, hyphens or not.

    Just as long as you don’t try to apply the term to Linux pthreads, of course.

    But still, there’s no harm in trying. You, oiaohm, have mangled the English language repeatedly and in more ways than I can count. So let’s see which one of these is not recognisable as the underlying concept:

    * copy-on-write
    * copy on write
    * copy no right
    * copyright on request
    * capybara asky asky writey writey

    (OK, that fourth one takes a while to think about.)

    Dismal, futile, and pointless as always, oiaohm.

  83. DrLoser says:

    DrLoser really you are presuming a lot where does GPL say that you have to distribute your binaries to anyone other than your clients/users.

    Did I really claim that? How silly of me. It sounds like U2’s bit of megalomania in downloading their latest oeuvre to everybody with the latest iPhone …

    Of course I didn’t claim that. Although I am unaware of any circumstance whatsoever in which a GPL3 licensed executable has, is, or will ever be denied to “anyone other than your clients/users.”

    That would sort of defeat the whole purpose of FLOSS, wouldn’t it, oiaohm? At the first hurdle.

    Feel free to defend this gibberish. And, in case you are prepared to substitute “source” for “binaries,” which given your inability to think straight you might well need to do …

    Be warned. I have an argument against that one, too. A fully GPL3-compliant argument. Read the damn thing first.

  84. DrLoser says:

    That Exploit Guy really I did not need to mention thread.

    It didn’t stop you, though, oiaohm, did it? Presumably you hoped to gain something by mentioning “thread” — perhaps silk, perhaps cotton, perhaps hemp, who knows? — and then were shot down in flames.

    Fork just happens to be one of the classic examples to find out if someone knows the difference between MMU and MMU Less.

    Oh, really?

    Do please explain the difference between a Linux fork on a MMU-based platform and a non-MMU-based platform. That would prove your assertion, wouldn’t it?

    And you can’t because there is no difference whatsoever.

    Try again, kiddo.

  85. DrLoser says:

    A rather heavyweight property for an in-process thread, I would think.

    And, I forgot to add, an extraordinarily undesirable one.

  86. DrLoser says:

    That package is pst-utils, not readpst-utils. I’m away from Beast…

    Good to know that I could get you the first steps of the way to a solution with only fifteen minutes of googling, Robert. Hope it all goes well — can’t see why it shouldn’t, really.

    And I appreciate the reciprocity of quoting pst-utils. That might very well come in handy for me in the future.

  87. DrLoser says:

    .Dr. Loser, I suspect Ohio is trying to best his “HIMEM.SYS creates the first page table so it can access more than 1MB.”

    Is he? You’re a better man than I am, Deaf Spy.

    I couldn’t make head nor tail of that gibberish … oh well, for posterity:

    Threads in 9x function like threads in uclinux not general Linux either. No copy on write.

    If I can descry anything from this total garbage, it seems to indicate that Linux threads — my current understanding is that Linux standardised on pthreads back in about 1998, after a couple of hiccoughs — have the interesting property of “copy on write.”

    A rather heavyweight property for an in-process thread, I would think.

  88. DrLoser says:

    Kolab is only the server. The client on Linux is Kontact, Kontact binaries for Windows that are supported are not free of charge. Kolab Enterprise gets you Kontact Enterprise with all the updates not the 2013 sample.

    Ah, oiaohm, oiaohm, oiaohm. When will you learn to think before committing nonsense to paper (or web)?

    First of all, the stated Munich project goal was not to save money as such. It was to free the City Government of dependence on a single Vendor, ie Microsoft. To which end, I might note, they spent rather a lot of money over ten years funding the private lifestyles of a bunch of bozos who clearly didn’t know how to develop a working Linux system, let alone anything else.

    There is therefore no obvious reason why they shouldn’t have invested a small sum in a third party (a purveyor of Kolab clients, which might or might not be Kontact — the beauty of FLOSS!). Even if it came out at the same price as Outlook, they would still have achieved their stated goals.

    But they could have gone one better than this, couldn’t they? Remember the Four Freedoms?. Here, for example, is a fine link offering the chance to get a rancid “beta-huge-debug” version of the Windows Kontact client — for free!

    (It’s described as the The Kolab Enterprise 5 Client, btw, so you’re plumb wrong on that one, too …)

    Here’s where the Four Freedoms come in. Examine, modify, correct and distribute this code, et voila! A working unified calendaring and mail client for Munich!

    Assuming that you truly believe that Kolab is any more use than a steaming pile of donkey dung.

    Personally, I doubt it.

  89. That Exploit Guy says:

    Something to be very aware of there is a important reason why 9x was design the way it was. MMU less existed in a lot competition computers been sold when 95 enters the market.

    The minimum CPU requirement for Windows 95 is 386DX. In other words, you simply cannot run Windows 95 unless your CPU has already got what an Intel 386DX has got, including a functional MMU.

    Also, by the time Windows 95 came out, Pentium had already been around for 2 years. Your assertion that people were still selling CPUs functionally equivalent to a 286 during that time is simply ridiculous.

  90. That Exploit Guy says:

    Because this complete line in the man page is only true until linux kernel 2.5.46 where uclinux was merged providing the other option.

    Let me guess: you made up this bit of nonsense basing on what you saw in this Wikipedia article. Obviously, the article tells you that the merge between mainline and μClinux came in the form of additional architecture supports. It had nothing whatsoever to do with architectures with functional MMUs that had already been supported (e.g. 386+). Also, in case you haven’t realised, let me clue you in on one thing: there is no fork() in μClinux. Instead, you use vfork() to spawn child processes. In other words, when you run Linux on an architecture that has no functional MMU, fork() simply becomes unavailable, and whenever you need to perform a fork/exec, you simply use vfork() instead of fork().

  91. Deaf Spy says:

    Threads in 9x function like threads in uclinux not general Linux either. No copy on write.

    Dr. Loser, I suspect Ohio is trying to best his “HIMEM.SYS creates the first page table so it can access more than 1MB.”

  92. oiaohm says:

    That Exploit Guy really I did not need to mention thread. Fork just happens to be one of the classic examples to find out if someone knows the difference between MMU and MMU Less. If a person describes fork with copy on write alone means they don’t have a clue on what MMU Less fork looks like. POSIX standard does not define fork as requiring a MMU or using copy on write. The fact you said fork uses copy on write also show that you are incompetent. uclinux is a mainline Linux kernel feature. So you can choose with Linux if you have copy on write using page tables fork or just plain old copy fork same with threading. Of course if you use uclinux methods you have downgraded stability of the OS massively and also break application compatibility in so many ways its not funny due to how many cases applications depend on MMU copy on write style where the addresses of memory have not changed.

    Something to be very aware of there is a important reason why 9x was design the way it was. MMU less existed in a lot competition computers been sold when 95 enters the market. MMU Less fork and threading operations can be put on top of a MMU you just require the MMU to pretend to be flat memory and the MMU Less code works. Now a MMU version of fork or thread on the other hand cannot operate in a MMU Less environment. 9x and the prior versions of that tree of OS’s was designed on the possibility that home pc’s could end up very crappy. Finding a new personal device/computer without a MMU is insanely rare.

    When making an OS you have to make a choice. MMU Dependant or MMU Not Dependant in design. The price for choosing MMU Not Dependant is stability and code incompatibility issues. Yet for some reason MS Trolls attempt to defend 9x. Best defense for the existence if 9x is that Microsoft was allowing that the market could possibly go in a few different directions. The threading model of 9x has more in common with Windows CE than NT and why does it have more in common with Windows CE is because again that is another OS that was designed for MMU less. Windows CE devices also have a history of a strange issues resulting in lock-ups.

    That Exploit Guy saying I did not know what thread was is not correct. If you were getting what fork was wrong there there was absolutely no point talking about threading. Fork is implemented by the c library you are using by posix standards so can be MMU Less or MMU style.

    Most man pages from Linux about fork trap the incompetent googler.
    Under Linux, fork() is implemented using copy-on-write pages, so the only penalty that it incurs is the time and memory required to duplicate the parent’s page tables, and to create a unique task structure for the child.
    Because this complete line in the man page is only true until linux kernel 2.5.46 where uclinux was merged providing the other option. Of course as normal for FOSS no one has seen any particular reason to update the documentation. As the old rule goes read the source. Notice copy-on-write hypns guess where That Exploit Guy was coping his incompetence from.

  93. Joe.M says:

    DrLoser shut up on random guessing and do some homework.

    oiaohm,
    We know that if he was capable of doing that he wouldn’t be a M$tard in the first place.

  94. Joe.M says:

    If you happen to define “sociopaths” as “intelligent gregarious people with enquiring minds, a hatred of cant, and above all a sense of humour,” then …

    sociopath: (1) a person with a personality disorder manifesting itself in extreme antisocial attitudes and behavior and a lack of conscience (eg. you)

  95. That Exploit Guy says:

    maybe oiaohm or other experienced person

    Experienced in what? Manipulating Wi-Fi signals with one’s own thoughts?

  96. That Exploit Guy says:

    No copy on write.

    So you don’t know what a “thread” is. Got it.

    ucLinux is the MMU Less kernel.

    Also known as “4 corner days” or “simultaneous 4-day”.

  97. oiaohm says:

    DrLoser by the way there is a long history of open source programs being free on Linux and paid on Windows to get access to tested binaries.

    QCAD, Kexi…… Its not a abnormality. Why free on Linux. Distribution build and testing servers pick up the production costs also there has no requirement to get decent performance to pay for compliers or development suites on Linux. Recent times a lot of these have changed to provide a trial edition or beta editions only on Windows.

    Lot have been helped that KDE project now runs build farms for Windows. Yes you don’t have to own the complier. Also the Windows version of open source programs has a very bad habit of being many versions behind the Linux versions or unstable in compare.

    DrLoser really you are presuming a lot where does GPL say that you have to distribute your binaries to anyone other than your clients/users. And where does GPL say that the installer has to be the same license. Remember the binary installer is an independent work. Like the WIX scripts you have created. Get the problem yet. They don’t sell you the open source program they sell you the installer and update system that looks after that program. Legal loop holes.

  98. oiaohm says:

    Deaf Spy Thread implementation in 9x does not function like fork in Linux or default Linux threading. Threads in 9x function like threads in uclinux not general Linux either. No copy on write. ucLinux is the MMU Less kernel. Please note That Exploit Guy raise fork http://mrpogson.com/2014/10/16/fallout-from-munich/#comment-206130 reason why I am using the term Fork is because That Exploit Guy was too incompetent to mention Thread. Its not like Linux kernel does not have native threading. I use the same term the other party uses. Yes if you read my descriptions I am description the 9x threading model. 9x is design around MMU Less design. You would not make a desktop Linux like OS using the uclinux because its going to have huge risks of instability due to the lack of memory protection. 9x displays all these issues.

  99. oiaohm says:

    DrLoser shut up on random guessing and do some homework.
    Probably not. There is absolutely no chance whatsoever that Munich-an-der-Linux will see anything remotely like SharePoint in 2015.
    https://summit.alfresco.com/barcelona/sessions/documentation-stadtwerke-munich
    Munich uses alfresco. So why do they need sharepoint they already have it replaced. This is the problem DrLoser most of what you are claiming about if I did far enough I will find video and white papers from the Munich personal showing exactly what they have been using over the time frame.

    DrLoser
    “Kolab isn’t free on Windows,” which would be a specious argument even if true (it isn’t).
    Kolab is only the server. The client on Linux is Kontact, Kontact binaries for Windows that are supported are not free of charge. Kolab Enterprise gets you Kontact Enterprise with all the updates not the 2013 sample.

  100. That package is pst-utils, not readpst-utils. I’m away from Beast…

  101. DrLoser wrote, “You are very welcome.”

    Yes, indeed. That seems to be reasonable advice consistent with what I’ve been reading since I wrote that comment. M$ has opened the specs to its file-formats, and this neighbour has used Outlook from home and work for ages. I don’t think it’s work-stuff he wants to save but I will check the next time I see him. He’s just retired so he may or may not have access to the server at work.

    I’ve found that Evolution and other e-mail clients can import .pst and I have a script to convert to formats that Evolution and Thunderbird can use, based on readpst-utils.

  102. DrLoser says:

    Not relevant to Munich, but whatever.

    I am not sure about that. On one hand, the way Peter Dolding lies is often so transparent it is simply unlikely that he is unconscious of what he is doing.

    I beg leave to differ, TEG. oiaohm is an incompetent idiot with delusions of … well, actually, the ability to do simple arithmetic.

    He’s not unconscious of what he’s doing at all.

    He just doesn’t have the brains to connect a logical proposition, via experimental evidence, to a practical conclusion.

    Even smart crows can do this. Without benefit of magical mystical microwave brain projection!

    I deduce that crows are perfectly capable of lying to themselves. There is no reason to suppose that oiaohm is not capable of lying to himself.

    I have two questions here:
    1) Why does oiaohm seemingly believe that his “blant,” oops, sorry, “blatant,” I used a spell-checker — as if I needed to — lies will be believed by any other even vaguely sentient human being on the planet?
    2) Why did Mr Pogson, just now, ask oiaohm for advice on how to transfer emails from an Outlook account to (say) a gmail account?

    This can only end in tears, Robert. You know that very well.

  103. DrLoser says:

    I did mention that he could likely run XP in a virtual machine on GNU/Linux and filter the networking to protect XP so as to continue running his current e-mail client. He has backups of everything so it should be doable one way or another.

    Sorry to say, Robert, but that’s the “mentioning” of somebody who has absolutely no clue what he is talking about.

    Do you want to revisit my concept of a $0 Windows 7 machine that gives you the joy and pleasure of experimentation?

    I can still help you with that, you know. And the best part of it is, you can help other people like your friend here.

    Oh, I forgot. Munich is ten years’ worth of pointless sucky shit..

    There, I feel all better now. I’d hate to drift off-topic.

  104. DrLoser says:

    He doesn’t know how to move that mail to any other client. I don’t either, but maybe oiaohm or other experienced person or Google could give suggestions.

    No, it isn’t worth another post, Robert.

    What your friend needs to do is to make sure that he has a copy of his Outlook mail on his local machine.

    The server version of an Outlook mail server file has the extension “.pst,” which is of course useless to your friend. The equivalent (OK, there are syncing issues, but let’s leave those to one side as probably irrelevant in this case) is a “.ost” file. You can force Outlook to download absolutely everything to a local OST file. (Details omitted but available on the Web.)

    He then needs to use something like this to migrate to, in this case, GoogleApps. There are various other migration options — which obviously depend upon the target email system.

    You are very welcome.

  105. DrLoser says:

    With everybody’s permission, I’d just like to reproduce the following:

    HIMEM.SYS creates the first page table so it can access more than 1MB.

    I vote oiaohm’s effort here up as this week’s gem. Not only on this thread, but on all the other threads on Mr Pogson’s blog.

    It is the single stupidest thing I have seen for a very long time. Compared to this, even the assumption that a South Korean SSID has enough entropy to be a worthwhile security item is pure genius.

