Why I Hate Wintel

12/30/99

So I bought a new computer recently. I had ordered several packages to go with the computer, and as it turned out the computer arrived first. So to begin with I had only the basic CPU and a Red Hat Linux install CD.

The first thing I noticed was that the Linux CD would not boot...I figured it was just the wrong format or that the CD writer had not been set up properly -- however later I observed that it seemed to be a problem with the CPU's internal L1 cache -- when it's enabled, the RH CD will not boot, but it will when disabled. The Win NT CD will boot either way, however.

So lacking Red Hat power, I proceeded to install DOS and then Win 3.1 on the HD. The installs were mostly happy, except that there were random lockups at times, and to start with Windows would lock up when I move the mouse pointer to the left or top of the screen (after turning the picture upside down)...this problem eventually went away, however (for no apparent reason). The only problems I did have were the occasional corruption of one of the program groups (again for no apparent reason).

With DOS installed, I proceeded to mount the Red Hat CD, make a Linux startup floppy, and try that out...Of course it didn't work -- gave an error about trying to delete the idle process (pid 0). So I went about setting up the modem using Win 3.1. After a good deal of jumper messing, I got it installed and working (although, for whatever reason, the 56k modem took a stupidly long time to connect -- 45 seconds or so, whereas my Mac 28.8 modem connects in about 10 seconds). Also, I was never able to get 56k speeds, or anything close, but that may have been due to my phone lines.

Later that day, I talked to knowledgeable folks about my Linux problems, and they suggested creating a new startup floppy (because the original could have been defective). So I used the Mac's Disk Copy utility to make a copy. It found no errors on the original, and created the copy. Then, to my utmost surprise, this second floppy did indeed start up happily -- why two supposedly identical disks could behave differently was a mystery. However, after much more messing around (which I'll get to eventually), I determined that Linux, or perhaps my CPU in general, has a motto of "If it doesn't work, wait and try again." I now believe the reason the first floppy didn't work was that I didn't shut down the computer completely and then restart from scratch....while copying the second disk, I had the Wintel computer shut off, and it had it's necessary downtime to reset whatever needed to be reset internally (as we shall see, this question of when the computer is actually starting fresh is one of the most randomizing problems I've dealt with).

So I proceeded to attempt a Linux install. The machine hung itself a few times during the install (first graphical, then text-based), but eventually I did get a successful Linux install. It started up all right, and booted me into the Gnome logon (since I had told it to do that automatically). This worked, and in general Linux was a happy experience. However, to start with, it started X in VGA (16-color) mode, which made graphics look horrible and made Netscape almost unusable and full of error messages. Also, I would at times get random hangs from the system -- usually vertical lines on the screen, often having a color similar to whatever was on the screen currently. After some playing with the Xconfigurator, I got a 1024 x 768 resolution to work -- fixing the color problems and noticeably speeding up the OS. I really would have preferred 800 x 600 resolution, but I could not get the monitor/card/CPU to sync at that rate. Even at 1024 x 768, I still had (infrequent) lockups, this time usually with diagonal lines across the screen, instead of vertical. Also the logout command always caused the screen to hang immediately, so the X server had to be killed with the ctrl-alt-backspace key combo. Finally, the mouse pointer in 1024 x 768 mode turned into a large, mostly black square (about the size of Netscape's "Back" button), with the real cursor in the top left of this box. ... Upon restart from a system crash, FSCK would always report that deleted inode 749900 had zero d_time, and usually this was the only error found (I have no idea what the significance of that is, but I give it in case someone else knows about it). But, in general, the system was functional and happy.

A few days later, I received my copy of Windows NT. Oh happy -- or so I thought. I set about installing it, and ran into the first of many problems. First of all, NT didn't like the way my partitions were set up, and I eventually had to delete my Linux setup to get it to work (which I didn't mind, since I hadn't customized it too much). As expected NT setup overwrote the HD master boot record (MBR) with it's own startup code (replacing Lilo from Linux). The problem is, it didn't do it correctly, so it only CORRUPTED the MBR, and the HD would hang with "LI" displayed (evidently some remnant of the LILO prompt). I wiped all partitions from the disk, returning it to a state of what I thought it was when I first started it up, but still the "LI" remained. After more knowledgeable assistance, I discovered the command fdisk /mbr, which wipes out the MBR. I used NT's setup to create 3 disk partitions for NT (500 MB NTFS system, 2047 MB FAT, and 2047 MB NTFS for programs), and later used its graphical program to create an extended DOS partition, which I intended to install Linux on later. One evil thing about NT: It only recognized my HD as 8064 MB, instead of ~10 GB, which it actually is. It seems that only Linux, the poor-man's public use FREE OS, is able to correctly address my (not so) large disk drive. Just great!

Anyway, after a couple install attempts, the "text-based" NT install completed, and I restarted into the graphical interface, only to find that the graphical setup stopped in Phase 0 with a message that the internal setup data structures were corrupted. I was baffled by this for a while. I restarted a few times, and (to my utmost surprise again) after awhile the message went away and setup continued. Stupid randomness! At this moment, I'm wondering if it didn't expect the CD-ROM to be in the drive, but that's a longshot for an explanation.

Setup took a while and a good deal of user interaction, but finally it was finished, and I restarted to find the familiar NT login prompt. Yeah!! It was a little slow, I thought, so I enabled the CPU's cache features, and speed improved dramatically...I proceeded to install Service Pack 4, then some programs and such, and then called it a night at about 5 a.m. (as I remember it, there was an error that seemed prevalent when I quit, but it didn't reappear after I got up the next day, so it was either a random error or another case of "if it doesn't work, wait awhile."

