Bruce R. McFarling

There are at least three types of people who assign web pages into the category of web pages that ‘suck’.

First, those who feel that a web page doesn’t hit the mark unless it has all the latest and greatest features that their browser and/or collection of plug-ins support. Unless a web page sings, dances, and generally dazzles at the same time that it is a stress test for computational resources, it ‘sucks’.

I am not in that group.

Second, those who disagree. They seem to feel that the Web is a forum for the promulgation of viewpoints that they agree with, and web pages caught in the act of cravenly serving the promulgation of the wrong viewpoints, ‘suck’.

I am not in that group.

Third, are the boring cross-platform fanatics, who want html to be what it was intended to be: a portable information presentation and cross-reference system.

I am one of these boring dweebs, and yes, you’re pages do ‘suck’ in terms of cross platform support.

They fall down in the most basic and what should be (for an author) the most natural of tasks: composing ALT tags to provide a universal means of expressing information that is originally expressed in a platform-dependent manner.

Looking at your site, you have ALT text. But you have alt text like (and I quote)

alt="emailblimp.gif (23130 bytes)"

with

then you will be getting a start on the only rule about ALT tags that everyone responsible for a web page has to know:

ALT tags for images should serve as a substitute for the image, and the image itself should only be described in the ALT tag if presenting that image is the purpose of the page.

About: libraries, webdesign, Linux. 11f

From : Svenson, [ qjsw@oce.nl ]

Jerry,

I did pay by the end of December (98). I know you did get the payment because I got the January column URL mailed. But only on the 20^th of January the money was actually deducted from my account. Guess whose money you got.

A note on your discussion about libraries. The thing that the Libertarians who want to abandon libraries are right about, more or less, is the fact that most public libraries carry books (or music, or …) that are indeed cheap enough to buy by almost every body (although ‘almost every body’ is not enough in my opinion). What they definitively miss is that books, even the non specialised entertaining types, become sold out eventually and then libraries become the only accessible places to get them. Apart from that, libraries are the best places to learn to appreciate other styles of literature and other cultures.

If you transfer libraries to the private, un-subsidised sector only books that are popular will remain because others will not bring enough revenue. The end result is a library that indeed only has books that are cheap enough to buy. And a few expensive ‘showcase’ pieces. You lose the variety which is needed to keep a society going.

I notice you still get comments about you web design. I think it is just right so. It is just a bit too chaotic for a casual visitor but people that want to visit you will learn the place soon enough. This means you have a kind of filter build in. Only people that are interested in what you say and do will come back as it takes some effort an dedication to get at the hidden treasures. It is fun in an Adventure type of way.

Another advantage of your reviews section (but also in your mail and view ) is that there is a very wide range of interests going from science-fiction to opera, covering history, education and computers. Most other sites are either highly focused and cover one area in depth or are unfocused and wide but easy going and casual, newspaper like. On the net we need both types of site of course but yours is more thought provoking. I visit it in the morning (in Europe, so your side of the world is just closing down) as a kind of cerebral fitness exercise. So <strong> DO NOT </strong> reorganise or divide it in topic pages. In a good stew everything is boiled together to blend the tastes. You may split up more (ex one mail page a day) to keep the site fast.

I installed Linux (Calderra 1.2) recently. It went faster than Win98. I installed StarOffice 5.0 on it, equaly smooth. The problem is that I am lacking X windows server support for my Mattrox G200 AGP card so I am down to 640x480 at 16 colour. This is not really usable if you got used to 1024x768 at 24bit colour on OS/2 and Win9x.

So once you get past the easy bits and want to set up a usable system the fun begins.

For inexperienced users Linux can be an enormous time-sink but it sure is a way to get cobwebs out off the brain!

BTW I originally installed it on my old 386 box, and old (1990) full tower thing (AMD386.40MHz, 8MB RAM, 200MB disk, Tseng ET4000, 14"display; a real speed king in its day) which I keep around because it is the last box here with both 5.25 and 3.5 inch diskettes. This went smooth enough, dual booting with DR-DOS 6.0, except I couldn’t print or use the mouse, probably because the old I/O controller (a separate board in those days) isn’t supported. Now I bought a new system (AMD K2-300 with 64MB RAM) and added a second harddisk (10GB totally) for it, when I told them in the shop that I intended to run Linux on it they proposed not to install Win98 which they normally install on all their systems. When I get time to do a reasonable review I’ll send a kind of report of my struggles ().