    Speaking of which:

    Even Gnu/Linux cannot beat entropy.

    Heck, it takes ten years to admit that FLOSS can’t even produce an everyday item like a unified calendaring and email system.

    Who could possibly have guessed?

  106. I don’t know whether this is worth another post or just a comment but I have just returned from a six-mile hike. About mile 4, I met a neighbour who was waxing philosophically about openness in government. The new slate of representatives gives us a local government potentially more open, efficient, responsive, etc. I mentioned that openness was being formalized in Europe, right down to the software… So, we ended up talking about the death of XP and his problems escaping M$.

    It turns out that the one and only obstacle for him using GNU/Linux is a decade or more of e-mail in the clutches of “Outlook”. He doesn’t know how to move that mail to any other client. I don’t either, but maybe oiaohm or other experienced person or Google could give suggestions. I did mention that he could likely run XP in a virtual machine on GNU/Linux and filter the networking to protect XP so as to continue running his current e-mail client. He has backups of everything so it should be doable one way or another. The guy’s retired and once winter sets in he will have plenty of time to work on the migration. He also has 4 PCs so he could move stuff around and work on a real PC rather than a virtual one. After hunting season, I will have plenty of time to help too.

  107. DrLoser says:

    I suppose now is not a good time to mention SharePoint either?

    Probably not. There is absolutely no chance whatsoever that Munich-an-der-Linux will see anything remotely like SharePoint in 2015.

  108. DrLoser says:

    Has anybody come up with a convincing explanation of why it’s taken Munich until 2015 to produce a workable version of “unified calendaring and mail” yet,” aka Kolab, which is actually free and under the GPL3 license — 1.0 better than the Linux Kernel itself! — yet?

    I haven’t seen one. I’ve seen oiaohm claim that “Kolab isn’t free on Windows,” which would be a specious argument even if true (it isn’t). Last I heard, some 10% of the Munich IT estate was still on Windows, because there are some things they still depend upon. Apparently the back-office drones appear not to consider Outlook as one of them.

    I’ve seen Robert claim that “they don’t really need that,” which is, I am afraid, not much of a defence. It’s an awful defense even if you substitute “a secretary and a PA system, which is probably more expensive than the commercial solution.

    And certainly far more impracticable. Let’s face it, we’re talking about Munich here. Munich is quite large. The 15,000 or so IT desks are distributed across, I don’t know, I’m guessing, a 100 KM square area.

    Let’s buy the (sole) secretary a Mercedes, so’s she can drive around at high speed and shout at everybody.

    No, wait, let’s buy the (sole) secretary a Lada. Cheaper. Still good for shouting. Probably enough money to spend on a bull-horn.

    08:00 prompt, it’s Freising. 09:00 it’s Erding. 10:00, it’s the turn of Ebersberg, which is a very annoying place with a chatty janitor, so it takes an extra half an hour.

    Rosenheim, Miesbach, Bad Tölz-Wolfratshausen, Starnberg and Fürstenfeldbruck follow at 11:30, 13:00 (half an hour for lunch!), 14:00, 15:00 and 16:00.

    Just time enough to get back to the centre and sync up with the Big Bods of Government, before going home and quaffing a traditional Litre of Strepsil to remediate the sore throat.

    Alternatively, of course, one might choose to use software. Assuming it was available.

    Which apparently it won’t be until 2015.

  109. DrLoser says:

    I guess those guys who use GNU/Linux for HPC, web servers and games just don’t have a clue… [SARCASM]

    Actually, “Win32 doesn’t have fork() because it has threads …” was quite clearly aimed at oiaohm, who had previously demonstrated his usual level of ignorance of the subject.

    Even I would not claim that oiaohm is a leading light in the architecture of “GNU/Linux for HPC, web servers and games.”

    But, judging by your sarcasm tag, you obviously know better. Perhaps oiaohm really is the eminence grise behind “GNU/Linux for HPC, web servers and games.”

    That would certainly explain a lot of things I wonder about. oiaohm hasn’t quite managed to wreck High Performance Computing yet, but if he persists with his “unique” theories of flat physical memory models, fully shared across all processes, and the central importance of the fork/exec family of APIs …

    Yes, he’d be the man to do that. Time to go back to World War Two, I think. Far less confusing times for people who have never really come to terms with the modern IT world.

  110. DrLoser says:

    P.S. Btw, what does HPC has to do with threads?

    Threads are vital to High-Performance Computers because threads are highly scalable.

    oiaohm will be along to help Robert to explain this obvious fact in a moment or two.

    Just after he’s finished confusing Cygwin (non-native) forks with HIMEM and some sort of von Neuman flat memory space that mandates shared memory even though, actually, any OS at all (yes, Linux does this too! Since 1995, I think!) has chosen to take advantage of the MMU and virtual paging space.

    Please don’t ask either of these IT specialists to explain memory caches and speculative look-ahead pipelines to you. Their apparent command of these High Performance Computing features would blow your mind. Much as the following comment blew mine:

    Indeed. These people are sociopaths, and you seem to have stumbled on some severe examples.

    Well, yes, sure.

    If you happen to define “sociopaths” as “intelligent gregarious people with enquiring minds, a hatred of cant, and above all a sense of humour,” then …

    Ya got it in one, Joe M!

    Hmm. Thinking about that, it’s an interesting question. Why does nobody involved in the FLOSS world seem to possess even a vestigial sense of humour?

  111. That Exploit Guy says:

    but if you ever know what you’re talking about, you’d never make such a blant mistake.

    I am not sure about that. On one hand, the way Peter Dolding lies is often so transparent it is simply unlikely that he is unconscious of what he is doing. On the other hand, sometimes he just seems to be so thoroughly convinced of his own crazy stories that I can’t help but think he suffers from some form of schizophrenia. The resemblance of his writing to Time Cube and Dr. Bronner soap bottles simply doesn’t help one bit.

  112. Deaf Spy says:

    I guess those guys who use GNU/Linux for HPC, web servers and games just don’t have a clue… [SARCASM]

    Sigh… Pogson, I wrote: “Unix had no notion about before pthreads”. pthreads are here since 95 or so, which is almost 20 years already. I can’t believe I need to state something that obvious.

    Gee, I wonder why the Unix world ever bothered to design and standardize phtreads, if fork(), according to Mr. Pogson, is so stupendous. [SARCASM, eh 😉 ]

    P.S. Btw, what does HPC has to do with threads? And, where are the games in Linux you are talking about? SteamOS? Ah, coming, just around the corner. Wait for another five years, and you’ll have the year of Linux for gaming.

  113. Deaf Spy wrote, “Win32 doesn’t have fork() because it has threads, something that Unix had no notion about before pthreads, and that fork() is the poorman’s wretched excuse for threads?”

    I guess those guys who use GNU/Linux for HPC, web servers and games just don’t have a clue… [SARCASM]

  114. Deaf Spy says:

    That Exploit Guy, shall we confuse the little Hamster even further by telling him that Win32 doesn’t have fork() because it has threads, something that Unix had no notion about before pthreads, and that fork() is the poorman’s wretched excuse for threads? 🙂

    Ohio, to try to name himem.sys a memory-manager is so pathetic that it even hurts. Gosh, I am aware that the readings in Wikipedia might be a bit confusing in this regard, but if you ever know what you’re talking about, you’d never make such a blant mistake.

  115. That Exploit Guy says:

    HIMEM.SYS creates the first page table so it can access more than 1MB.

    Absolute BS. As I said, the job of HIMEM.SYS is to provide a low-level API for program to access extended memory in discrete chunks called Extended Memory Block. What’s more – there is nothing in HIMEM.SYS or the base components of DOS that possesses the ability to create page tables because both of these things predate the 386 MMU. In other words, they simply do not and cannot utilise the 386 paging ability in any way and therefore have no use whatsoever for a page table.

    Its starting a new process from a independent binary in the shared memory space.

    Again, you cannot share a virtual memory space. What you can share are the frames that the pages map to. In other words, you can set up mappings so that the same virtual address in two difference virtual memory spaces map to the same physical address. You may even set up mappings so that two different memory addresses in two difference virtual memory spaces map to the same physical address. What remains, however, is that a virtual memory space is fakery performed by the MMU to give the appearance of a single contiguous address space to the running process – it simply cannot be shared among processes.

    Then VMM32.VXD is used to create the next page table set in 9x containing being a copy of what is allocated in HIMEM.SYS. HIMEM.SYS is the first point of page tables created then the on going coping.

    Again, there is no such thing as a page table for HIMEM.SYS. The idea of running HIMEM.SYS under a virtual memory model is simply preposterousness in the extremes.

    “shared flat memory” can contain sections of virtual memory. flat memory is referring to the addressing model. linear memory model is another thing I can call it. Virtual memory can make up segments of linear/flat memory.

    A virtual memory space in Windows 9x is a single contiguous space addressable with linear addresses. This is why it is called “flat” memory. This is also in contrast to the Windows 3.x memory model, which uses the segment/offset address method typical of pre-386 programs. Also, a virtual memory space is not made of anything – it is simply a space mappable to another space – and it is entirely possible for a virtual memory address to map to absolutely nothing. The fact you think a virtual memory space is made out of “segments of… memory” indicates that you understand neither the computer science concept of virtual memory nor the mathematical concept of mapping.

    Sorry to say That Exploit Guy fork under 9x don’t work anywhere near how you described.

    Of course 9x doesn’t work anything like how I described, you moron. Anyone with the most basic knowledge of the Win32 API knows that it doesn’t have a fork() function. What you have cited is a discussion on Cygwin, which is hacked on top of the Win32 API to provide a Unix-like API.

    Next, pay attention to what the other person is talking about before replying with BS. (Actually, don’t, because I am sick of your BS.)

  116. Joe.M says:

    Yep. The more scared they become the more shrill, repetitive and picky they become. Folks like this in the real world would be homeless loners, unable to get along with anyone.

    Indeed. These people are sociopaths, and you seem to have stumbled on some severe examples.

  117. DrLoser says:

    My general approach is to disagree with you fundamentally and on practically every supposed “justification” you offer for your obviously ill-thought out, unprepared and unconsidered claims, as here.

    I’m not even going to limit my complete lack of pickiness to this particular thread, Robert.

    I’m quite happy to extend my abhorrence of picking small tedious little details from the morass of completely incomprehensible wilful ignorance to your post on South Korean SSIDs.

    Regrettably, owing to a sad lack of adequate refutation, I have to resort to simple blind repetition.

    Even Gnu/Linuxc cannot beat entropy.

  118. DrLoser says:

    And some of us even compose music….

    Call that music? It wasn’t even composed with FLOSS! Without Transparency and Accountability, it’s just Plinky-Plonk!

    And let me tell you, you picky little swine, back in the good old days, back in World War Two when Men were Men and Computers were almost entirely Women*, nobody cared about swinish wastes of time like music!

    NO! Back then, we listened to Big Bands and Doo-Wop and There’ll be Unspecified Yet Representatively Free Avians Flying As Close As Shellshock Will Let Them Over The Possibly Blue But GiMP Hasn’t Quite Got There Yet Cliffs Of Eastbourne Whoops A Small Bug In FLOSS Geo-Positioning There …

    It was undoubtedly a fine lyric, but back then in 1940 we didn’t have Huge Monolithic Linux Kernels and …

    … well, we just ran out of notes.

    Tragic, really.

    Anyway, about ten years of Linux in Munich, Robert.

    It’s been unmitigated worthless shit, hasn’t it?

    The sooner it’s gone, the better it will be for every single citizen, and on an everyday basis, the better it will be for everybody working in the municipal government.

    * This is a true fact. Look it up.

  119. olderman says:

    “Yeah, that’d about sum us up.”

    And some of us even compose music….

  120. DrLoser says:

    ooh, you’re soooo … scary!

    Is that a purple latex rubber blow-up doll I detect with my HammySenses(TM) over the internet in the background?

    Naaaah …. forget that, Robert.

    She’s not your type.

  121. DrLoser says:

    Ah, the comfort of having a member of the peanut gallery who agrees with you unconditionally. No need for thought, eh?

    Yep. The more scared they become the more shrill, repetitive and picky they become.

    My ears are not as finely tuned as yours, Robert. I am unable to detect “shrill” over the internet.

    As for “repetitive,” it is only necessary to repeat a quote (from the author’s own cite) when the author (repeatedly) fails to address it.

    As for “picky,” I wouldn’t be too hard on yourself. I don’t spend my time analysing every small detail in your responses for grammatical errors or other trivialities.

    My general approach is to disagree with you fundamentally and on practically every supposed “justification” you offer for your obviously ill-thought out, unprepared and unconsidered claims, as here.

    To be fair, I will often be “picky,” and agree with you on various lesser points.

    Folks like this in the real world would be homeless loners, unable to get along with anyone.

    Believe me, Robert, the Black Country is about as gritty and “real” as it gets. You’ve not experienced proper reality until you’ve bitten into a lump of deep-fried pork gristle and lost a filling or two.

    Oh, but maybe by “real world,” you meant “being part of the private-sector workforce and able to purchase novelty blow-up dolls in purple latex to compensate for their lack of friends …”

    … wait a mo’, that didn’t quite come out as I intended …

    “respectable members of their local LUG”

    … hang on, that’s practically a tautology …

    “bog standard people who have jobs and houses just like anybody else who is blessed with professional qualifications and an open mind.”

    Sad people, us self-sufficient employed professionals with an open mind and an active social life that doesn’t involve either purple latex dolls or hanging around the nearest LUG and begging for small change.

    Yeah, that’d about sum us up.

  122. olderman says:

    “Folks like this in the real world would be homeless loners, unable to get along with anyone.”

    You know better than this Robert Pogson. In fact I would suggest that you look in the mirror before you presume to make such statements about what you do not know.

  123. DrLoser says:

    There’s no particular need for “integration”. In the old days one would have monolithic programmes sometimes without an OS underneath/standing alone.

    First of all, you are (unpardonably) confusing business/government activity with, of all things, batch processing. Or even 1970s IT in general. Thus my reduction ad absurdum to the Sassanid Empire, which worked tremendously well via clay tablets and not much else.

    But that’s a tremendously stupid and inefficient way to run a government of any size in the 21st century. Stupid as in wasting colossal amounts of time, effort and money. Inefficient as in wasting colossal amounts of time, effort and money.

    And secondly there is “a particular need for integration.” Your customers demand it. You, Robert, should anybody take the dubious step of putting you in charge of a government of a municipality 1,000,000 strong, are not the arbiter of need. Everybody else is.

    Your argument boils down to “the entire work force of the Munich government should accept that they are not able to use this bog-standard everyday tool, because FLOSS is too broken to provide it.”

    And I note that you would not be using this argument if FLOSS was not too broken and was able to provide it.

    This is a grotesque and feeble argument and a very sad state of affairs in Munich.

  124. Joe.M wrote, “You can sense the fear and paranoia of the M$tards”

    Yep. The more scared they become the more shrill, repetitive and picky they become. Folks like this in the real world would be homeless loners, unable to get along with anyone.