The next day, I received some dictation software by mail. I went about installing and customizing it, and at what seemed to be the very last step of customization, I got a stop message, something seeming to have to do with the NTFS system. I restarted as coldly as I thought I needed to, but the error remained. Now, dumb me, instead of following my principle of "wait awhile", I immediately grabbed my ERD (emergency repair disk) and set about "repairing" the installation. I noticed that a lot of files had apparently changed, and then remembered that it was probably undoing the effects of Service Pack 4 (SP4). I *thought* I had remade the ERD disk after installing SP4 -- so either I was wrong, or the NT setup programs is even stupider than I had thought. Either way, after "repairing" the installation, I started up, and got some random lockup problem. I did a more thorough repair, making sure to verify the system configuration, and this time it did start up, but it couldn't log me onto the system (error C00000DF). Later that day, I looked up the error, and found it to be an error with the security subsystem...it seemed to be related to reverting the SP4 system to the original NT 4 setup...apparently some necessary part of the registry was not reverted as needed. I followed some online instructions about modifying the ERD (I modified it on a Mac, if that matters), and then reran setup as it instructed, and it did appear that the correct files were copied from the floppy. But, on restart, the logon process generated a stop code (incorrect termination of program -- 0x80). Some part of this evil problem still remained. I then tried a more troublesome fix, initially telling Windows to install over top of the old installation -- this worked for the text-based mode. When I got to the graphical mode, again I ran into the "corrupted Setup data structures" error msg, and again after a while it randomly went away. But this time graphical Setup did not complete -- it ran into the error "Cannot initialize security subsystem" -- apparently my security problems were unfixable. And don't you just love the descriptive error messages!!!

As a near-desparate effort, I attempted reinstalling NT in a companion directory (on the same partition), but this generated stop codes again, and I left it at that for the moment.

I then went back to attempting to install Linux. To my surprise (as mentioned before), with the internal CPU cache disabled, the RH CD would boot, but to begin with even the CD kernel crashed. This is when I abandoned most of my ideas about corrupt floppies and such causing crashes. I left the mess for a while, and when I tried again after restarting, surprisingly (but it was getting less and less surprising) it worked, and graphical setup started. But now Linux got stuck at the partitioning screen. It didn't like my extended DOS partition, so I deleted it (it was unused anyway). But I could not create the recommended partitions, because "the boot partition is too large". Now I tried resizing the supposed boot partition (/boot) down to 1 MB, but the error remained. I later learned (as I had remembered vaguely) that the Linux boot partition had to be within a certain amount of space on the HD physically -- any farther "up", and the booter in the MBR wouldn't recognize it. I tried creating other partitions, and sometimes they worked, sometimes they didn't (at times the error "not enough free space" would appear, but I believe that error message was in error). What annoyed me was that I could not change a partition from DOS/NT FS to Linux FS. If this had been possible, I could have used one of the 3 NT partitions as the Linux boot partition (possibly). But the RH installer would not allow that, and so I was stuck. I didn't actually try installing any further on one of the partitions it said it could create, because from what I can tell all the partition programs are evil and half the time don't know what they're talking about. NT and Linux and DOS all disagree. I've had a 500 MB hard drive, and 8 GB hard drive, and a 10 GB hard drive, depending on which program you believe.

So, now I have 2 installations of NT on my computer, neither of which will work, and no partition space available for Linux. In other words, in all this work, I have gained absolutely nothing, except for knowledge. Now I know why I've always liked Macs -- because they WORK. Sure, they crash, and the OS's memory architecture leaves something to be desired. BUT, when the Mac crashes, almost inevitably a restart fixes the problem. On this POJ Wintel box (POJ = Piece of Junk), a restart usually only confuses the human debugger, who thinks "I restarted, therefore the system should be returning to its original state" -- which is almost certainly NOT the case. I now understand why I've always heard folks talking about reformatting their hard drives every other week (or so it seems) -- because Wintel OSes, from my perspective, are evil and tend to corrupt things and make it very difficult to recover. One wrong move, and "bye bye OS." To top it all off, this apparently almost always means "bye bye data" as well, because application programs are tied into the OS through the registry, and Linux requires a partition reformat for installation.

Now, this setup process has taken entirely too much of my Christmas break -- it has sucked time out of my life. I expected it to take time, but I assumed it would be in _setup_, not in setup, repair, replace, debug, resetup, debug, resetup, and end up back where I started -- with a dead computer...450 MHz of worthless CPU. However, I have been in this debugging mode for less than a week, and I have already learned a lot. So the time is not completely wasted (just 95% so). Perhaps one day I will understand the great mysteries of Wintel and actually understand why my computer operates as it does -- I will cease to see errors as random and be able to associate them to a certain cause. At times, though, this debugging has seemed COMPLETELY random...which, I guess, is typical of debugging. But, until I start seeing the stable operating systems I expected to see, operating systems that are the leaders of the information revolution that I once believed they were -- I am going to continue my stance and say that "I HATE WINTEL."

- - - -

Your comments are welcomed and encouraged. If you are smart in these matters and can debug my problems, I would be very grateful for your help. If you are just a passerby or if you have your own horror stories, I would be interested to hear from you as well. Bon voyage, my noble computer users. Godspeed to your hard drives. Godspeed to your CPUs. And most of all, Godspeed to your OSes. We all know they need help from someone.