Ps Congratulations and good luck with your new car. Just don’t test its rolling capacity too soon ;~)

Svenson

Mine however is permanently ‘out of memory’. .

>I don’t know what an ALT Tag is; apparently Front Page assigns the ones

you don’t like much. I am >not at all sure what I ought to do about it, but

more to the point, WHY? I can see how replacing >some of the tags with

their links makes sense, sort of, as it certainly conveys more information

>than a description of the file length; but this is the first I have hard

from ANYONE on this >subject, and I don’t know why I should care what the ALT Tags say. Enlightenment welcome.

What an ALT TAG ‘is’ is the little pop-up box that appears if you let your cursor sit on a graphic (jpeg) for a second or so.

How/why they’re used is to provide for those people who either a) browse with graphics turned off in the browser (this can speed things up considerably on graphics heavy pages) or b) use non graphical browsers (yes they exist and people use them) like LYNX (the original WEB Browser that ran on unix systems before the WEB became so highly graphics oriented) or c) are visually impaired and browse the web with text-to-speech enabled browsers.

Frontpage, by default, puts in the filename and size of a jpeg for the text in the ALT TAG, but entering your own text is quite easy. In Front Page, right-click on the graphic to bring up a context menu. Select ‘image properties’ from the context menu. Then in the pop-up config window you can enter text in the ‘alternate representation’ section to put whatever you want in the ALT TAG.

John

Jerry,

You and Roberta might try to get in touch with the University of Chicago and work with them on their suggestions for preparing for The Core, for my money, probably the best undergraduate preparation for the thinking adult citizen an American college has ever put forth. Sadly, however, it is being diluted, in order for "Chicago" to be known as a more "fun" school.

Also, to the issue of your difficulties in formatting and posting to your Web Page (since you are now running a REAL OS (LINUX)), you might try Tamu’s rather creatively brilliant WetPerls (URL: http://people.physics.tamu.edu/jsmith/wet-perls/ <http://people.physics.tamu.edu/jsmith/wet-perls/> ) for a look at how this might be solved. You also might enjoy his link to the "Reagan Leadership Society" (grin).

Best regards,

Mike

Dr. Pournelle, I need clues. I wouldn’t dream of asking you for the gory details, but, can you clue me in who to ask? Question is, what is the trick to hooking up an HP Scanjet IICX to a single, non-networked Windows 95a machine?

Did you think Microsoft had run out of ways to make itself look bad? Think again.

Sample quote from the _Washington Post_:

"Around the courtroom where the government’s antitrust suit against the software giant is being heard, Microsoft’s lawyers looked crestfallen. A spokesman, Tod Nielsen, rushed into the hallway with a cellular telephone. Boies appeared jubilant as he pressed onward: ‘How in the world could your people have run this program? ... You do understand you came in here and swore this was accurate?’"

Hello Jerry,

Bearing in mind the recent discussion about computer dealers who only offer Windows, the following URL provides a list of dealers who offer OS/2, and sometimes other operating systems, preinstalled on their systems.

http://budgetweb.com/os2/preload.html

Bob Griswold in King George

Jerry,

I've just discovered your "View from Chaos Manor" and when reading your Week 2 comments you were wondering if there was a "return to where I just was" command in Word.

Maybe this problem is already solved by somebody else but I didn't found anything on that subject in your Week 3 to 6 comments.

If what you want is analogous to the "Return to a previous editing location in a document" command in Word, than you can use the "SHIFT + F5" command. This two-key command keeps track of the last three locations in your Word document where you typed or edited text.

The "SHIFT + F5" command works in Word for Windows 2.0 and in Word 95. I hope it also works in Word 97 to which we are upgrading next month. I use this command on a daily basis when editing Word documents and found it very useful.

I hope this can still be of use to you.

Regards,

Filip Verberckmoes

Jerry,

ODBC: _O_pen _D_ata_B_ase _C_onnectivity. Used by many apps to access databases not in their native format.