  125. DrLoser wrote, “used to working with a unified mail and calendar application back in 2004, and has been staggering along since without one.”

    Obviously, they were not staggering on. They have e-mail. They have schedules. There’s no particular need for “integration”. In the old days one would have monolithic programmes sometimes without an OS underneath/standing alone. They were cumbersome and inflexible. So would the concept of a unified e-mail/calendaring system be if an organization operated in any way different than the writers imagined. No such application can be right for 90% of organizations. That’s why there are different choices.

  126. Joe.M says:

    You can sense the fear and paranoia of the M$tards in this comment section.

  127. DrLoser says:

    It [a unified mail and calendar application] is a completely obvious requirement for any corporate body, commercial, non-commercial, government, or even Dr No, anywhere on the planet with more than, say, 50 people involved.

    That’s utterly ridiculous.

    Sadly you are correct and I am wrong, Robert. What I should have said is that it is a requirement that is completely obvious to everybody on the planet except for Mr Pogson and a dwindling number of retired World War II generals.

    For the extremely large set of organisations I defined, it is indeed completely obvious. So completely obvious, indeed, that this is what practically all of them (90%+) use.

    A secretary and a PA system can do wonders.

    TEG has already pointed out how ridiculous this is.

    And it’s not only ridiculous, it’s expensive. A secretary costs $20,000 pa minimum. Even Microsoft wouldn’t dare charge that much for Outlook. Not to mention that we are talking about the free Kolab.

    I’ve been in schools with ~1600 people in one building and stuff happened without that other OS and without M$’s applications.

    Both anecdotal and frankly medieval, Robert. And note that you are the one insisting on Outlook, not me. I’m just arguing for similar functionality, as provided by the free Kolab.

    An announcement would be made. All the affected people noted the message in their Daybooks and carried on.

    Doesn’t sound very … automated, does it? Also doesn’t sound easy to handle time-slot bookings, co-ordination of groups of individuals, rescheduling, moving to another meeting room, etc etc. Which is what modern day computer calendaring systems do, at far less cost than the minimum $20,000 your notional secretary would charge.

    But that’s just calendaring, Robert. Nobody is accusing Munich of not having a calendaring system in place. In fact it’s probably Kolab, which is free software.

    The present discussion is about a unified mail and calendar application.

    Which apparently will not grace the corridors of power in Munich until some time in 2015. Presumably in Beta.

    Heck, the whole of WWII was organized with very few computers, certainly not one on every desk.

    And the Sassanid Empire performed wonders of organisation with little more than a series of clay blocks and cuneiform inscriptions.

    So what? We’re not talking about the Sassanid Empire. We’re not even talking about World War II.

    We’re talking about an entire municipal IT organization that was presumably used to working with a unified mail and calendar application back in 2004, and has been staggering along since without one.

    There is genuinely never a good reason to take away an everyday facility like this from your users and fail to offer them a viable alternative for the next ten years.

  128. oiaohm says:

    That Exploit Guy without a page table real mode x86 can only access 1 MB. Welcome to the stupid world of x86. HIMEM.SYS creates the first page table so it can access more than 1MB. This due to nice little design flaw fails at 64 MB. Then VMM32.VXD is used to create the next page table set in 9x containing being a copy of what is allocated in HIMEM.SYS. Plus means to access more. HIMEM.SYS is the first point of page tables created then the on going coping. Still does not change the fact that HIMEM.SYS is the first turtle in the stack of turtles making up 9x.

    “shared flat memory” can contain sections of virtual memory. flat memory is referring to the addressing model. linear memory model is another thing I can call it. Virtual memory can make up segments of linear/flat memory.

    http://octave.1599824.n4.nabble.com/cygwin-fork-bug-on-Windows-9x-td1643241.html
    Sorry to say That Exploit Guy fork under 9x don’t work anywhere near how you described. Screw the idea of normal copy on write to start off with as you need per process page tables todo that what you don’t have. 9x in fact duplicates and reallocates when you fork. So the forked process has different address to it parent for the same blocks under 9x. Pointers in 9x fork instances can get very interesting to say the least. Anyone who has played with cygwin or msvc porting unix based code to 9x find this out fairly quickly. Fork under NT based like 2000 and XP worked how you described.

    MMU fakery can be run in all kinds of way. The way the 9x developers decide todo it is extremely odd ball. In fact the old ball ways comes from the fact Windows 1.0/2.0 run in real mode without access to any pages tables or MMU fakery so no means at all todo copy on write. 9x has the historic odd ball behaviors. That Exploit Guy the history of where 9x comes from is required to understand why its behaving so odd ball. Does it make those odd ball behaviors sane no it does not.

    That Exploit Guy just think for one min how would you implement fork if you had no copy on write. Reality that is what 9x is doing. You are apply too modern of methods to how you are thinking 9x works.

  129. That Exploit Guy says:

    “shared flat memory” is a feature of dos applications and 9x.

    In case you are not catching on: there is no such thing as “shared flat memory”. Again, a “page” is nothing more than logical construct that exists solely for the purpose of representing a mappable area in a memory space, which we refer to as “virtual” because it does not exist in reality.

    Duplicating a page table won’t cause two processes to share the same virtual memory space. This is obvious considering that a virtual memory space is merely fakery performed by the MMU when the process is running: it is inherently not sharable between processes. The only things sharable are the frames that the virtual memory pages are mapped to, and copying the page table allows the mapping between virtual memory addresses and physical addresses to persist across the fork() process. Any attempt to write to a frame simply results in the frame being copied to a new location in the physical memory, which can then be written to without affecting other processes sharing the original frame, and the page mapped to the original frame is then remapped to the new frame. We call this process “copy-on-write”.

    Very simply by dos being HIMEM.SYS and this is what creates the first page table structures.

    Bawhawhaw… Oh, mercy!

    HIMEM.SYS is what we call an “eXtended Memory Manager” (XMM). Its primary function is to allow real-mode programs to access physical memory beyond the first megabyte through a low-level API. Memory allocated through an XMM is anything but flat, and there is nothing that HIMEM.SYS does that even comes close to creating a page table. After all, HIMEM.SYS is a DOS driver, and DOS is anything but a multitasking OS.

  130. oiaohm says:

    That Exploit Guy file object issued happened in 9x all the time. Same issue the Linux fork thing describes except on a worse scale.

    “shared flat memory” is a feature of dos applications and 9x. Its not a feature of Linux, VMS, BSD, Unix, NT or any other sanely designed OS. NT applications takes slightly longer to start up due to having to populate page table with items that are shared. All proper protected mode operating systems take this time. Problem is 9x losses almost all the same amount of time relocating the binary. 9x is just not a proper protected mode operating system.

    Then how exactly the first ever process is started, mind if I ask?
    Very simply by dos being HIMEM.SYS and this is what creates the first page table structures. Sad reality here 9x does not create the first process memory space itself that it keeps on replicating to everything. It is not even part of the windows binary so the first turtle has to exists before windows 9x win binary even runs. Yes no HIMEM.SYS no 9x windows.

    That Exploit Guy 9x design is just completely nuts. Notice something NT memory manager is part of the OS kernel. 9x Memory manager is not part of the kernel but a part to use with DOS with all the limited security features Dos single application at time applications expect.

    So yes turtles all the way down is a very good way to describe 9x since it is truly that insane. Its not a feature of 9x that makes it work it a feature running on Dos that makes it work.

  131. That Exploit Guy says:

    That’s utterly ridiculous. A secretary and a PA system can do wonders.

    Absolutely, if the world was just one huge supermarket.

    “Cleanup on Aisle 4!”

  132. That Exploit Guy says:

    Except 9x is doing for every application run. Also unlike fork its not duplicating the calling process. Its starting a new process from a independent binary in the shared memory space.

    Really? Then how exactly the first ever process is started, mind if I ask?

    Does the operation involve a very tall stack of turtles going all the way up from the ground infinitely below?

    On NT every application binary starts at the same appearing address to the application why because each application is getting its own page table.

    So now only applications in Windows NT get their own page tables. What happened to all the duplication crap you mentioned about “shared flat memory”?

    Yes Linux programing guides and books tell you to avoid fork due to the race conditions that happen and do happen. Guess why 9x was so unstable.

    The race condition described in the cited webpage involves an execve() taking place between open() and an attempt to flag the open()‘d file descriptor with FD_CLOEXEC using fcntl(). It has nothing to do with the stability of anything.

    Try writing better stories next time.

  133. DrLoser wrote, “It’s a completely obvious requirement for any corporate body, commercial, non-commercial, government, or even Dr No, anywhere on the planet with more than, say, 50 people involved.”

    That’s utterly ridiculous. A secretary and a PA system can do wonders. I’ve been in schools with ~1600 people in one building and stuff happened without that other OS and without M$’s applications. An announcement would be made. All the affected people noted the message in their Daybooks and carried on. Heck, the whole of WWII was organized with very few computers, certainly not one on every desk.

    No, M$’s salesmen made such things seem essential just like V-8 engines were essential. I drive just fine with a 4-cylinder engine and I’m very gentle on the throttle. Of course, I can add ~50KW from the battery with just a twitch…

    Is life better with such toys? Sometimes not. I was at one place that had a combined e-mail/calendar/forum thingy. I won’t mention the name of the software. It was junk. I got into trouble once because I was zooming text for my old eyes and sent an e-mail which intimidated the recipient because this wonderful system copied the zoom… Another time in a forum, I questioned the success of a conference ripped by food-poisoning and was publicly scolded for not cheering… Such tools in the wrong environment can cause more problems than they solve. They may be fine for a small group working closely together but they are just a waste for a large group who work rather independently. I’ve seen lots of systems that don’t scale, like the place where a memo was distributed to 3000 employees on demand and then countermanded the next day ( turf war…). It took a minute or two at least to read and understand the two memos. That’s man-days of lost productivity in one transaction. Further, the two memos were policy so both had to be inserted into every departmental policy-manual which was required reading for all staff. It was a four-inch binder when I was there.

  134. oiaohm says:

    That Exploit Guy true what 9x is doing is very much like fork under Unix. Except 9x is doing for every application run. Also unlike fork its not duplicating the calling process. Its starting a new process from a independent binary in the shared memory space. Relocating the start of the executable. On NT every application binary starts at the same appearing address to the application why because each application is getting its own page table. What 9x is doing is perfectly inside what the MMU allows. Its just not a very secure way to run. Just because you have MMU and virtual memory does not mean you cannot do actions to under mine the protections virtual memory offers.

    http://www.linuxprogrammingblog.com/threads-and-fork-think-twice-before-using-them
    Yes Linux programing guides and books tell you to avoid fork due to the race conditions that happen and do happen. Guess why 9x was so unstable. Understanding the memory management that 9x was using explains why it crashed a lot. Race conditions due to lack of memory security. So anyone complaining about having 9x crash a lot that was status normal and to be expected from 9x design. Add in a few leaks and you have a complete disaster.

    Note That Exploit Guy first claimed that Microsoft had to be doing something odd. With a little more homework he then finds out no its not special it what hardware can do. Just because hardware can do something does not mean you should. The description you dug up of virtual memory describes more how NT works. Did not describe overlapping. Its kinda in face that 9x is overlapping memory when it relocating the start of executables address to the application. Non overlapping you don’t need to relocate the start of executables to the application why because the address is virtual page tables mean that address can be anywhere.

  135. That Exploit Guy says:

    In any case, Peter, you can always stick to writing fantasy epics on the subject NDISWAN, though it seems that you prefer others to just forget about it.

  136. That Exploit Guy says:

    Instead of creating a new page table per process you just duplicate it between processes.

    Again, you are failing to take into account that a “page” is nothing more than a logical construct in the MMU. More hilariously, you somehow believe that, by duplicating the page table, you’ll somehow make two processes share the same virtual memory space.

    Those who aren’t ignorant buffoons, on the other hand, are perfectly aware that this is essentially what fork() is in a nutshell. That’s not quite a Windows 9x thing, I am afraid.

  137. oiaohm says:

    DrLoser
    Did you, for example, go to “Programs and Features,” click on MYOB, and choose the “Repair” option?
    The repair option in Programs and Features is gone because because the system restore removed it in fact the MYOB entries are removed. This is a problem with a program half in user account and half in system mixing with system restore. All your general cure it instructions are broken by using system restore bar returning to newer version or completely reinstalling and hoping there are not conflicts installing program over itself.

    The damage to the data files can be already done by the time you see error messages at times. This is why system restore should be last option. In fact restoring a clonezilla/ghost image of the complete system can be safer. Applications on the machine decides if you should attempt system restore or not. Reinstalling the complete machine from scratch might sux but sometimes that is the best option.

    Which is how the “magical” repair function works — on absolutely any MSI you care to install on Windows, even though the software author has no clue how it works.
    You have just stated another myth. Software author with no clue how MSI works can wrap a general exe installer inside that happens to work in ways MSI cannot track and happens not to function again after its run once. MSI are not like DEB files that have validation tools to make sure they are not doing any action that is untracked. Debian dedicates servers to installing and uninstalling packages and comparing what is before and after to find issues. A random deb file from some where that does not validate can damage debian as well. It is really required for software author of the installer to under stand how MSI works to avoid doing anything wrong.

    If MSI file that did the install you are messing with is audited and knowing to function properly yes you can use MSI repair. If it not you just might have made matters worse. This is something I particularly hate about windows. Debian repositories packages are audited for correct install, uninstall and repair.

    This is what I get sick of from MS Trolls over stating functionality of system restore and MSI and other parts. They have flaws. You need to know flaws so they don’t burn you.

    Now, to be fair to oiaohm, he is the only person here who has tried to defend this utter abnegation of forward planning. (Ten years!)
    The Munich forwards planning had that every machine would be running Linux so Windows only solutions was not required because the machines with Windows would be running Linux as well. That has not happened so they just need to change the plan in a minor way to include a Windows desktop client.

    If confronted, in 2004, with an evident need for an integrated mail/calendar system, what do you do?
    In 2003 Munich the answer is provide it by web browser interface using kolab community. Its not that there was not a integrated mail/calendar its that particular people wanted an application on the desktop and the solution was browser. Application on the desktop undermines central backup by the way. Of course this seamed like the obvious choice at the time.

    DrLoser you are using the advantage of hindsight every can say what was perfect todo with hindsight because every thing about how the plan went south was covered.

    Yes there is a reason to use the web interface when you are still migrating machines. You don’t end up with data on users computers that gets formated when changing OS with all the stuff provided by browser. Basically until the client OS stabilized it was not possible to deploy local application safely. Yes they were doing something over all those years working out the desktop OS ratio and deployment locations. It is not that they were doing nothing and that no solution was deployed. The solution deployed the current major does not like that is all. So it makes absolute sense what Munich IT did but you can also see how it could make some users unhappy.

    DrLoser “We do nothing.” is for every year you stated because the functionality was done in 2003. Functionality was meet. Form of functionality was issue.

    https://www.kolab.org is the community version/open source edition. This is not kolab enterprise. If you build from source you do not get support from kolabsys.com and access to particular items like current working builds of the Windows client. Just because you can build from source does not mean you get everything. This is a true fact about FOSS.

