Who is to blame for the fact that Windows crashes a lot?

This page is a response to Walter Mossberg's article on the Wall Street Journal's website titled I'm Tired of the Way Windows Freezes!, dated September 30th, 1999.

I believe that the problems Mossberg writes about are all due to three basic causes:

1. WINDOWS DESIGN FLAWS: Microsoft has knowingly (and by design) built their OS's with an architecture that is easily rendered unstable by allowing application software to modify OS components.
2. BAD SOFTWARE: Software vendors write their software with the same assumption they had when they were writing DOS software: my application is the only one running on this PC. They alter shared components and consume unreasonable amounts of resources, with the result that they don't "play well with others." And an awful lot of commercial software has really bad memory leaks, often exacerbated by badly written video drivers. In other words, BAD SOFTWARE. This is further exacerbated by hardware manufacturers who lard up their default installations with software that users don't need and will never use, but don't know how to remove.
3. FRAUD: The computer industry has sold the public on the fraudulent idea that computer users need have no more knowledge than that in the empty heads of angels being born.

   "It's Intuitive! It's Easy!"

   It's a FRAUD.

   In the current state of technology, users need to have some understanding of how their computers work and how to maintain them. But no one gets any training any more because "It's Intuitive! It's Easy!"

My commentary on the article follows. For each of the problems, I've included a reference (e.g., (2) or (1)) to which of the causes above is in play there. You can assume that (1) is a contributing factor in all cases, but especially where I explicitly cited it.

    THE OTHER DAY, I was sitting in my office when it happened:
    the dreaded Microsoft Windows Blue Screen of Death. I wasn't
    doing anything special, just trying to print a document from
    Microsoft Word 97 running on a new, powerful Hewlett-Packard
    Pavilion PC.

First off, my experience with HP Pavilions is that they ship larded up with all sorts of unnecessary software, much of it running at boot. It doesn't suprise me that such a machine would be unstable. (3)

    But suddenly, the screen turned entirely blue, and this message
    appeared: WINDOWS. A fatal exception 0D has occurred at
    0028:C000B25A in VXD VMM (01) + 0000A25A. The current
    application will be terminated.

This kind of error should never happen with the standard Win9x installation. In my experiences, the Virtual Memory Manager VXD never causes problems except when third-party software (particularly scanners) is doing something to modify the in-memory VXD image at boot. It's quite common for drivers designed for Win95 to produce precisely this kind of error when installed on Win98 because they were written for a different VXD than the one they are currently running with. (1) (2)

    So here are some excerpts from my week with Windows. These
    experiences occurred on computers from IBM, Dell, Compaq and
    Hewlett-Packard. Most were new PCs with at least 64 megabytes of
    memory.

First of all, 64MBs is the minimum for multi-tasking with modern applications under Windows. Twice that RAM very often eliminates all kinds of instabilities. (2)

Furthermore, those four manufacturers tend to ship PCs that are, out-of-the-box, borderline stable. That is, change nothing and they'll run like a dream. But add a piece of hardware or any memory-resident software, and the whole thing can break down because of incompatibilities with what's installed by default. (3)

As a matter of course, I strip any new PC down to the minimum running software at boot. That means cleaning out the Startup group and the RUN line in the System Registry. It means eliminating all sorts of unnecessary processes that soak up system resources and can potentially conflict with other applications. On an ideal machine, the PC will boot with the My Computer Properties showing 95% or more available resources. Devices like Winmodems and Windows Printing System printers (any laser under $400 or any inkjet under $200) are important offenders here, because you can't remove the processes that support those devices from memory without disabling the devices (which use your CPU in place of processing power installed in the device itself).

My home computer boots with 98% resources available, and it doesn't crash. Ever. My primary work computer boots with 94% and it crashes only when I do something really nasty during software development/testing. It's very uncommon for me to be required to re-boot any of the computers I maintain.

    In almost every case, I was
    running multiple programs
    when the problems occurred.
    But this wasn't an unfair test:
    Microsoft touts Windows as being able to handle such multitasking.
    And all of the programs I was using are well-known. They included
    Lotus Organizer, America Online, Netscape Navigator, MusicMatch
    Jukebox (a digital music player) and Microsoft's own Word, Internet
    Explorer and Outlook Express e-mail program.

AOL and Netscape are two notoriously unstable and ill-behaved applications. Netscape crashes on stable systems and is very sensitive to RAM. I wouldn't recommend running it in less than 64MBs by itself and less than 128MBs in a multi-tasking environment. (2) (3)

I find MS's programs to be basically stable, but, on the whole, you need more than 64MBs of RAM to have them "play well together."

    Sept. 22, 4:45 p.m., San Francisco hotel room: I try and wake
    up my IBM ThinkPad 240 laptop from its battery-saving
    hibernation" state. The computer never revives. A reboot is
    required. 

Well, I learned long ago that "hibernation" modes in laptops just don't work. Recent experience has not changed my mind on that. Well, actually, my 2nd laptop, a 386SX25 Zenith, had a working hibernation mode -- I could just close the lid and re-open it at will. But that laptop died in 1995, and I've not seen one since that actually worked.

My advice: turn off all power saving features and never use hibernation.

This is another example of the computer industry perpetrating a fraud on the public. (3)

   Sept. 23, 9:55 a.m., hotel room: I boot up the ThinkPad, but the
   sound doesn't work. I turn "Mute" on and off a few times. The
   sound resumes. 