Something installed ODBC drivers. If you installed MS Office Pro then I'm certain MS Access did it. You can go to the install program for MSO Pro and reinstall, refresh the install, or select just the ODBC for reinstall. That should/may fix your problem. You can also download the most recent ODBC drivers (and a whole lot more) from the MS site at:

http://www.microsoft.com/data/download.htm

You want the most recent MDAC file.

I would start by trying the Office CD approach first though. You will get stuff you may not want in the MDAC download.

I share your pain with your Windows configuration problems. It does seem that a minimal change can lead to lost hours, if not days. MS has done a good job of trivializing many operations, making incredible computing power available to the masses.

Linux, OTOH, has a looonnnggg way to go in this regard. When you stated that installing Linux was not a trivial affair you were right on the mark. I think I'm a pretty sharp guy when it comes to PC's (and others would verify this) but Linux has really tested my mettle. It's a good thing though, because I feel it is forcing me to get back in touch with the machine in a way that Windows no longer (or at least rarely) requires. And, being a software developer, I take some (sick) pleasure in actually compiling pieces of my OS.

Anyway. . . when you really think about it, all of those electrons buzzing about, set into action to do our bidding. . . well, I'm still amazed that any of it works. Especially as well as it usally does. . .

Regards,

Brian W. Harriger

bharriger@austin.rr.com

I have been a great fan of yours for some time, and I used to read your columns in Byte magazine quite regularly. I run a small computer consulting company in Cambridge, MA that provides consulting services to large companies all over the U.S. I recently had some experiences with Diskeeper v4.0 and Windows NT that I thought you might want to know about.

We recently took an older Pentium 133 computer that we had lying around and decided to make a FAX server out of it using Windows NT Wrkstn v4.0 and Symantec WinFAX Pro v9.0. We got the system up and running properly, and it trundled along quite happily for several months.

We also resell computer hardware and software as a service to our clients, and one of the things we resell is Symantec software. We received a v1.0 copy of Norton Utilities for Windows NT, which we installed on our FAX server because we wanted to get the disk defragmenting features of Norton Utilities. Turns out this was a BIG mistake, as the v1.0 Disk Doctor program under Windows NT will scramble the FAT on a NT drive and leave it in an unrecoverable state.

Ok, so having learned our lession - we rebuilt the machine (took about 3 days over a couple weeks . . . hey, we work for a living...). Having read the positive things that you had written about Diskeeper, and also having had experience with Diskeeper under VMS on various VAXes &; Alphas—we thought, okay we’ll format our drives with NTFS and use Diskeeper on them.

Turns out this was also a big mistake. We had 3 disk drives on this system; C: was a 1 Gb FAT formatted volume, D: was a 1 Gb NTFS volume on which NT resided, and E: was a 2 Gb NTFS formatted volume used to hold user data. We ran the disk directory consolodation feature on drive C: and everything worked great. We later ran the disk directory consolodation feature against drive D: and that’s when DISTASTER struck again. (I must admit I was for a moment so annoyed, I nearly took the computer and threw it out the window when I found out what the problem was...read on.)

The crux of the problem is that when Diskeeper goes to do the directory consolodation on an NTFS volume, it attempts to run CHKDSK against the volume before and after the consolodation. Unfortunately, the AutoCHK functionality in Windows NT (both server &; workstation) seems to have a serious bug in it that was introduced in Service Pack 4.

When Diskeeper invokes AutoCHK, either AutoCHK or Diskeeper does something to the registry on the machine so that it winds up being permanently and unrecoverably corrupted. What the end-user sees is that the system comes up, and then when the blue screen appears before the system boots into the GUI and AutoCHK runs - the system simply hangs after AutoCHK finishes running. Pressing the spacebar to stop AutoCHK from running, stops AutoCHK but then hangs the system.

We tried using the Recovery Disk built during the installation (and subsequently updated several times) to restore the system to a bootable state to no avail. (Turns out there are problems with using a recovery disk under certain conditions to repair a system, since not all of the files get "repaired" during a repair—check out Microsoft’s Knowledgebase articles Q168015 &; Q158423 for more information.)

Basically, we are now getting ready to rebuild this system for the THIRD TIME in three months. I think I’ve had it with all of the utility vendors, since between them and Microsoft—who needs enemies? (I’ve also recently had some serious war stories with Norton Anti-Virus v5.0 that Symantec has admitted are a problem, and haven’t been fixed yet... The Ultimate solution was to go back to using v4.0 of Norton Anti-Virus until they get the bugs in v5.0 worked out.)