  138. oiaohm says:

    That Exploit Guy
    Really, so you are saying that Windows 9x shifts the silicon of the CPU and makes it do untold Microsoft voodoo that only you know rather than what it’s hard-wired to do?
    Shared Flat memory mode is just a method of setting up virtual memory. Instead of creating a new page table per process you just duplicate it between processes. 9x is not quite perfectly flat memory shared in protected mode but its close for kernel and share libraries is very close to shared flat memory. The X86 cpu is designed todo isolated protected memory and shared protected memory. 9x uses shared protected memory where ever possible NT based used isolated protected memory where possible. Completely different models and different set of issues. Shared protected memory with flat declared zones as 9x uses means a lot of leaks don’t in fact become unaccessible. Its just like a x86 is design to operate a Microkernel using 4 different protected mode levels yet most OS’s decide not to use this. x86 design is able todo a lot of things.

    That Exploit Guy thinking that Microsoft is doing some untold voodoo to make the memory mess that 9x is really shows lack of knowledge. x86 architecture guide by intel includes many virtual memory setup model examples including the horible insecure version 9x uses. There is also a reason why 9x uses is “DOS Protected Mode Interface” Also uses the horrible insecure lack of processes memory allocation. A lot of “DOS Protected Mode Interface” application break under NT 3.5 and 4.0 because it is they enforcing proper separation.

  139. DrLoser says:

    I am not as wise, nor as assiduous a researcher, as oiaohm, and therefore I will take the following on trust:

    If you wish to use [Kolab on] Windows pay money.

    And I am therefore minded to quote the equally believable Richard M Stallman:

    Freedom 0: The freedom to run the program for any purpose.
    Freedom 1: The freedom to study how the program works, and change it to make it do what you wish.
    Freedom 2: The freedom to redistribute copies so you can help your neighbor.
    Freedom 3: The freedom to improve the program, and release your improvements (and modified versions in general) to the public, so that the whole community benefits.

    Oh, but maybe the license got in the way?

    Vendor: Kolab Systems GmbH
    Developer: Kolab community, Kolab Systems
    License: GPL-3.0 and other
    Web: https://www.kolab.org

    I guess it didn’t, did it?

  140. DrLoser says:

    And now, back to our regular feature: Munich. Specifically, a sad lack of planning, over ten years, for a completely obvious requirement for any large bureaucracy, such as the City Government of Munich.

    No, strike that. It’s a completely obvious requirement for any corporate body, commercial, non-commercial, government, or even Dr No, anywhere on the planet with more than, say, 50 people involved. Need I repeat the wise and brutally honest words of the FSFE again? Needs must:

    Regarding the missing unified mail and calendar application criticised by Schmid, it became clear that the implementation of the corresponding Free Software solution Kolab is currently in the works. It was only initiated in early 2014 and is expected to be used starting in 2015.

    Now, to be fair to oiaohm, he is the only person here who has tried to defend this utter abnegation of forward planning. (Ten years!)

    Let us consider oiaohm’s defence:

    DrLoser you attempt to drag back on topic does not understand there are two kolab products. Kolab open source and Kolab enterprise. Kolab enterprise comes with Windows client that has support. Munich was using Kolab open source for a long time problem no supported windows client. If you wish to use Windows pay money.

    Note: I quote the whole paragraph. I didn’t pick and choose.

    Question: If confronted, in 2004, with an evident need for an integrated mail/calendar system, what do you do?

    And let’s stop right there.

    What was the answer in 2004? We do nothing.

    What was the answer in 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013?

    We do nothing.

    The belated attempt to resuscitate a doomed project, which has only recently become apparent because there is considerable political opposition, plus a review process, by plonking all the chips on Kolab is completely irrelevant.

    At this point, it doesn’t matter whether or not Kolab is the appropriate choice. It doesn’t matter whether there is a FOSS version of Kolab. It doesn’t even matter whether Munich would be better or worse off using Microsoft Outlook.

    What matters is that this was an obvious choice that had to be made back in 2004.

    And the operational functionality is STILL NOT THERE.

  141. DrLoser says:

    (I believe the worthless spell-checker on this version of whatever framework it is corrupted “systemd” into “system.”
    (Oh well, baby steps. The overall layout of posts is infinitely better than it was. I can live with correcting a moronic and unavoidable robot if necessary.)

  142. DrLoser says:

    Needless to say, I was remiss and did not check lower down for TEG on MYOB Error 9004.

    My heart-bled apologies to TEG. That pretty much settles oiaohm’s hash in this particular instance.

    Nevertheless, oiaohm, my point about Restore Points and the (rarely necessary) availability of the MSI “repair” function still stands. Not to mention the STR thing, although I doubt we will ever see one from you.

    About the MSI, btw. The whole point about software installations on Windows is that they are mediated by a transactional system involving several discrete internal (yet accessible) tables that define every single step of the process.

    Personally, having spent the last couple of weeks working with a WiX script, I consider them a little rough around the edges. But, what with being transactional, you can roll those transactions (and there are several transactions within the main one) backward and forward.

    Which is how the “magical” repair function works — on absolutely any MSI you care to install on Windows, even though the software author has no clue how it works.

    Alternatively, of course, you could just rely on a primitive 1980s architecture like system.

    Wait, wait, system is heresy round these parts, isn’t it?

  143. DrLoser says:

    I can’t stop myself. I am intrigued by your complete and utter nonsense, oiaohm. I have a deviant fascination for the limited capacity of an incompetent mind to absorb and accept the obvious logical conclusions of the evidence in front of them.

    In other words, I’m going to be a traditional Software Developer, faced with a reported issue, oiaohm. I’m going to ask you for the Steps to Reproduce, typically abbreviated to STR.

    So, if MYOB deemed you spongeworthy, and you were the lucky recipient of a message such as a Myob Error 9004, what did you do next?

    Did you, for example, go to “Programs and Features,” click on MYOB, and choose the “Repair” option?

    If not, then why not?

  144. DrLoser says:

    Well, I have better uses for $150 and I would probably be criticized for not working with M$’s latest and greatest OS, ever, “8” at what? 8% global share of the desktop? No, I would rather not waste my time.

    I’m not entirely sure where Windows 8 comes in, Robert, considering that you spend so much of your time in comparative antiquity, reminiscing about the Good Old Days when you could poke fun at Windows 95.

    So, let’s settle on a happy medium. A large proportion of the desktop-using world, right here and now, are happy using Windows 7. You are fearlessly advocating that they are wrong to be happy doing that.

    Which one is your problem? Paying $150 to purchase an experimental rig, on which you can prove your various points?

    Or the horrid possibility of admitting to yourself that you are wrong, and that Windows 7 suffers none of the technical issues that you bang on about?

    You can go with admitting the horrid possibility if you like.

    But, if you’re going with the “I won’t spend more that $0 on a Windows 7 machine” thing … I can help you with that.

  145. DrLoser says:

    Oh, “creative”, you say? I hope that doesn’t involve Dave Culter travelling to a distant land to slay the dragon and save the princess, because that’s just clichéd.

    It’s David Cutler, TEG, you ignorant slut.

    And it’s not clichéd at all. I sold Dave the travel tickets for that particular jaunt.

    Admittedly, this might be considered anecdotal and dubious, but not quite as dubious as the Tales of Easterville that seem to crop up here on a regular basis.

  146. DrLoser says:

    DrLoser you never read MYOB install instructions there is a warning note about this.

    There are many otiose things in life that I choose not to do, oiaohm. That is indeed one of them. In fact, reading pointless documentation about broken software products is very definitely on the top of my list for things not to bother doing.

    Don’t let me stop you doing so, though.

    If you are lucky you get a Myob Error 9004 that it has detected damage to the system files it depends on.

    And if you are unlucky? Don’t tell me. Cthulhu rises up and informs you that your mystical microwave brain-control powers have been revoked.

    Probably a glitch in the Windows Registry.

    I’ve just detected follicle damage in your left armpit, oiaohm. Take it from me, it’s true. Your left armpit has follicle damage. I am an authority on the subject of follicle damage, particularly as it applies to the left armpit.

    In much the same way, MYOB is the world’s leading authority on damage done to Windows system files after reversion to a Restore Point.

    Except that it isn’t.

  147. That Exploit Guy says:

    Not reallocatable yes but still readable and write able.

    In case you are too think to understand the linked webpage I gave you, a memory “page” is a logical construct that represents a mappable area in the virtual memory space. It itself is not supposed to be readable or writable, but you are welcome to come up with a fantasy mechanism for Windows that completely defies the architecture of CPU, like you did here.

    Principles of virtual memory good write but 9x is that crap that is does match how 9x operated.

    Really, so you are saying that Windows 9x shifts the silicon of the CPU and makes it do untold Microsoft voodoo that only you know rather than what it’s hard-wired to do?

    NT leaks were a little more creative

    Oh, “creative”, you say? I hope that doesn’t involve Dave Culter travelling to a distant land to slay the dragon and save the princess, because that’s just clichéd.

  148. oiaohm says:

    That Exploit Guy
    Ah, so it seems that “Error 9004″ occurs when MYOB can’t locate JRE. It’s an error message commonly seen by those bootlegging the software by copying it straight from one computer to another.

    If you cannot work it out with the new MYOB that installs it core program into the User Profile. Performing a system restore to a point prior to when MYOB was installed results in MYOB running as if you had just copied straight from one computer to another. All the random disasters that can happen from copying program from 1 computer to another all can happen. Programs like this you have to be very careful of using system restore. This is a real issue that bites people who random-ally use system restore and don’t take anywhere near the due care.

    You mean particular programs depends on pages getting leaked out and becoming usable by neither the system nor the process? How interesting!
    This is wrong for 9x because they did not become not usable. Not reallocatable yes but still readable and write able. In the 9x case there was never a case that leaked memory become unaccessible to process. Heck in 9x you could access other processes memory directly. 9x vm really did not work from a security point of view. There are quite a few shim in windows 2000 and XP emulating some of this behavior. 9x accessing not allocated memory was also acceptable. Why 9x even run with how it vm was constructed was more fluke than good management. 9x was more a flat memory system than a virtual memory system.

    Principles of virtual memory good write but 9x is that crap that is does match how 9x operated.

    NT leaks were a little more creative I wish I could remember one of the number. One was super creative because it was a lost lock resulting in application assigned memory not respecting free command because it was in usage. Of course some applications were performing use after free then after the application closes then it becomes lost in vm space. So fix the bug and programs crash while running don’t fix the bug leak memory to no mans land when applications closes.

    If I had the list of windows 2000 and xp shim in front of me I would be able find the bugs involved. There are quite a list of shims for emulating NT and 9x issues that application depend on. Most at the time on NT and 9x were resulting in memory leaks. Windows 2000 with the compatibility system fixes so much it not funny. Even then Microsoft released Windows ME because there were too many applications from 9x doing strange things to make the compatibility system handle it in 2000. It took until windows XP for Microsoft to expand the compatibility system to cover most of the horible 9x was. Please note the word most.

  149. That Exploit Guy says:

    If you are lucky you get a Myob Error 9004 that it has detected damage to the system files it depends on.

    “Error 9004”, huh? Let’s Google this sucker and see what we get…

    Ah, so it seems that “Error 9004” occurs when MYOB can’t locate JRE. It’s an error message commonly seen by those bootlegging the software by copying it straight from one computer to another.

    Hmmm….

  150. That Exploit Guy says:

    Don’t tell me that other OS was perfect on our computers. It crashed randomly in normal use.

    Yeah, right.

  151. That Exploit Guy says:

    The typing tutor one is funny.

    I prefer typing monkeys. They are more fun just for being monkeys!

    What is nasty about them yes they are leaks but particular programs depend on the behavior.

    You mean particular programs depends on pages getting leaked out and becoming usable by neither the system nor the process? How interesting!

  152. oiaohm says:

    DrLoser you never read MYOB install instructions there is a warning note about this.
    If you are lucky you get a Myob Error 9004 that it has detected damage to the system files it depends on.

    Regarding the missing unified mail and calendar application criticised by Schmid, it became clear that the implementation of the corresponding Free Software solution Kolab is currently in the works. It was only initiated in early 2014 and is expected to be used starting in 2015.
    This is not referring to what you think it is.

    DrLoser you attempt to drag back on topic does not understand there are two kolab products. Kolab open source and Kolab enterprise. Kolab enterprise comes with Windows client that has support. Munich was using Kolab open source for a long time problem no supported windows client. If you wish to use Windows pay money.

    https://kolabsys.com/products/kolab-enterprise . This is not the open source product at all. Based on the open source product correct.

    Schmid complaint was lack of client on Windows machines paying for Kolab enterprise fixes this. But the reality they did not have budget to pay for very much at all.

    Basically news writers like
    http://www.infoworld.com/article/2610363/open-source-software/munich-opts-for-open-source-groupware-from-kolab.html
    have got it wrong all over the place.

    All the city’s LiMux PCs and the remaining Windows PCs will be using the Kolab Desktop Client in combination with the Kolab web client based on Kolab Enterprise 13, it said
    At least infoworld includes the correct information with miss leading title.

    https://kolabsys.com/news/introducing-kolab-enterprise-13 Yes they are switching to a new kolab version in 2014 to 2015. This is not changing group-ware product as such. This is really like someone doing a exchange upgrade.

    https://fsfe.org/news/2014/news-20141016-01.en.html Never trust fsfe to give unbias reporting.

    DrLoser please read the end of one of my posts it contains the information why 9x and NT 4.x and before are such a disaster.
    http://technet.microsoft.com/en-us/library/dd837644%28v=ws.10%29.aspx
    This shim system only comes into existence in Windows 2000. The shim system is required to work around programs that depend on memory leaks. So effectively everything Windows prior to Windows 2000 is absolute crap because you could not patch memory leaks for all programs when you had 1 program that depended on it.

    If you knew your topic you would have shut Robert down on talking about 9x properly by accepting its junk and pointing out why DrLoser.

  153. DrLoser wrote, “it is clearly not true, is it?” as if the BSOD never existed.

    see Bill Gates
    see How many bugs will be in Microsoft’s Windows Vista? That estimate applies to Lose ‘9x too.
    see Lose 2K also

    Don’t tell me that other OS was perfect on our computers. It crashed randomly in normal use. GNU/Linux did not.

  154. DrLoser wrote, “why don’t you help your arguments out by toddling down to your local downtown second-hand shop and splashing out $150 on a used Windows 7 machine?”

    Well, I have better uses for $150 and I would probably be criticized for not working with M$’s latest and greatest OS, ever, “8” at what? 8% global share of the desktop? No, I would rather not waste my time.

  155. DrLoser says:

    (Other than oiaohm, who as usual made no sense at all on the subject.)