Not enough information on this one to comment. Probably something to do with hibernation mode. It's often quite difficult on laptops to know whether you've turned the damned thing off or put it in some kind of hibernation mode, since the power switches function for both. (3)

    Sept. 24, 1:43 p.m., same hotel room: While I check e-mail, the
    whole PC slows to a crawl. The cursor freezes and unfreezes. I
    reboot. 

Perhaps a power-saving problem? Hard drive went to sleep? Or, perhaps there's a communication problem, and it's causing problems with the serial port, and, hence, the modem. That in turn might block some message queue and lead to this kind of symptom. My bet is that logging off and logging back on would have solved the problem without a reboot, but there's not enough information. (1) (2)

    Sept. 25, 10:11 a.m., United Airlines Flight 230: I am sitting
    next to Dan Gillmor, technology columnist at the San Jose
    Mercury-News. Between us we have decades of experience
    with PCs. Yet we are unable to get our two ThinkPads to
    exchange a file via their infrared ports without fear of
    disabling other functions. 

There's no real standard for IR ports and it's really not completely supported. I don't know anyone who uses their IR port, to be honest. Should it be this way? Well, I don't know. Who cares? Not me. The number of times this would be convenient for me I could count on the fingers of one hand. On the other hand, I might not appreciate it until I had it.

It's an underdeveloped, untested technology. Who should be surprised that it doesn't work? (3)

    Sept. 26, 9:30 a.m., home in Washington: My son's Compaq
    freezes up when he tries to run both AOL and Netscape
    Navigator -- a frequent problem. A reboot is required. 

Answer to problem: AOL. Don't use it. It is garbage, buggy, unreliable software. And the service itself stinks, bombarding the user with ads, while providing substandard tools for most Internet functions. As I said before, Netscape is notoriously unstable, in just about any version. Mixing these two unstable programs together is almost guaranteed to cause problems. (2) (3)

    Sept. 26, 1:54 p.m., home: On my month-old Dell Dimension,
    with an obscene 256 megabytes of memory, I am suddenly
    unable to print my monthly calendar in Lotus Organizer. I get
    this message: Painting Screen. You have insufficient system
    resources. Close another application and try again. System
    Error: (87). The parameter is incorrect. I close another
    program. 

A memory leak somewhere -- it has ABSOLUTELY NOTHING TO DO WITH THE AMOUNT OF RAM INVOLVED (it says "insufficient system resource," which means it's out of stack space). Could be caused by Lotus Organizer or could be caused by video drivers. Or the combination of the two. If it goes away when you close Organizer, you know what produces the symptom, but not really how to solve it. I'd contact Dell for the latest video drivers and check with Lotus to see if they know anything about the problem. The websites of both would be the best place to start an investigation. (1) (2)

    Sept. 26, 1:55 p.m., home: I try again to print the calendar,
    only to get this message: Printers Folder. There was an error
    writing to LPT1: for the printer (HP DeskJet 722C): There was
    a problem printing to the printer due to an unknown system
    error. Restart Windows, and then try printing again. A reboot
    is required. 

HP Printer Drivers are notoriously unstable. An IT friend of mine was once told by a Microsoft tech support person that the single largest source of calls for them is HP Printer Drivers. HP actually has excellent technical support and they are constantly updating the drivers. (2)

I fear, though, that the problem can't be solved that way. I'm pretty sure that the DeskJet 722C is one of those "brain-dead" printers that I spoke of above that depends on your CPU to do all it's "thinking" for it. That means that there are probably four running processes (perhaps loaded as System Services in the System Registry) that support that printer. I just wouldn't waste my money on these cheap printers. What you save in $$$ you quickly use up in time and headaches. (3)

    Sept. 27, 8:45 a.m., home: My wife receives, via AOL e-mail, a
    document she needs for a 10 a.m. meeting. She can't open it,
    read it or print it. Neither can I. She tells her colleagues just to
    fax her such documents from now on. 

Don't use AOL. Its mail reader does not properly support the most common MIME standards. Let me repeat this: AOL IS GARBAGE. Its software is badly designed and unstable. It is slow and larded up with advertisements. It is an exercise in masochism to stay with AOL once you've learned the basics of e-mail and the Web.

Get a real ISP and use a real e-mail reader. (2) (3)

    Sept. 28, 4:30 p.m., office: I can't print an e-mail from Outlook
    Express, running on a new HP Pavilion PC. I get this
    ridiculous error message: Could not set the colSpan property.
    Invalid property value. Solution: Enter a value between 1 and
    1000. Huh? A reboot is required. 

"colSpan" suggests that Outlook Express is using HTML to render/send e-mail, since "colSpan" is an attribute of cells in HTML tables. OE defaults to using HTML for sending mail. This is a mistake. Turn off HTML.

Or, get a better mail reader. OE is barely adequate, in my opinion. (2)

    Sept. 28, 6:07 p.m., office: My week of keeping records ends
    when the Pavilion suddenly shuts down while I am trying to
    copy some files -- including a draft of this column -- from its
    hard disk onto a Zip disk, so I can take them home. A reboot is
    required, yet again. 

I wonder if it's a parallel port external Zip drive? In combination with the HP inkjet printer, this is a recipe for disaster. The two probably cannot coexist on the same parallel port, because both demand ownership of it. (3)

If it's an IDE Zip drive, then it's a more serious problem. (2)

©1999-2000, David Fenton Associates. Created October 1, 1999. Last updated July 5, 2000.