Executive software has also realized the error of their ways, since they have just released an update to v4.0 that disables the AutoCHK functionality so that disk directory consolodation can do its thing.

Mr. Capps in Executive Software’s technical support department can confirm what I have stated above, and more information about this "undocumented feature" (sic) can be found in the Microsoft Knowledgebase articles Q173322 &; Q151376.

I have been working with computers of all sizes and shapes since 1978, and it sure seems to me that the quality of the software coming out seems to be going down in proportion to the price of hardware going down. Although 10 years from now computers will be so cheap that anyone will be able to afford them, the software will work so badly that no one will want to. (mumble)

In any case, I thought that you would want to know about this bug, and also let your readers know about it so that they might be able to avoid the horrible fate of having to rebuild their systems again.

Continuing the discussion on Education:

Anyone contemplating the education of the young may find Dorothy Sayer’s essay "The Lost Tools of Learning" valuable. Sayers makes a convincing argument for a return to the medieval "trivium," grammar, dialectic, and rhetoric, in that sequence, as a foundation for the abilities to think, communicate, and learn. The essay sketches a curriculum in which young children memorize grammar and other things, pre-adolescents practice the skills of logical debate, and adolescents follow a more and more individualized path in which in-depth knowledge, and the ability to synthesize and present results, is emphasized. Sayers sees this "trivium" phase of education ending at age 16, at which point the young person proceeds to either trade or university. In either case they have been equipped with much more than just a basic store of knowledge. The main point of their education has been to provide foundational abilities in language and logic that will enable both further learning, and rational participation in society.

Subject: possible explanation of why ODBC is in Windows

Dr. Pournelle,

Realize I am not sure, but I think that the registry uses the ODBC component, which is why a basic Windows install includes it, It is one of those nifty features that MS includes, like Paint, or a browser, with their OS. I think that since the cards had been moved, it was a "new" system, and Windows began the Plug N Play shuffle. The ODBC version was not the one it was looking for, so it complained. Believe me, Windows 9X and Plug n Play and upgrading components on same can be a real pain in the nether regions. We here at my place of business try not to upgrade systems if possible, instead recommending a new purchase. Many of our customers are not tinkerers, and couldn’t handle the stresses of the continuous restarting of an upgraded windows box. They would panic on the first PNP shuffle. Better to ease their mental pain. Now you know why resellers don’t work on their friends’ machines, much like lawyers and doctors.

George

paulbeaver@csi.com

Subject: ODBC woes and the card shuffle

Jerry,

Just a brief comment on why your computer was confused when you did the "card shuffle" whe moved to the new case.

The problem nay have been caused by your system. When the PCI bus was introduced one of the big points was that PCI cards would be "self configuring" when the system starts up all the pci cards in the system communicate with the system bios, each card would take available resources, if there was any conflict it would be resolved by the bus arbitration. Like most new standards it was not implemented well. In some combinations it just didn’t work.

In some older computers you could overcome this by going into the BIOS and assinging an interrupt to a PCI slot. Normally cards don’t care what slot they are in, but if plugged into a PCI slot that has been given a set interrupt, then windows will reinstall the drivers for any card that has changed its interrupt.

well, I can certainly sympathize with your woes regarding moving the hardware to a new case.

I recently installed a cpu fan on my video card, an agp tnt. To do so, obviously, you have to remove the card. The installation went quickly and painlessly, and I reinstalled the card to the agp slot, taking great care to seat the card fully. I then rebooted the system, which showed the bios of the card correctly, and in the right place during boot, and everything seemed fine.... until I hit the desktop, where I was now in 640x480x16 colors. As I run at 1024x768x32bit color, that was not good. I attempted to change resolutions, but the system wouldn’t allow any greater than 640x480.

I shut down, and despite the fact that it made no sense, made sure the card was properly seated. Yes it was.

Reboot. Check drivers. Win98 says Diamond Viper550, version xxx, all of which is correct. Still nothing above 640x480. So, I reinstall the drivers and viola, all is well. Given the fact that the drivers were present, and the system never showed any "installing new hardware/software" messages, I still don’t know what happened. Argh.