  156. DrLoser says:

    Dragging this topic back to its origins, with some difficulty, I might add:

    Regarding the missing unified mail and calendar application criticised by Schmid, it became clear that the implementation of the corresponding Free Software solution Kolab is currently in the works. It was only initiated in early 2014 and is expected to be used starting in 2015.

    Anybody out there have any observations on this sad lack of planning?

  157. DrLoser says:

    Gibberish, gibberish, gibberish. Specifically:

    Of course using system restore to a restore point prior to in time of installation is disobeying MYOB installation instructions.

    No it isn’t, oiaohm.

  158. DrLoser says:

    Speaking of that Dark Side, Robert:

    I am all but convinced that everybody here labelled a “M$ shill” has, at some time, had cause to use the Linux Desktop. I am all but convinced that we are more than willing to use the Linux Desktop again, to test out any theory propounded.

    This, I believe, is the Scientific Method, as applied to experimental observation concerning IT. We’re all quite happy to do that.

    So, why don’t you help your arguments out by toddling down to your local downtown second-hand shop and splashing out $150 on a used Windows 7 machine?

    You wouldn’t be polluting your working hardware. And you’d have oodles of fun proving us all wrong!

  159. DrLoser says:

    My estimate is out by a magnitude, but, let’s face it, 499,999,950 is still quite a noticeably large number when you are talking about PCs falling over left right and centre twice a day or so.

  160. DrLoser says:

    Every one on the planet was having similar problems with the particular build of Lose ’95 that we had.

    I think that’s a marginal overestimate, Robert. I don’t actually remember the number of PCs, planet-wide, that had Windows 95 installed at any point in their life, but I’m slightly dubious that it was anywhere near the global population at the time.

    Let’s say, for the sake of argument, 1 billion people? And let’s go a little further and assume that an average of 2 people shared each machine? (Windows 95 wasn’t that hot at multi-user login accounts, so that sounds reasonable.)

    The problem was with each and every PC, not just one. There was no way to fix the problem. It seemed to be the OS losing resources, a memory leak. No updates were available to fix that. All lose ‘9x had that problem.

    (Your link to google books appears to be broken, btw: at least from this end.)

    Here’s what you do when confronted with what appears to be a systemic problem, Robert.

    1) Since there’s no way of telling that the phenomena you observe are of purely local origin, you find an equivalent test subject that is isolated from your current environment. In this case, a Windows 95 machine set up by, say, some bloke in another part of the Canadian School System. I would recommend Manitoba: presumably you flew back during vacations and would have been able to test your theory thusly.

    You didn’t do that, did you?

    2) Rather than leap to the conclusion that “there is no way to fix the problem,” you consult a peer. Once again, I would have recommended someone else in the Canadian School System.

    You didn’t do that, did you?

    3) Under absolutely no circumstances whatsoever do you make wild all-encompassing claims, such as “all Lose 9x had that problem.”

    Which of course you just have. But it is clearly not true, is it?

    By my estimate, above, there were 4,999,999,950 Windows 9x machines out there, and it is clearly not the case that they all went pear-shaped every two hours or so. Some proportion might have done so, I grant you, but you have yet to offer any proof that this proportion was particularly significant.

    So, basically, none of what you say, which is any case purely anecdotal, makes any sense.

    The only logical conclusion to reach is that you didn’t try very hard at all, because you didn’t want to. (There is a slightly more offensive conclusion, but I’m aware that you are technically competent in general, so I refuse to offer that imputation as an alternative.)

  161. DrLoser says:

    “You” is the person who bought the OEM PC, not some stranger who has not “accepted” the EULA.

    Let’s review the EULA, shall we, Robert?

    we grant you the right to install and run that one copy on one computer (the licensed computer),

    This is the main clause of the sentence. The subject of the sentence is “we.” “We” is, fairly clearly, Microsoft. The transitive object of the verb (which should really be “grant to,” but let’s not quibble) is “you.”

    Now, interestingly, what with English being the way it is, there is no way to distinguish the singular second person pronoun from the plural second person pronoun. Well, there is, but the Microsoft Legal Department cops enough grief without referring to thee and thou.

    Even more interestingly, there is absolutely no qualification to the pronoun.

    Does it say, as you imply, “you, who bought the OEM PC?” Or “you, who are about to create a login for yourself?” Or “you, who promise to be a meanie and not let your neighbour anywhere near the machine?”

    It does no such thing. Funny, that. And therefore, by inference, it allows “you” (a place-holder to be determined somehow) to create as many login accounts as you like and to assign those login accounts to as many other people as you like.

    …for use by one person at a time, but only if you comply with all the terms of this agreement.

    Here we see a sub-clause, dependent upon “the licensed computer.” Again the pronoun is important: in this case it is a pronounal phrase, one person.

    I imagine that any rational human being would agree that there is no necessary implication that you and one person are identical in any way.

    I’m almost certain that you are well aware of this, Robert. I am also almost certain that you are aware that there are countless Windows-infected PCs out there that sit in (for want of a more general term) the family living room and are accessible to the entire family, not to mention anybody in the vicinity who drops by to ask for a metaphorical cup of sugar.

    Let’s be honest: if your neighbour wanted to, there’s no legal obstacle at all to his/her letting you log in to his/her Windows 7 PC to check out some detail or other.

    I’m not even recommending that you dirty your hands by doing so. I’m merely suggesting that it might help you check some of my more outrageous and fantastic claims, were you to take a Brief Trip to the Dark Side and back …

  162. TEG wrote, ““If the problem was neither the hardware nor the software, then what was it? Bad feng shui?””

    You gain nothing by omitting the phrase, “on a single PC”.

    If I had a CPU overheating on one PC, I could reduce the problem by shutting that machine down or fixing a fan or… If I had a bad hard drive on a single PC, I could do the same. If I had malware one one PC, I could do the same but it was the same problem on all the PCs and it was M$’s software that was the problem. The PC that crashed in the middle of my classes could be any one of the five and with five-minute booting it caused a problem for me and my students. There was no way to fix M$’s software. Read the EULA. Changing to GNU/Linux did fix the problems. We had uninterrupted service for the rest of the school year.

  163. oiaohm says:

    Did it install magic pixie dust to prevent monsters from eating the PC whole as well?
    That Exploit Guy really its not magic pixie dust. Its defects in Microsoft Libc and other libraries. Features included with windows are not always perfect. The typing tutor one is funny. Programmer deleted scores directory to clear high scores the issue is the way it looked up application location. Broken code in windows returns user profile instead of application path for applications inside the user profile.

    MYOB installer installs fix locking bug fixes for Windows and install SQL server embedded particular versions as well as particular versions of other things shared problem here is MYOB does not call SQL server embedded or any of the other parts by version. Yes you can have more than 1 installed. One can be the wrong version. Result is unreadable data.

    If the system is missing particular windows updates MYOB ships bundled with their installer. Application installers do in fact install Windows Updates at times because they depend on those updates to prevent bugs in program. MYOB is not the only program doing this. The reality is installer of program alterations to OS cannot be undone and expect to run the application.

    Nothing I am talking about here is magic. Basically you want 20 page write ups right exploit guy detailing every bit myob installed and how all this interaction results in massive data loss. Of course using system restore to a restore point prior to in time of installation is disobeying MYOB installation instructions.

    9x has a stack of leaks like http://support2.microsoft.com/kb/242161
    What is nasty about them yes they are leaks but particular programs depend on the behavior. So you prevent the leak a group of applications don’t run. 9x is a horible time. I can also pull leaks like this for NT 4.0 and 3.5.

    Windows 2000 was not much better but it started getting better. The introduction of the shim system/compatibility mode system in Windows 2000 to allow applications to have individual hacks started sorting this problem out. So leak free could be provided to most applications under 2000 with a shim hack for those that depended on the hack.

  164. That Exploit Guy says:

    Here’s what I wrote, shooting down his point in flames: “It wasn’t defective hardware or defective software on a single PC but some random PC so taking one out of service wasn’t an answer.”

    And here’s what I wrote, shoot down Robert Pogson argument in flames while Van Halen is playing guitar in the background: “If the problem was neither the hardware nor the software, then what was it? Bad feng shui?”

    The problem was with each and every PC, not just one. There was no way to fix the problem. It seemed to be the OS losing resources, a memory leak. No updates were available to fix that. All lose ‘9x had that problem.

    Windows 95, huh? Given that Windows 9x had generally lower resource limits (e.g. GDI objects) than NT, anything could have caused that. I dunno, but I’ll need more than hand-waving to be convinced that the problem was caused by a memory leak in the operating system. I am pretty sure someone with supposed training in the scientific method and academic vigorousness can provide some concrete proof for that, right?

  165. That Exploit Guy says:

    That Exploit Guy Current MYOB updates parts of Windows when it installs. These updates prevent MYOB doing nuts things to its own files.

    Absolutely. Did it install magic pixie dust to prevent monsters from eating the PC whole as well?

    Here’s is how I am going to read your posts from now on, Peter Dolding: for every detail you neglect to substantiate, I’ll take what you say as a lie. Giving your track record, I think that’s only reasonable.

    Understood?

  166. TEG, desperately trying to score a point, quoted part of my sentence, missing important stuff… “It wasn’t defective hardware or defective software”

    Here’s what I wrote, shooting down his point in flames: “It wasn’t defective hardware or defective software on a single PC but some random PC so taking one out of service wasn’t an answer.”

    The problem was with each and every PC, not just one. There was no way to fix the problem. It seemed to be the OS losing resources, a memory leak. No updates were available to fix that. All lose ‘9x had that problem.

  167. oiaohm says:

    Mind you myob was the best behaved example. There is a typing tutor program that install in profile that when damaged instead of just deleting the high score it does equal to deltree on the complete profile directory. So myob is a minor brain screwup.

    That Exploit Guy what distribution where you running with that WAN failure. Because it was not debian or you did not know how to use debian because you have different Linux kernels to choose from. Debian provides more than 1 kernel version. Latest kernel from the mainline is really last restort you have 8 different options before you get there with Debian. Also being a wifi problem lot of case this can be solved by blacklisting the kernel driver and using ndiswrapper.

    The claim of try different distribution or Linux kernel mainline is rarity with Debian.

    By the way there were many motherboard that XP and Linux runs on but Vista does not as well. Random lockups with Linux releated to wifi/WLAN does always require more detailed investigation did you speed check you WLAN I think you will find its under performing. Linux has habit of completely dieing when a network card does. Windows on the other hand shows it up as poor transfer rates. This is a preference in a server room better to completely die than function poorly.

    Just because Windows runs does not mean everything is OK. You WLAN card could be defective and by replacing it your WLAN performance could improve massively.

    Broken hardware + Linux = Kernel panics.

  168. oiaohm says:

    That Exploit Guy Current MYOB updates parts of Windows when it installs. These updates prevent MYOB doing nuts things to its own files.

    The It that destroys the MYOB data is MYOB itself when you open it because its run time is damaged by the restore point reversal. MYOB the newest versions install the core of the program in the Users profile. Not shared for all users on the computer. This is where trouble starts. MYOB is a per USER license. So you install it per USER ACCOUNT on the computer. Not everyone remembers todo backups when they should. MYOB is not the only program out there that installs in the users profile directory. System Restore needs to be used with extreme care when you have cases of applications in the users profile directory. There are typing tutor programs and other odd programs that install the same way as MYOB.

    That Exploit Guy by the way items like MS Office by group policy and directly can also be installed per User Account. There is a big cascading explosion you can cause by using system restore if you are not aware of the traps.

    The problem here is System Restore is not a cut or dried option.

    That Exploit Guy I was not talking about data files as such. I was talking purely about executables. Of course this does not physically destroy the machine but it can more than ruin your day.

  169. TEG wrote, ” you didn’t even attempt anything close to figuring out that the OS was what caused the system crash.”

    Nonsense. Every one on the planet was having similar problems with the particular build of Lose ’95 that we had. I wasn’t going to wait a year or two for the budget to contain an upgrade. Eventually, I met Lose ’98 which was no better. Students doing ordinary things crashed that one every few hours. I could browse and word-process and get it to fail in minutes. Both shipped with thousands of bugs and no security.

    There were no updates for Lose 95 B, which we had, released to the public. They released new features only.

  170. That Exploit Guy says:

    I create a restore point. I install MYOB. A issue appears I roll back to restore point. Guess what. 1 MYOB is only partly uninstalled. 2 what is missing means it destroys the MYOB data file. 3 accountants very upset.

    Accountants, huh? Did these accountants look at your email and tell you that you typed worse than their teenage children on SMS? I dunno, but unless they were in somewhere so remote from civilisation they had to hire the same person that fixed their roofs to fix their computers, I suppose they could always find someone that could write properly to do stuff for them.

    Then there is the problem of not using the backup function to save everything from MYOB to a zip archive, because, well, who needs backups anyway?

    Then there is also the problem that MYOB has its own way to save data on the local hard disk drive. Trusting restore point to not muck with them obviously wouldn’t work, and of course there would be remnant of the data where System Restore wouldn’t touch, but apparently the person you have taken this story from simply didn’t pay attention to this detail.

    Then there is also the problem of not simply undoing the restore point rollback. Seriously, that would be the obvious thing to do once you have realised that something has gone wrong, but, hey, I suppose Peter Dolding would just come up with some BS story about undoing System Restore causing half of the computer to get eaten physically, and I am simply tired of debunking his transparent lies.

  171. That Exploit Guy says:

    Wrong link. Here is the proper one.

  172. That Exploit Guy says:

    @Robert Pogson

    I can’t parse that. What’s your point?

    I had a laptop that came preinstalled with Vista, and I heard Linux was so good and Vista was so bad that I just had to give a popular Linux distro of the time a try.

    Because everything I ran on my laptop at the time was all Windows software, the only thing Linux had to do was to be the host of a virtual machine running XP. Simple enough, alright? Well, as it turned out, the laptop would simply crash several times a day with the capslock blinking like it’s in some kind of seizure. As it turned out, the kernel panicked, and the culprit was the WLAN driver. At this point, I had two options: one involves moving to another distro, and the other involves trying my luck with the latest kernel from the mainline. My choice? I simply kicked Linux to the curb and reinstalled Vista. Here was the same machine with Vista installed having run 4 months in a row without powering down or restarting.

    My live couldn’t have been happier since.

    It wasn’t defective hardware or defective software

    I can’t parse that. If the problem was neither the hardware nor the software, then what was it? Bad feng shui?

    You see how I managed to go from a frozen screen to the diagnosis that a bad driver was at play? That’s effort for a piece of junk operating system that I was more than happy to get rid of in a heartbeat. You had everything from Event Viewer to an array of kernel debugging tools and the best you could come up with was this? That’s simply did not help with the impression that you were simply looking for an excuse to force Linux and FOSS ideology onto your students, did it?

    I had to get rid of the OS.

    No. You didn’t have to. In fact, you didn’t even attempt anything close to figuring out that the OS was what caused the system crash. You simply jumped to the conclusion that the OS was at fault as an excuse to promote FOSS ideology.

    So much for someone supposedly well trained in the scientific method and academic vigorousness.