Win98 is better with hardware than Win95 was, and both are better than the dos days, but it aggravates when the system just don’t work. Makes me leary of messing with the hardware, which may be the intent. Paranoia strikes deep in the heartland.

Subject: CDROM on IDE

Jerry,

One possible problem you may be encountering is having your CDROM as the MASTER of an IDE channel. CDROM’s should always be slaves (even if there are NO masters on a channel). From what I read on the web page, I had exactly this configuration (CDROM as MASTER), and had months of good service. Then, the system started behaving badly. Short lockups at random times. Then the sytem crashes would start. Eventually, it would "loose track" of the CDROM under windows. After lots of headaches, and a lot of "education" about windows and the registry, I asked someone at the local computer games store about this, and when he dragged out that I had the CDROM as MASTER, his eye’s went wide.

Good Luck, and Enjoy!

Thanks for telling me about CDROM slaves.

The trouble is, he told you wrong. Most CD-ROM drives come jumpered Slave on the assumption that they may be connected to a channel that already has a hard drive installed as master. But because many CD-ROM drives will be installed as single devices on an ATA channel and many of those will not have their default jumpering changed, most ATA interfaces are designed to function properly when a CD-ROM drive jumpered Slave is connected as the sole device on a channel. They’re doing so outside the ATAPI specs, however, and this configuration is not guaranteed to work.

When I build a system that has one hard disk and one CD-ROM drive, I always put the hard disk on the Primary ATA interface, the CD-ROM on the Secondary, and jumper both as Master (per the ATA/ATAPI spec). I’ve never had any problem doing it this way. I have, on the other hand, seen numerous systems that don’t work properly with a single Slave CD-ROM drive on a channel.

My understanding, which appears to match yours, is that you must always have a master if there is a slave.

In fact, there was a period of a few fixpacks where IBM enforced this with OS/2, failing to boot properly if it found a CD-ROM drive as a slave without a master.

This rumor probably started for a couple reasons. One, many IDE CD-ROMs ship already jumpered as a slave to help ensure success for an inexperienced user who adds it to an existing hard drive. Also, as you’ve likely noticed, sometimes things don’t work when everything is done correctly. So, all it would have taken is for someone to connect it as a slave and have it work to make a rule out of it.

For myself, I’ve had cases where an IDE CD-RW drive didn’t work as a slave to a hard drive, but making the CDRW drive the master and the hard drive a slave worked. (Should have spent extra for the SCSI version, but I didn’t think I would use it enough to warrent the extra $150).

The business about strapping IDE CDROM units as slaves might be nothing more than pidgin superstition but in my experience it holds true. Same goes for IDE ZIP drives. I’ve even had both a CDROM and an IDE Zip drive on the same IDE channel and both had to be strapped slave before either would work reliably.

To Windows Magazine:

"I’ve known Byte Magazine for years, and believe me, you’re no Byte Magazine"

I’ve been a fan of yours since October, 1983, which featured a cover story on my first computer: HP 150 with Touchscreen (for only $ 8,000 including business software, "leased at $200/month for ever"). I never discarded any issue of Byte as I considered it in the same class as a professional journal. With the arrival of each issue, I immediately turned to Chaos Manor, ignoring all else until I had gleaned the "inside story" or suffered through the latest insult to documentation intelligence and digested the valuable lesson in patient, logical troubleshooting.

My biggest surprise at the Fall ‘89 Comdex:

That so many "Professional" marketers of hardware and software were unaware of your column!

Most memorable column:

The day Thor struck Chaos Manor and "field tested" your collection of UPS and surge protection equipment. Since discovering them at a remote hotel at Comdex (Where the small, but interesting companies are found) I use:

Control Technology, Inc.

7608 North Hudson

Oklahoma City, OK 73116

405-840-3163, FAX 405-848-0489.

They also make mil-spec UPS’s for the armed services. I used their mil-spec units to protect the PC in a transportable micro-refinery destined for use in the third world. Their commercial units have nearly the same components, are less expensive than sine-wave, and respond in less than 0.3 milliseconds (Zero Switch option). My 750 watt model survived a near miss from lightning that hit my neighbors aluminum siding less than 100 feet away and knocked out all unprotected electronics. The same strike killed my $25 modem surge protection, but not my $400 SatisFaxtion board. A few years later, I learned that it successfully ran my cleaning ladies vacuum cleaner for about 10 minutes before I unplugged the beast. No damage. A similar load killed a 450W UPS (American Power Conversion) at work. (Replaced under warranty, but not without downtime. We made it up to APC with more purchases for "less critical" workstations. Control Technology replaces batteries (twice in and does upgrades and repairs (I overloaded the guy and held down the circuit breaker switch—shame on me!). My UPS has kept me going 24 hrs/day for since June, 1991. Thanks again for the "tip" about how to find useful technology at Comdex.

Most useful tip:

Dielectric lubricant/grease for circuit boards and chips (memory, CPU, bios). I dabbed a bit onto all of my household telephone connections to reduce crosstalk and modem static. Canadian firm. Starts with P? They sent me a 30ml free sample, and a full literature set. I can’t remember their name. Blush! I’ll check an old Byte (see why I keep them?). Maybe they have a web site.

Last word:

I first noticed your computer column because I’m also a Sci-Fi fan, a Navy junior, and a USAF Vietnam Vet who appreciates your fair and frank insights on military philosophy and the space program. Sorry, I’ve not written before to thank you for sharing your thoughts and working so hard to bring a little common sense into the world.

Thank you very much!

Bob Huddleston

Subject: Liunx does not boot

I had the same problem. RedHat insisted that it was a 1024 cylinder issue (Linux partition crosses 1024 boundary which some older systems’ BIOSes can’t handle). This was not the case for me.

Jerry,

I remember having a similar problem. It turned out that I had forgotten to set the "bootable" flag in in DiskDruid for that partition.

--

Linux boot problem

Jim Griebel [jgri@earthlink.net]

Jerry

Jerry,

You mention you re-partitioned and re-installed RH5.2. In the installation there is a choice for installing lilo (boot manager) onto either the mbr or the start of the first HD partition. Sounds like you chose the start of the first HD partition instead of the mbr since you can boot from a floppy. Get your boot floppy and boot the machine.

then: emacs /etc/lilo.conf

The file should read something like this

boot=/dev/hda #boot from master boot record

map=/boot/map # name and location of kernel files

install=/boot/boot.b # the boot sector file

prompt # gives you the prompt to choose labels

timeoutP # how long to wait before continuing if no input

image=/boot/vmlinuz #the name of the kernel file to boot

label=linux # the friendly name

root=/dev/hda1 # the location of the root partition

read-only

I suspect that your lilo.conf will show

boot=/dev/hda1 or /dev/hda2

I believe that you can fix the problem by simply altering this line

back to

boot=/dev/hda

saving the file

and then typing

/sbin/lilo at the command prompt

But definitely check with the many experts among your readers

On the ODBC drivers, this sounded familiar so I checked back and found (using the viewdex ;-) in view 7 you mention some problems relating to Outlook and these drivers, perhaps this has something to do with the present problem?

Well, I understand how you feel, but if you keep hitting yourself in the head you can’t reasonably complain when you end up with a headache. Just remember, we’re doing this so they don’t have to...

I thought you’d have problems using MS-DOS fdisk to set the active partition for just the reason you describe. Just a suggestion, but why not just boot MS-DOS, run fdisk, pick a partition at random, and set it active? Restart the computer from the hard disk. If it doesn’t boot to Linux, just try making a different partition active until you find the one that boots. It’s not elegant, but it should take a minute or less to test each partition, and that’s less than you’re spending on it now.

Bob

Mr. Pournelle,

I find the best tool for partition related stuff is the Linux fdisk.

From what I’ve read in View 34 you have tried to used it, as you say, "RED HAT fdisk doesn’t seem to have a ‘list existing partitions’". Well, it does. At least if we’re talking about the same program. The command is "p" (print current partition table).

When you run RED HAT fdisk, it outputs something like this:

Command (m for help):

The command you need to use is "a".

If you need more help, please send me (or publish in MAIL or VIEW) the output, errors etc. you get when using fdisk &; the rest.

You’re right, of course, FDISK doesn’t tell you the names. I am PRETTY SURE the one you want to set active is Partition 1, which is at the start of the disk and will be where whatever’s telling you the disk isn’t bootable is looking for the partition table/boot loader it can’t find. This shows as "Non-DOS" in MSDOS Fdisk, with a size of 18 MB, which happens to be the /boot partition on my Linux disk. It may also be the only partition FDISK will allow you to set active, but I’m less sure about that.