  173. oiaohm says:

    I would argue – that a standard with several years of history behind it and wide adoption is more likely to be functional, secure, and low on bug count than something knocked up yesterday by some know-nothing dipstick in a basement.

    DrLoser every security protocol has a live span. There is a reason for this. The old the protocol the more likely it will become flawed due to computer evolution. Security protocols are not old the the better. As computer power bigger the possibility that a single person will have access to a enough computer power to brute for it someway comes possible.

    Security protocols are true shades of grey. You don’t want them too young and you don’t want them too old. Too young could not be audited enough. Too old flawed due to technology improvements and possible errors in protocol(protocols are written by humans so a percentage always need to be killed off). If several years of history is less than 10 the protocol is mostly likely alright. Please be aware what the current SSL versions are.

    http://en.wikipedia.org/wiki/Transport_Layer_Security

    Most current SSL version is TLS 1.2 released in 2008 so it has over 5 years of testing behind it and we still don’t have everywhere migrated to it. SSL 1.0 SSL 2.0 SSL 3.0 are old out date the TLS contains revisions and fixes in fact TLS 1.0 does not have the flaw that SSL 3.0 has. Really from a practical point of view browsers should be rejecting TLS 1.0 and displaying warning messages for TLS 1.1 connections to encourage upgrading. And when the next TLS gets like 4 years old star displaying warnings about TLS 1.2. Lets not let them bit rot until they start failing under our feet.

    Security protocols require a constant cycle of replacement as they are updated for to protect against attacks from current and future hardware.

    Heartbleed we should not keep quite about. But it is very important not to just make out its a FOSS thing only. There are closed source programs effected that should be end of life and removed from all usage because of Heartbleed. Heartbleed is a symptom of the fact we have been running without proper security protocol and application/libraries audits in place. The new Google flaw should not have existed if proper cycles were happening. Its also very clear closed source companies like Microsoft and Cisco systems has not been doing it because they have not called for end of usage of particular standards.

    Heartbleed is kick everyone up ass. Shellshock is give OS X (apple) heavier hit since larger market share they should be faster.

    DrLoser unified mail and calendar application when it comes to kolab is the application called Kontact. Kontact predated outlook. Kontact is scary to install on windows as you watch 2G of harddrive space disappear. Kolab was used in Munich when they were still running NT 4.0 by its Web interface. Munich was never an exchange environment.

    System Restore does not affect any of your documents, pictures, or other personal data. Recently installed programs and drivers might be uninstalled.
    DrLoser cannot read that complete line is a problem in fact the first time I read this I did not understand what this line really means. Best way for me to show what it means is write up the real world example of how this interacts.

    I create a restore point. I install MYOB. A issue appears I roll back to restore point. Guess what. 1 MYOB is only partly uninstalled. 2 what is missing means it destroys the MYOB data file. 3 accountants very upset.

    How did this happen MYOB is one of the nasty programs that install some of its self in the USER ACCOUNT and the rest of it Runtime in the global system. The rollback by system restore undermines MYOB runtime but still leaves the program in place for the user to click on but now broken without any sign that it is.

    The key words were “might be uninstalled” you need to be very careful before using restore points. Particularly for applications that half install in the user account and half in the system. There are also ways for drivers to install under Windows to avoid being uninstalled by system restore. That note is a big huge warning because that action of system restore can blow your feet off. Using a clonezilla image of a complete machine is safer so you don’t end up with half state. If user is using programs like MYOB system restore should be completely not recommended even under Windows 7+.

    MS Trolls always hold up system restore as magic bullet. The reality its not a magic bullet. System restore can ruin a business. You don’t see people recommending backing up before doing a System Restore as you should with items like MYOB.

  174. DrLoser wrote about the EULA of “7”…

    ” we grant you the right to install and run that one copy on one computer (the licensed computer), for use by one person at a time, but only if you comply with all the terms of this agreement.Typically, this means you can install one copy of the software on a personal computer and then you can use the software on that computer.”

    “You” is the person who bought the OEM PC, not some stranger who has not “accepted” the EULA.

  175. DrLoser wrote, “every single last one of the various government and corporate organisations I can think of have come to expect the existence of a unified mail and calendar application”.

    Nope. I’ve worked for government and there has never been a need for unified mail+calendar. That’s just M$’s marketing. e.g. Just about every teacher I know is required to keep a paper journal and it’s usually provided by the employer and includes a calendar. Some employers require employees to use the company e-mail service but others allow personal e-mail, but not for “official” purposes. I can see some kinds of employees would definitely benefit from unification, like salespeople who spend all day meeting people, but a lot of us have periodic duties written on paper or Post-It notes, or daily plans. As a teacher I had meetings no more often than weekly and anything else went into the daily planning book.

    On my present GNU/Linux desktop I use a calendar for the odd meeting and I use Gmail but there’s no need at all for one to know about the other.

  176. DrLoser says:

    An interesting little gem from the FSFE cite at the OP, btw:

    Regarding the missing unified mail and calendar application criticised by Schmid, it became clear that the implementation of the corresponding Free Software solution Kolab is currently in the works. It was only initiated in early 2014 and is expected to be used starting in 2015.

    Correct me if I am wrong, but every single last one of the various government and corporate organisations I can think of have come to expect the existence of a unified mail and calendar application (a la Outlook) since well before 2004.

    Was there any half-way decent reason not to plan this particular item in to the ten-year odyssey of continuous failure that was LiMux?

    FOSS! It works for you, provided you are prepared to wait … oooh, let’s say ten years, guv.

  177. DrLoser says:

    But, back to Munich. This column is purportedly about Munich, isn’t it?

    Does oiaohm have any evidence whatsoever on the median age of a municipal desktop in Munich? Can he contradict my supposition that a goodly part of the inventory is easily modern and oomphy enough to run Windows 7?

    He has none and he cannot.

    That’s [keeping PCs 10 years is a terrible way to run a government] not obvious at all. Thin clients, PCs with few moving parts, can easily last that long if they don’t have fans sucking in dirt.

    What a pity that Munich didn’t go for a thin client solution.

    Since they didn’t, I’d say that the veracity of my claim is pretty bloody obvious.

  178. DrLoser says:

    Can’t do that and you know it. The EULA forbids anyone other than the holder of the licence to run the software.

    Perhaps you have misread the EULA? I quote from Windows 7 Home Premium, as a vanilla version:

    License Model. Subject to Section 2 (b) below, the software is licensed on a per copy per computer basis. A computer is a physical hardware system with an internal storage device capable of running the software. A hardware partition or blade is considered to be a separate computer.

    Surely the concept of “borrowing a neighbour’s PC” is not totally alien to you, Robert? That PC will come with a EULA, and that EULA allows you to log on.

    But you’re right. We must all keep on constant watch for the Black Helicopters of Redmond. The price of Freedom is Vigilance!

    M$ has prosecuted folks from loaning or renting the software.

    Is that, strictly speaking, relevant?

  179. DrLoser says:

    Any documents the malware has modified or spied upon have been modified and spied upon and Restore Points does nothing to fix the harm. Neither can it make an unbootable system bootable.

    I understood your point to be about Patch Tuesday, Robert. I can only correct one of your uninformed misapprehensions at a time.

  180. DrLoser wrote, “I suggest you borrow a neighbour’s PC and do the following”.

    Can’t do that and you know it. The EULA forbids anyone other than the holder of the licence to run the software. M$ has prosecuted folks from loaning or renting the software.

  181. DrLoser quoted, “System Restore does not affect any of your documents, pictures, or other personal data. Recently installed programs and drivers might be uninstalled.”

    Exactly my point. Any documents the malware has modified or spied upon have been modified and spied upon and Restore Points does nothing to fix the harm. Neither can it make an unbootable system bootable.

  182. DrLoser says:

    As for “simplicity,” I suggest you borrow a neighbour’s PC and do the following:

    WindowsKey system restore

    The GUI wizard will take you from there. Not quite as simple as the Linux CLI, I admit, but I suppose I could whip up a Powershell script to do the same thing.

  183. DrLoser says:

    That’s not simple at all. M$ has no way to make all your data and backups work.

    A completely unsubstantiated proposition which in any case has nothing to do with Patch Tuesday, Robert.

    They may be able to put the registry right for them but what about all the changes the user(s) have made in the meantime? Poof!

    Apparently I was correct in my surmise that you never looked into the use of System Restore Points in Easterville. They’re not about the registry (though they use it — a handy little tool for recovery is the registry). To quote the utility itself:

    System Restore does not affect any of your documents, pictures, or other personal data. Recently installed programs and drivers might be uninstalled.

    Now, it’s quite difficult to see how you can uninstall a recently installed program without uninstalling it, isn’t it? I regret that I have to point out this simple syllogism, but that’s basically what you need to do to remove a Patch Tuesday Patch.

    Furthermore, Patch Tuesday will often create a System Restore Point of its own, for precisely this purpose.

    Don’t take my word for it: just check your own copy of Windows 7, or in your case XP SP3, or … ah well, some cites work better than others, don’t they?

    Remember the stack of unbootable PCs of which I have written.

    How could I forget, Robert? That particular Crime Against Humanity is as a recurrent toothache to a woolly mammoth.

    And with just about as little relevance to Patch Tuesday.

    If M$’s software won’t boot, it can’t be used to rescue itself.

    Well, there are these things called Recovery Disks. (These days they’re normally Recovery Partitions — a handy little bit of bloat, I’m sure you’ll agree.)

    But since your cite makes no mention of Windows machines that cannot boot, I think we can safely conclude that this observation, too, has absolutely nothing to do with Patch Tuesday.

    You really didn’t try at all hard to make XP work at that school, did you?

  184. DrLoser wrote, “you simply roll back to a Restore Point that works.”

    That’s not simple at all. M$ has no way to make all your data and backups work. They may be able to put the registry right for them but what about all the changes the user(s) have made in the meantime? Poof! Further, I would bet not more than 10% of users of that other OS have any sort of backup or known-good “restore points”. Who’s to know when the malware got in and started biding its time? The worst malware spends its days spreading and then does damage gradually or in one fell swoop. Nothing M$ can do after the horse has left the barn can get it back. If the malware was a keylogger, for instance, all passwords/accounts may be compromised and M$ has nothing to do with those.

    Remember the stack of unbootable PCs of which I have written. If M$’s software won’t boot, it can’t be used to rescue itself.

  185. DrLoser says:

    I remember dreading Patch Tuesdays. I get shivers thinking about them… My Utopia doesn’t have Patch Tuesdays.

    My Utopia doesn’t have Patch Tuesdays either, Robert. I don’t imagine that anybody’s Utopia does.

    Sometimes the requirements of the real world in which we all live, and some of us work, just gets in the way of Utopia.

    Interesting word, though. Did you know that Sir Thomas More’s original polemic “Utopia” was in fact a satire? And that most writers since (say, Aldous Huxley) have used “Utopia” interchangeably with “Dystopia?”

    Nobody likes Patch Tuesday, Robert. But then again, nobody likes Heartbleed or Shellshock, do they?

    Now, here’s the thing. I’ve mentioned it before. Windows has this nifty feature called “Restore Points.” On those rare occasions (I promise you they are rare) that Patch Tuesday stomps on your machine, you simply roll back to a Restore Point that works.

    And then you wait for the various patches to catch up with themselves.

    This really isn’t that difficult for even the average Joe in the street, Robert, let alone a seasoned professional like you.

    Albeit a seasoned professional who hasn’t used Windows NT in many, many, years, and therefore, to put it politely, doesn’t know what he’s talking about.

    Incidentally, does this have anything at all to do with Munich?

  186. Further on Utopia. Last week M$ broke a bunch of systems with updates…

    I remember dreading Patch Tuesdays. I get shivers thinking about them… My Utopia doesn’t have Patch Tuesdays.

    see Microsoft yanks botched patch KB 2949927, re-issues KB 2952664

  187. DrLoser says:

    I’m sorry you cannot visit this website because it runs on FLOSS.

    I’m not bigoted, Mr Pogson. I am not one to dictate what technologies or products other people use. It’s of no consequence to me whether you spend $10,000 or $0 on setting up your site.

    By the same token, incidentally, obviously Microsoft Windows “works for Robert Pogson,” given that I and many others access it via Microsoft platforms.

    Curious that your choice of server technology has a more generous and all-encompassing view of the IT world than you do yourself, isn’t it?

  188. DrLoser says:

    Since when is being cheap and thinking that you have the right to dictate how much anyone who sells software can make from the product of their work constitute being “proud and brave” Robert Pogson?

    Silly question, olderman. Since 2001, obviously. I would give you a cite, but I’m afraid it’s not up to the standards required by Wikipedia …

  189. DrLoser says:

    Sorry DrLoser holding Heartbleed up and a flag will result in cases like POODLE being pulled against you.

    So, we should just keep quiet about massive security issues like Heartbleed and wait for them to go away, just because some bloke in northern NSW has access to Google Search and is in a tetchy mood and consequently wishes to lash out?

    Not, I would submit, a particularly constructive suggestion.

    If closed source was truly better they should be winning the common standard bug reporting game.

    It’s not a game, oiaohm. Only an idiot would call it a game. Commercial companies are general more interested in fixing problems than they are in beating other people to the tape when finding them.

    Notice the age of SSL 3.0 its 18 years old.

    Not only is software not subject to bit-rot, oiaohm; a fortiori, standards are not subject to it either. It could be argued – indeed, I would argue – that a standard with several years of history behind it and wide adoption is more likely to be functional, secure, and low on bug count than something knocked up yesterday by some know-nothing dipstick in a basement.

    Did Microsoft implement ssl 3.0 under IIS yes they did.

    And here we come to the crux of the matter, oiaohm. It’s eerily reminiscent of your dismal efforts on UTF-8.
    Do you know what the purpose of a communications standard is? It is to enable communication between machines sited at opposite ends of the wire. You can posit the finest ever protocol at one end of the wire … and if the other end of the wire doesn’t recognize it, you are shit out of luck.

    How would you propose talking to an internet client via HTTPS when the client does not implement SSL 3.1? That’s right – you use SSL 3.0. You have little choice.
    I will leave it to you, oiaohm, with your encyclopaedic knowledge of such things, to deduce which protocol I refer to when I use the term “SSL 3.1”.

    . DrLoser by this should we not presume that Microsoft own proved SSL is in worse disrepair than OpenSSL.

    In that you have given precisely no evidence whatsoever for this quixotic presumption, I am afraid the answer is no.

  190. olderman says:

    “For the few, proud and brave, non-FLOSS doesn’t work for a similar number/kind of people that FLOSS doesn’t work for, precious few. ”

    Since when is being cheap and thinking that you have the right to dictate how much anyone who sells software can make from the product of their work constitute being “proud and brave” Robert Pogson?

  191. DrLoser says:

    IIS has had on going security issues with plugins.

    Then don’t use plug-ins. Problem solved.

    Really, oiaohm, you do go out of your way to make things appear more complicated than they actually are. Besides, the question wasn’t “are commercial substitutes as secure* as their FOSS equivalents?” It was “is there a valid commercial substitute for Apache?”

    Considering that IIS is in use on 30% or so of web-servers worldwide, I am very much of the opinion that the answer is a resounding, “Why, yes, Robert, indeed there is such a substitute!”

    * A propos which, see:
    CVE-2014-3581,
    CVE-2014-0074, and of course the beloved
    CVE-2014-6278,
    inter alia.

  192. TEG wrote something, ” My Utopia does not include one crash per hour in the classroom.
    Which is obviously something not achievable with Linux.
    Or Robert Pogson.”

    I can’t parse that. What’s your point? The ONLY reason I switched to GNU/Linux was that I had a cluster of five PCs running that other OS and one of them would crash every class. It wasn’t defective hardware or defective software on a single PC but some random PC so taking one out of service wasn’t an answer. I had to get rid of the OS. I installed GNU/Linux from CD and never had a crash on them for six months. Those machines were in regular use for years after I left with no in-house GNU/Linux expertise to keep them running. They just ran and were a great asset to education.

  193. olderman wrote, “What does this have to do with the point I was making, Robert Pogson, which was that your contention that FOSS works for everybody is incorrect?”

    For the few, proud and brave, non-FLOSS doesn’t work for a similar number/kind of people that FLOSS doesn’t work for, precious few. Even IBM found that for 80% of users, GNU/Linux can be a “drop-in” replacement. In schools it’s even higher, probably more than 90%. For the folks that live with FaceBook and e-mail it’s 100%. So FLOSS works for people. For you to claim that’s not so is splitting hairs or something akin. Your applications could be written for FLOSS systems as are many turn-key systems that are part of some equipment. That they weren’t is someone’s choice, the developer’s, not yours and that’s certainly not any fault of GNU/Linux.

    In particular, M$’s OS doesn’t work, legally, for people who need more than N computers networked without paying extra. That’s just silly in an organization with any size, like almost any school. It’s also arbitrary. There was at one time no limit. Then it was 10, 15, now 20 at M$’s whim. An organization increasing its use of IT should find economies of scale not per-seat increased cost. Quoting “8”‘s EULA for OEM OS: “You may allow up to 20 other devices to access the software installed on the licensed computer for the purpose of using file services, print services, Internet information services, and Internet connection sharing and telephony services on the licensed computer.” The purpose of that is to sell server licences and CALs, an undue burden on individual users and organizations. That doesn’t work for most folks who don’t even know what a server is. I’ve been in several schools that did not have a server yet had more than the limit of PCs sharing files/printers. M$’s OS didn’t work for them.

    On a similar note, Sears rototillers don’t work for me. The one I bought worked well for a year and its transmission broke, the worm gear. It was repaired and broke again a year later. My next rototiller will be more expensive and more robust. Obviously Sears tillers work for city-folk with tiny gardens made of loam but some people need something more robust. This may be overkill. It has more than twice the horsepower at 2/3 the RPM (diesel engine), weighs 1000 lb and I can ride on it. It has 3 forward speeds and reverse in two ranges and can be used to drive a backup generator by belt-drive. Yes, you steer it with your feet… It will probably last my grand kids’ lifetimes. The picture is of the “economy” model. I’m going with the largest model with electric starting. It’s a little more costly but the freight on this thing will exceed the purchase-price.

  194. olderman says:

    “Show me the non-FLOSS equivalent of Linux or Apache or LibreOffice or VLC.”

    What does this have to do with the point I was making, Robert Pogson, which was that your contention that FOSS works for everybody is incorrect?

    Libreoffice and VLC are supported on windows, as is apache (www.wampserver.com/en) . My music packages are not supported on linux.

    SO it would seem that your statement needs to read

    “I claim that FOSS works for most.”

  195. That Exploit Guy says:

    My Utopia does not include one crash per hour in the classroom.

    Which is obviously something not achievable with Linux.

    Or Robert Pogson.

  196. oiaohm says:

    http://googleonlinesecurity.blogspot.co.uk/2014/10/this-poodle-bites-exploiting-ssl-30.html
    DrLoser also the fall out around SSL just keeps on coming. ssl 3.0 is broken at the standard level. So every one who as implemented ssl 3.0 is broken. Yes until the heart-bleed bug no one noticed. Did Microsoft implement ssl 3.0 under IIS yes they did. Notice the age of SSL 3.0 its 18 years old. This is a very big thing to come aware of software is not being reviewed and replaced in a timely manner for security.

    The existence of POODLE comes out from the openssl audit. So exactly why has Microsoft and others making SSL implementations not reported this issue so beating Google to the punch line. DrLoser by this should we not presume that Microsoft own proved SSL is in worse disrepair than OpenSSL. Yes a case of identical bugs in all implementations because it comes from the standard.

    Sorry DrLoser holding Heartbleed up and a flag will result in cases like POODLE being pulled against you. Were the closed source implementation require the Open Source implementations to find the standard bugs. This completely undermines the idea that some how Closed source is better quality than Open Source. If closed source was truly better they should be winning the common standard bug reporting game.

  197. oiaohm says:

    DrLoser IIS is not a good thing to raise. IIS has had on going security issues with plugins.

    3) “Where did my $500 music composition software go? Where did my $500 photo composition software go? Where did my $5,000,000 geological survey software go?”
    When talking a government with sub departments do all departments need these applications. Also DrLoser please know your topic. Geological Survey software for the expensive stuff is Linux/BSD/Unix based. Why you need a Linux/BSD/Unix cluster to be able to process the data. The Windows client is only 1 of the front ends. So yes deploy a Windows only network where is my geological survey software.

    Pure Windows or Pure Linux is effectively not usable to a government. It comes down to the split ratio. The most Windows you can in fact do in a government is about 90 percent because you need the other 10 percent of items like Geological Survey software that require something Unixish. DrLoser as normal you just pulled out something expensive that under mines your case. Please do your homework. The world is grey not black and white get use to it.

    Oh, and you can add Apache, LibreOffice, VLC, all that good stuff to your heart’s content. Not forbidden by the license! Won’t cost you a penny!
    It will cost you a penny because you will have a shorter operational life until you find yourself with no drivers for the latest Windows OS sooner than you will find yourself without drivers for Linux.

    Also installing all those application individually on Windows has another set of teeth. Each of applications on Windows has independent update software so you can now get users out of version sync so causing issues. So claiming it will not cost you a penny has no understanding of the issues you can trigger choosing Windows. Gooubuntu and LiMux has the repositories set to internal out box as well as auto connecting to the management solution. This is why Gooubuntu and LiMux is not much use to anyone outside. They would be like having redhat enterprise without subscription for repository but its also giving a remote group direct control of configuration of your computer. People working at Google don’t use the Google provided Gooubuntu, OS X and Windows machines for personal usage.

    Non-technical people with Windows desktops and laptops are baffled enough as it is. (Trust me on this one: I’ve spent a week trying to remove a dialog saying “this is a major upgrade. Please uninstall the previous version” simply because the average user cannot uninstall a program.
    This is why Google uses puppet. So end users never have to see this stuff under Linux, Windows or OS X. Even so Linux solution you have less update software to deal with so making having X and only X version deployed of some software way simpler and most cost effective to pull off.

    DrLoser Microsoft discontinued the stater products why the limited feature set caused more problems than it was worth when the limitation edition mixed with full editions. Rule one install full products or not a all. If Windows shipped with Libreoffice default that could be workable because that is a full product. But even shipping how would you solve the remain synced problem. Remember under windows user does a reinstall and machine is disconnected from management solution. The advantage of LuMix and Gooubuntu is a reinstall for the company results in connected to management solution by default so avoiding sync issues.

  198. DrLoser wrote, of Utopia, “mine is already here”.

    My Utopia does not include one crash per hour in the classroom. That’s what I had when I chose to go to GNU/Linux. My Utopia does not include further enriching a business with huge margins for a mass-produced product for which GNU/Linux is a great competitor. My Utopia does not include agreeing to connect no more than X PCs to my PC… My Utopia does not include errors in maths in spreadsheets retained for backwards compatibility. My Utopia does not include waves of malware taking over millions of PCs because the creator of the software had achieved monopoly….

  199. DrLoser says:

    Here’s my counter-proposition, Robert.

    Imagine a world in which you buy your $300 desktop or $450 laptop from a friendly local retailer. Imagine that this desktop or laptop comes complete with an OS that you really have no need to fiddle with.

    Imagine that said personal computer comes with a starter pack for a word processor that everybody uses, a spreadsheet that everybody uses, a free firewall that you don’t even have to think about, a top-quality Web browser …

    I had to go through serious contortions to imagine your future proposition, Robert. You don’t have to go through any contortions at all to imagine mine.

    Because mine is already here.

    Oh, and you can add Apache, LibreOffice, VLC, all that good stuff to your heart’s content. Not forbidden by the license! Won’t cost you a penny!

    Y’know, you can pick a guy of the street and ask him to choose between your rancid apocalypse and my representation of the real world.

    Guess which one he or she would pick?

  200. DrLoser says:

    Imagine a world where M$ and its “partners” disappeared tomorrow.

    Duly imagined, Robert. Although I personally see no difference between partners of the quoted kind and partners of the non-quoted kind. No matter. Duly imagined.

    There might be some weeks of disruption but the world would carry on without.

    Considering that you would knock out the immediate usage of 80% of desktops, 20% of Web Servers, and practically 80+% of corporate in-house servers and LOB applications …

    … Yes, you are absolutely correct, Robert.

    Ignoring the obvious catastrophic effect this would have on global commerce (hey, I’m not the sort of absurd fool who would wish this on the world overnight), I can see a few issues with this “some weeks of disruption” thing.

    1) Every last one of those LOB applications goes out the window. Start again from scratch. Hopefully in a CMMI sort of way you can recreate them using suitable FLOSS technology, but I would suggest that six months is a minimum. Going by the relatively small-scale experience of Munich, even ten years might not be enough.
    I know I would and I know everyone else could too. Therefor FLOSS works for everyone.
    2) Non-technical people with Windows desktops and laptops are baffled enough as it is. (Trust me on this one: I’ve spent a week trying to remove a dialog saying “this is a major upgrade. Please uninstall the previous version” simply because the average user cannot uninstall a program.
    2a) There will be a wailing and a gnashing of teeth if these people are forced onto a Linux distro. Short-term, at least.
    2b) Which Linux distro? It might take the world four weeks just to pick a default.
    3) “Where did my $500 music composition software go? Where did my $500 photo composition software go? Where did my $5,000,000 geological survey software go?”

    Oops. You’ll just have to wait, guys. Nobody said the Rapture is going to be easy.

    I know I would

    Duly noted; a given.

    and I know everyone else could too.

    Need I point out the difference between an anecdotal “would” and an unsubstantiated “could,” Robert? Apparently, needs must. The two modals are (infuriatingly) orthogonal.

    I “could” live with all four limbs chopped off. I “would” rather not. See my representative arguments above.

    Therefore FLOSS works for everyone.

    A completely unsustainable logical conclusion.

    But, hey, at least we’ve moved on from petition principia to a false synthesis!

  201. DrLoser wrote, “FLOSS does not work for me.”

    I’m sorry you cannot visit this website because it runs on FLOSS. I’m sorry you can’t read my text because it was generated on FLOSS. I’m sorry the web is a shrivelled place for you as it runs on FLOSS.

    Imagine a world where M$ and its “partners” disappeared tomorrow. There might be some weeks of disruption but the world would carry on without. I know I would and I know everyone else could too. Therefor FLOSS works for everyone. It’s in the licence:“You may make, run and propagate covered works that you do not convey, without conditions so long as your license otherwise remains in force.”

  202. DrLoser says:

    I do sincerely hope that I am not “putting words on your keyboard,” Robert. In an effort to avoid this practise, I have resorted to quoting you word-for-word. (Actually I almost always do this.)

    I claim that FLOSS works for everyone.

    And I claim that it doesn’t. Your claim is once again petitio principii, or in the vernacular, “begging the question.”

    Ev”er*y*one` (?), n. [OE. everychon.] Everybody; — commonly separated, every one.

    Hmm, Webster’s 1913 wasn’t much help there, was it? Let’s go the extra step:

    Ev”er*y*bod`y (?), n. Every person.

    Not wishing to fall into the trap of circular argument, an unfortunate fate that appears to have befallen you several times recently, Robert, I am going to set out my argument based upon an assumption.

    I qualify, in Webster’s terms, as “a person.”

    Therefore the set of “everybody” includes me as a member.

    Therefore the set of “everyone” includes me as a member.

    Therefore the statement “FLOSS works for everyone” can be easily falsified.

    FLOSS does not work for me.

    Which do you want to challenge, Robert? My assumption or my conclusion?

  203. DrLoser says:

    I say option #1 would be a viable solution to cure his pangs.

    Better still, Dougie, he could

    1) Pay an expert like you to build it for him
    2) Spend $500 or whatever it is for a license for something that makes his hobby something less than a coruscating horror of inadequate functionality and barely comprehensible command-line fixes. Outside the wonderful world of FLOSS, we evil types like to get stuff done. Even in our spare time.

    Point (1) was, of course, a cruel joke. I’m sure you could design, write, test and install an equivalent bit of FOSS software for $500, Dougie.

    I’m also sure that you would balk at the 0.0005¢s an hour for doing so.

  204. DrLoser says:

    Show me the non-FLOSS equivalent of Linux or Apache or LibreOffice or VLC. I’ve used what M$ offers and find it difficult/inflexible and certainly more expensive.

    Aha, a challenge on functionality! Bravo, Robert!

    1) Windows XP SP3. This one even comes free, as long as you pirate a copy and don’t care about the EOL! (Which makes it roughly the equivalent of any Linux desktop distro on the market, except that it will outlast most of them.)
    2) Microsoft IIS, as long as you don’t mind something that is both more secure and better loaded with features. And is more scalable. You can use asynchronous network comms (“interleaved”) on IIS; Apache might do that on a Windows platform. I haven’t looked. I doubt it.
    3) Leaving aside the Beast of Redmond, you can use Kingsoft or Softmaker. You can even work in parallel, with non-broken ODF (apparently a Redmond issue) and non-broken OOXML (apparently a LibreOffice issue). All for just $20!
    4) OK, I’m lost for an alternative to VLC. Quite how broken and limited do you want your media functionality to be? I could probably scrape something equally as pathetic up, had I only the need.

    And that wasn’t olderman’s point, was it? He claims that there are some things on a Windows/Mac platform that simply have no equivalent on Linux.

    As my list shows, the reverse is clearly not true, at least on the functional examples you provide.

    Oh, but you were talking about “Free as in beer,” weren’t you?

    Well, you gets what you pay for. And with the program suites I listed above, you get an awful lot of very easy to use functionality for very little cash outlay.

  205. olderman wrote, “Show me the FOSS equivalent of”.

    Show me the non-FLOSS equivalent of Linux or Apache or LibreOffice or VLC. I’ve used what M$ offers and find it difficult/inflexible and certainly more expensive.