Subject: Still More Linux Boot 11f

From: Jim Griebel (jgri@earthlink.net)

You’re welcome, but let me follow up so you don’t think you have a Linux problem because I don’t think you do: It’s that expletive deleted Award BIOS. Or that’s how I see it.

When I set up Red Hat 5.2 I did it on the Windows machine here. I bought a new motherboard at Fry’s to use in the Linux machine. These both come from the same maker – EFA – and both have Award BIOSes, but different versions. On the old motherboard Linux booted just fine. When I moved that HD to the new motherboard, I got that error message, which in full is:

"Not found any (active partition) in HDD. SYSTEM BOOT FAILURE. INSERT SYSTEM DISK AND PRESS ENTER."

I suppose the Taiwanese English should have been a major clue, but still I spent most of a day convincing myself that it was the BIOS that was refusing to boot Linux. I even did a new minimal Linux install on another old HD. Wouldn’t boot on the new motherboard. Booted fine on the old one. Swapped motherboards and called it a day.

From what I understand of the way the boot process works, here’s what’s wrong: The BIOS on the old motherboard just pulled in the boot sector and ran the little program that’s in it. That little program doesn’t care if there’s an active partition; it gets told in explicit detail where to find the program it’s supposed to load as the next step in the startup process, goes ahead and loads it and starts it, and we boot. The BIOS on the new motherboard looks at the partition table on the HD, discovers there’s no active partition, throws up its hands and refuses even to load the boot sector. If it would just load the furshtunkene thing, the way that BIOSes on Wintel machines have been doing for the last 15 years, we wouldn’t have this problem.

That’s why it worked with a disk originally formatted for DOS: You set an active partition in FDISK when you originally set up the thing. (This is a reflex action when running FDISK, at least for me, if I want a bootable disk.) The BIOS saw this and happily loaded the boot sector.

I suppose that there’s room for argument about whose problem this is, but I’m inclined to blame whoever thought up this "enhancement" to the BIOS. On machines which don’t bother looking at the partition table – which is probably about 99% of them – this problem just doesn’t happen. As already noted the problem doesn’t even occur on all Award BIOSes – one of my machines has the problem, the other doesn’t. Keeping up with this kind of thing is at least very difficult, especially for a small company like Red Hat. And they’re up against the old "fix something, break something else" problem – Who’s to say if they changed Disk Druid to automatically set a partition active to keep Award happy, they wouldn’t give themselves a problem with some other BIOS? I agree that it’d be nice if the fix were easier to come by; from the e-mail they sent me, Red Hat is aware of the problem and the fix, but go try to find it on their Web site.

You may find places on your new vehicle to use Stabilant 22A as well. It is widely used as part of the hop-up procedure on hot rods and race cars now that everything is microprocessor controlled. Underhood and off road is an interesting environment.

Incidentally, do you happen to know whether some of the aircraft electronics were potted with asphalt which would weep and mildew when they parked the aircraft on the ground in South East Asia? Heard that but not from any reliable sources.

Dear Jerry:

You have encountered one of the banes of a support person’s existance. But it is not specific to RedHat or to linux. It afflicts all Operating Systems and much of the other software we use.

The key premise is that all PCs are supersets of a well-defined and rigorously specified Architecture and Reference Platform. At one time, that was true. But some current designers treat as interchangeable "meets the specification" and "runs with today’s Windows release". That is why NT, and RedHat, and SCO, and everyone else is forced to maintain "Approved Hardware Lists".

My notebook has a Microsoft-logo "Designed for Windows 95" label on it. I have not been tempted to switch to Win98 - even though that migration "usually" works. I once wasted too many hours trying to install NT 4.0 on a Compaq Prosignia VS (which was once on the NT Approved list - but was deleted for exactly the same bug that I had been fighting. It took additional hours with the "Knowledge Base" to find the details). I am sure your Award BIOS is no longer on the Redhat list, either.

We all have horror stories. But the root cause is a lack of quality. As long as vendors ship "probably good enough" products, support people will be busy. And not merely linux support people.