  206. dougman says:

    Well, as it seems that OLDman is griping about the lack of music composition software equivalents in FOSS, perhaps instead of whining about it he should:

    1. Build his own software
    2. Take up a charity donation
    3. Quit trolling on this blog

    I say option #1 would be a viable solution to cure his pangs.

  207. olderman says:

    “I claim that FLOSS works for everyone.”

    And that is either ignorance or deception on your part. Robert Pogson. It doesn’t work from me or most composers using computers to engrave or compose music.

    Show me the FOSS equivalent of:

    http://www.finalemusic.com/?_ga=1.191850218.1188473800.1413650366

    and

    http://www.garritan.com/?_ga=1.121596616.1188473800.1413650366

    and

    http://www.soundsonline.com/Symphonic-Choirs

    This is just one example of many.

    Perhaps you wish to modify “everyone” to “many”, eh?

  208. oiaohm says:

    DrLoser an operation working on a model of replace when failed does not need a huge hardware list.
    However, since this solution has about the same hardware requirements as the comparable Windows solution
    About is the big word in this sentence. Problem is its not about. 64 bit Ubuntu only requires 1G of ram where windows 64 bit requires 2G of ram. Harddrive requirements 10G for Ubuntu 20G for Windows.

    That is not allowing for firmware issues. Linux kernel contains lots of detection and fixes for old firmwares.

    Even so cases of machines not having enough ram for standard Ubuntu systems they can be setup thin. Old machines can have failing hard drives without being a major bother to a Linux deployment. Boot from network with ubuntu means old machine does not need new hard drives. Windows is not really designed for running from network hard-drive.

    DrLoser sorry the response from Munich to the HP/Microsoft so called costs mentions this stuff. But of course you TMR guys would not be up on this stuff. Besides the HP report only compared XP to Linux because otherwise there was major hardware cost differences.

  209. DrLoser wrote, “Hewlett-Packard and Microsoft couldn’t obtain a useable hardware inventory”.

    Because M$ didn’t have a spy in the organization, you can’t blame oiaohm or Munich. Munich knows what it has and has published the savings. Live with it.

  210. DrLoser wrote, ““It works because it works, except in those rare cases where it doesn’t.””, putting words on my keyboard.

    I claim that FLOSS works for everyone. Munich and others have not yet made the effort to free themselves from all non-FREE software. That’s the only reason FLOSS doesn’t work for ~10% of Munich’s PCs. Read their reports. They got rid of all kinds of M$-only apps they didn’t need, consolidated others, switched others to web-applications and managed quite well with many FLOSS-only PCs. They probably will eventually do away with all of M$’s stuff. M$ and “partners” do not have any magic. The world can and does make its own software.

  211. DrLoser says:

    Quit begging the question.

    I wasn’t aware I was doing any such thing, Robert. However, as you know, “begging the question” is essentially the logical fallacy of “petitio principii.” Your conclusion:

    “Clearly it [LiMux] works for most employees except the few still using that other OS to run certain applications.”

    … is perilously close to petitio principii. “It works because it works, except in those rare cases where it doesn’t.”

    The City of Munich has not ground to a halt under GNU/Linux

    I never said it did, Robert. In much the same way that, to your credit, you have never claimed that any of the hundreds of thousands of other municipalities running Windows have “ground to a halt.”

    No argument there.

    … which took years to phase in.

    Ten years, to be more or less precise. That’s rather a long time to produce what oiaohm claims is a simple internal disk image, isn’t it?

    You have to wonder what a proper IT organisation would do with all that time and money.

  212. DrLoser says:

    DrLoser the ages of Munich hardware was in one of the Munich IT offices media reports disproving HP cost claims.

    What, so the majority of the City Gov is still running on the same machines as they were in 2004, just before the migration? It would be tragic if so. But I notice that you don’t substantiate your claim with a cite, and I will therefore kindly suppose that you have misread your source, oiaohm.

    I don’t know the number either, but I can offer this, translated:

    The cost for the hardware can not listed because the City of Munich has no inventory which would allow hardware requirements to be determined. However, since this solution has about the same hardware requirements as the comparable Windows solution — €50 million — it can be assumed that the hardware cost has no impact on cost comparison.

    If Hewlett-Packard and Microsoft couldn’t obtain a useable hardware inventory, how on earth can you do so, oiaohm?

    DrLoser LiMux and Goobuntu are not different to a internally Windows Image for Company usage.

    Once again, something you cannot possibly know, oiaohm.

    Because Google simply does not release Goobuntu.

    Perhaps everybody in Google has a pet name for their chair … you know the sort of thing, “goo-cushiony-wushiony.” No? Well, if not, why do they pick a single item such as a vanilla (Ubuntu) distro “company image” to give a pet name to?

    The real answer is LiMux is a internal image solution that has no vendor.

    No vendor? Really? Would you care to tell me what IT@M, under director Karl-Heinz Schneider is, if it isn’t a sole vendor? It’s touted as a”städtischen IT-Dienstleisters” or “Municipal IT Service Provider,” but it might just as well be a sole vendor as far as I can see.

    And what sort of idiot spends ten years and €50 million on an “internal image solution?”

    So no different to Windows custom install image.

    The cost of a Windows custom install image is maybe six man months of integration and system testing: I’d say €50,000 would do it.

    A factor of 1000 in cost indicates to me that there is quite a substantial difference, oiaohm.

  213. oiaohm says:

    DrLoser the ages of Munich hardware was in one of the Munich IT offices media reports disproving HP cost claims. The acquirement comment about NT 4.0 was covered in the write up about costs.

    Apart from the obvious fact that this would be a terrible way to run a government IT department …
    You might say obvious but it the way the Munich IT department has successfully run. So even that it seams terrible it manages to work.

    DrLoser LiMux and Goobuntu are not different to a internally Windows Image for Company usage. IT Policy to connect to network demands company images deployed.

    The one is supported by a single vendor with absolutely no third parties available.
    DrLoser want to think LiMux is this. What is not.
    The other is supported by a single vendor with a massive number of third parties to choose from.
    When in reality LiMux is not this either.

    The real answer is LiMux is a internal image solution that has no vendor. So no different to Windows custom install image. Of course you get no direct support on a Windows custom install image. The vendor on current LiMux images is Ubuntu so there is a third party option. LiMux is in the same camp as zentyal you can buy commercial support from ubuntu if you have the budget. Munich does not have the budget to pay for Ubuntu support yes this is less than Windows licensing costs. So if Munich cannot pay for Ubuntu how can they pay for Windows????? Ubuntu is the primary vendor that they cannot pay. That has massive number of third party vendors who can be chosen from who support Ubuntu. So claiming less vendors than Windows is possible. Claiming no third parties is false.

  214. ram says:

    Mr Pogson wrote: “Thin clients, PCs with few moving parts, can easily last that long if they don’t have fans sucking in dirt.”

    That is right! They have plenty of computing power and are energy efficient as well. Recent (last 5 years of so) Shuttle fanless Atom boxen have multiple AMD64 architecture cores, run quite a few simultaneous threads, and have a very good ratio of computing power (say MFLOPS) per Watt. Now Intel has fanless NUC boxen, and other companies have entered the competition as well. All these boxen use under 15 Watts. That is less than 10 percent of a Microsoft PC which lowers the electricity bills directly and by reducing the waste heat that more often than not must be taken away by office airconditioning systems.

    Linux thin clients have been show time and time again to reduce IT related electricity consumption by more than 95 percent in tropical countries. That is a big savings!

  215. DrLoser wrote, “LiMux clearly doesn’t work for a large number of Munich city employees, and they have no choice in the matter.”

    Quit begging the question. The City of Munich has not ground to a halt under GNU/Linux which took years to phase in. Clearly it works for most employees except the few still using that other OS to run certain applications.

  216. DrLoser wrote of Goobuntu, “nobody outside Google can even download Gubuntu, can they? Which I would have thought is a strange position for a champion of the Four Freedoms to be in.

    Simply put: if oobuntu works for Google, then, fine. People who work for Google make that choice.”

    That’s silly. Google is using oobuntu internally so they configure it and use it. They don’t have to distribute the source because they are not distributing the code outside. It’s nothing to do with Free Software except anyone, including Google can use it.

    SJVN did an expose of Goobuntu a while back: “Goobuntu is simply a light skin over standard Ubuntu.” In particular, Google uses the latest long term support (LTS) of Ubuntu. That means that if you download a copy of the latest version of Ubuntu, 12.04.1, you will, for most practical purposes, be running Goobuntu”

  217. DrLoser wrote, of keeping PCs 10 years, “Apart from the obvious fact that this would be a terrible way to run a government IT department”.

    That’s not obvious at all. Thin clients, PCs with few moving parts, can easily last that long if they don’t have fans sucking in dirt. I imagine government is not too different from schools in that CPUs idle while folks sit around or type or click so speed is not an issue, just whether or not there is enough RAM to hold applications and data. I was in schools that had 1gB per PC ten years ago, plenty for a thin client or a clerk with just a few simple applications running. The big stuff is likely to be on servers anyway.

    The Wintel treadmill was invented to sell PCs and licences, not to do IT right. Munich has shown it wanted independence as far back as 2003 or so and that attitude may give them the courage/clarity of thought to keep hardware for its entire life.

  218. DrLoser says:

    DrLoser large section of the computers at Munich are over 10 years old.

    Apart from the obvious fact that this would be a terrible way to run a government IT department …

    You know this, how, oiaohm?

    I’m going to go out on a rather sturdy limb and suggest that you are making it up out of thin air.

  219. DrLoser says:

    Windows NT at its core is a cleverly arranged sequence of bits.

    See, I can do meaningless syllogisms with the best of them, oiaohm.

  220. DrLoser says:

    Like Gubuntu is for google.

    But nobody outside Google actually uses Gubuntu, do they?

    In fact, nobody outside Google can even download Gubuntu, can they? Which I would have thought is a strange position for a champion of the Four Freedoms to be in.

    Simply put: if Gubuntu works for Google, then, fine. People who work for Google make that choice. Indeed, apparently, they can choose other operating systems, up to and including OSX, as far as I am aware.

    LiMux clearly doesn’t work for a large number of Munich city employees, and they have no choice in the matter.

    Is this lock-in, or is it freedom? You get to choose!

    Be honest, now.

  221. DrLoser says:

    Ignoring the rest of the standard oiaohm wall-of-gibberish:

    LiMux at its core is standard distributions.

    And Windows at its core is standard NT.

    The one is supported by a single vendor with absolutely no third parties available. The other is supported by a single vendor with a massive number of third parties to choose from.

    Guess which one is LiMux and which one is Windows?

  222. oiaohm says:

    DrLoser large section of the computers at Munich are over 10 years old. Windows 7 might work fine on computers acquired since 2007 but the problem is that is not the Munich fleet of computers. Linux hardware compatibility equals longer service live before replacement of computers. €3.1 million is to replace the machines that were made before 2007 that are currently in active usage. Installing Windows 8 moves that date forwards so requiring more hardware replaced.

    LiMux at its core is standard distributions. Like Gubuntu is for google. So LiMux is self does not need to be supported by vendors. So as long as vendor supports Debian and Ubuntu that support also applies LiMux due to the fact its is just a custom installer of those distributions. Support companies that know how to handle Ubuntu and Debian can also handle LiMux its really no difference.

    If you want to be “interoperable” with all the other city administrations in Germany, not to say the rest of the world, would you choose a set of formats and applications that none of them use? Or would you … horrors … I cannot bring myself to expose you to the obvious conclusion …
    This is because you have not read the IBM white paper on the topic. 90 percent plus of all business and government operations is internal. Munich did not get rid of every copy of Windows. 10-20 percent Windows is enough to meet interoperability needs. Leveling up the internal that everyone is using the same office suit internally does reduce issues.

    What do the hardware procurement guys in Munich do for a living? Drive around in a horse and cart begging for cast-off Pentium V machines?
    What IT hardware procurement guys that don’t exist in Munich. There has been no budget to go out and buy new hardware in any form of volume in Munich for over 20 years before the migration to Linux. When hardware dies it is replaced this is the Munich operation model. Before the Migration to Linux Munich was having to run on all second hand acquired machines and licenses. The migration to Linux freed up budget for them to buy new hardware but they are only buying it in small lots following the rule of replacing failed. Heck Munich was that tight they were even buying second hand Windows licenses of the types Microsoft had not locked to machines.

    The Munich IT department is going to fight tooth and nail to avoid being pushed back into that existed before. The olds of Munich reversing is insanely low. Major will have explain where the money to fund the IT will come from to go Windows.

    This is just the case the arguement that Munich would reverse as soon as government changed was always never taking account of how big of a mess Munich was before the Linux migration. Think of this year 2003 Munich had to acquire Windows NT 4.0 to run internal software. Munich had a completely failed IT system. Munch now has a IT system that at least somewhat works and is buying some new hardware. Munch IT system now is possibly healthy. A full review may find areas to improve but most likely will find that Munich does not have the budget to return to Windows.

    Please note the Migration to Linux has also allowed the computers for “interoperable” operations to be running the latest Microsoft software. Munich was not “interoperable” with other cities before the Linux Migration because they were running out of date software compared to everyone else. Every metric is better now than it was.

    DrLoser what you have to explain is how Munich will have the money to maintain windows to current version if they convert the lot back. Or maybe Munich should just stay with the 80-90/20-10 split between Linux/Windows. This current combination Munich has the money to support. Old Windows machines that no longer can run current version of Windows get a second life in the LiMux system operations. Hardware from 2004 in LiMux is not a problem. Life cycle of a computer in Munich is 10+ years on current operating system being either LiMux or Windows.

    Yes this is a case they don’t have the money. What the Major can do is limited by what the money allows.

  223. DrLoser says:

    This is, of course, seen through the prism of a posting by FSFE; not exactly an independent journalistic source. Even so:

    One thing is clear: the evaluation needs to take vendor independence and interoperability into account instead of limiting its perspective on cost and user satisfaction, since the former have been central arguments for the switch in the first place.

    Sans citation, I assume this is clear to the FSFE and not necessarily anybody else.

    But it’s a fascinating perspective, isn’t it?

    The hell with cost and user satisfaction! Who cares about cost and user satisfaction? Not the FSFE, apparently.

    Let’s just fall back on vendor independence and interoperability, shall we?

    Well, not being the FSFE, and although I personally rate cost and user satisfaction quite highly, I’m prepared to tackle both of these issues.

    1) How many vendors support LiMux? Is it more than ONE?
    2) If you want to be “interoperable” with all the other city administrations in Germany, not to say the rest of the world, would you choose a set of formats and applications that none of them use? Or would you … horrors … I cannot bring myself to expose you to the obvious conclusion …

    Loved the bit about the cost of a hardware upgrade, though. €3.1 million? A bulk purchase of Windows 7 licenses will work fine on just about any x86 box bought since, er, 2007.

    What do the hardware procurement guys in Munich do for a living? Drive around in a horse and cart begging for cast-off Pentium V machines?

Leave a Reply