This article is more than 1 year old

Bring us the head of Michael Dell

Sick of them 'Jim Crow' Windows-only boxes

Letters Those who've followed what Linux Today Managing Editor Michael Hall knowingly calls our Red Hat "installation tragicomedy" will recall that it came down in the end to a Dell computer with two CD drives set by default to 'cable select.'

I took a swipe at Red Hat for not knowing this, and for not warning me in its documentation or taking steps to correct it, but the majority of our readers would lay the blame squarely on Mikey's doorstep.

It seems that Hewlett Packard, Compaq and Dell all default to cable select, as this thread at Total Tech Forums indicates. That accounts for how many million machines? If the practice is this popular, something, clearly, has got to be done by someone. But who? I nominate Red Hat, who should know about this and at a minimum warn us, but our readers largely disagree.

Herewith a sampling of your letters:



I'm rather incredulous that Dell still use CS for their drives; I had no end of grief with my old Dimension when adding a DVD/CD-RW drive to it last year (and that was under Windows, no less!). So I'd be inclined to say it's Dell's fault for using such a bletcherous and completely useless system such as the CS jumper, but Red Hat certainly has a problem if it causes so much trouble. It might be a Linux problem, but I installed Slackware 97 on my Dimension without hassles... Moot point for me now, anyway, since I moved to an Optiplex with SCSI throughout.

Personally, for a home user, I'd recommend Debian (and have, in fact, and the home users in question are very satisfied); but then I don't like Red Hat at all (my machines all run Slackware). IMO Red Hat is the Windows of the Linux world; full of junk with loads of people using without realising that it's just causing them grief. I've lost count of the number of people I've spoken to who found Red Hat difficult to install and assumed Linux in general was too difficult...

Regards,
--Ben A L Jemmett



I'd expect it is Dell's fault. As far as I know, using cable select instead of master/slave is not a 'vanilla' config. My understanding is that cable select capability is only there for older machines which require it when a master/slave config. will not work. Thus I'd suggest that for Harry Homeowner Red Hat ( minus Dell) is fine.

--David

ps. just for fun, I did a search for cable select on Maxtor Support Knowledge Base - result - not recommended!!

Cable Select (CSEL) is an optional feature per the ANSI ATA specification. It is an alternative method of identifying the difference between device 0 and device 1 on an IDE interface cable. Hard drives configured in a multiple drive system are identified by CSEL's value:

If CSEL is grounded, then the drive address is 0.
If CSEL is open, then the drive address is 1.

Additionally, CSEL requires a specialized (and more expensive) 40-pin IDE interface cable, unlike the standard EIDE interface cable that is far more commonly used. Maxtor does not recommend using the CABLE SELECT feature unless your specific system documentation requires its use.



While I have no great love for Red Hat, I think Dell is the one to blame for this...in my 6 years as a tech, I've never seen cable-select to work reliably with any OS, and to be frank, any monkey that can't be troubled to learn the intricacies of moving one fucking jumper on a drive for master/slave-select should be shaved hairless, horse-whipped, and set back down in front of the TV for their daily dose of radiation therapy.

Just because the manufacturers would like the squirming masses to believe that computers are just glorified toaster ovens doesn't make them so. Anyone who's actually installed Linux on a computer and done real work with it knows what their machine is capable of, instead of devoting a godzillion fp/s to making things "pretty", "friendly", and "easy for you fucking retarded sheep to use".

Come to think of it, Dell isn't to blame at all...they are, after all, in business to make money, a retard's money is as good as a genius's, and there's a hell of a lot more retards than geniuses. So...the monkeys that drive the market are only getting what they deserve, and they obviously don't deserve Linux. Unfortunate for Linux from a money-making standpoint...but who really wants a Linux that's been dumbed-down? Oh, yeah...

--not signed



Like all I.T. "fault" issues, the answer depends on what level you look at the problem. From a technical perspective, one could argue that Red Hat didn't test the ability of Linux 7.2 to install with IDE devices in CS instead of M/S mode. As cable select is a valid and standardised (if not regularly implemented) mode of device selection, then the "blame" appears to lie with Red Hat, and not Dell.

However, from a higher level, consider the plethora of combinations of hardware device settings that an OS must be tested against to ensure it will install correctly. Your summary of e-mailed suggestions highlights the list - everything from BIOS settings, DMA HD drivers, IRQ conflicts, PCI slot issues, etc.

Even though we now have plug & play, PCI auto device contention etc, and other new technology that supposedly relieves us from the "bad old days" of interrupt conflicts and manual device configuration, we now have firmware and device driver layers to content with above the hardware. In addition, much of today's technology (such as IDE) is a carry over legacy standards from more than 10 years ago. As such we still to contend with cable select vs master/slave jumpers.

There are so many vertical layers of a complete system to get from hardware to operating system that it is impossible to test fully at a device level. For the moment the only way to guarantee compatibility is to test each system against each operating system version to ensure compatibility. Microsoft has the money, relationships with suppliers, and infrastructure to be able to do this, and most Linux vendors simple cannot compete at this level.

You argue that you deliberately chose a "bread and butter" standard Dell box for this reason, and yet if you just add up the number of standard PC configurations offered by the top manufactures during say the past 2 years, you still have thousands of "bread and butter" variants to test.

You are right that the "Moms and Dads" out there in consumer land don't know and don't care about the technical issues - they want to be able to insert a CD in their computer and get it working! On that basis I also wouldn't yet recommend Linux as a Windows alternative for non I.T. enthusiast users.

However, I.T. enthusiasts see your problems and this entire configuration situation as a grounds for a debate as to which is an architecturally better O.S. The summary of my points is that is not an issue of which O.S. is fundamentally "better", but an issue of whom has the resources to conduct exhaustive system testing and verification before shipping a new product. Microsoft has the huge resources to win out here, as your case demonstrates.

One final point. As I'm sure you are aware there is currently an industry push to remove legacy devices from new systems (such as serial, parallel ports, floppy drives) and only have systems with USB 2.0, firewire, bluetooth etc., interfaces supporting truly universal and compatible drivers and firmware. Whilst the cynics see this as another push to make the old obsolete and sell more of the new, it does offer hope of a more even playing field for Linux vendors. A universally compatible hardware interface standard would mean that with much sociability testing, Linux could be as easily installed on all new computers as Windows is today.

It may also leave many PC support engineers looking for a new job!

Regards,
--Tim Epstein



glad to hear everything finally worked out. I had forgotten about the cable-select settings that dell uses on all their desktop hardware, i only wish i had thought of that a few days ago and recommended that solution. (perhaps i has just 'not paying attention' and thought you were trying to do this on a laptop).

as for the 'who's fault is this... dell or redhat?' question... i can only come up with one answer: dell. i've been a field service tech for about 4 years now, and dell is the only pc manufacturer that i've run across that's so adamant about shipping their drives in cs-mode. if you ask any professional, they'll tell you that they should be set as either master or slave... and that the cs jumper is for podunkers. (they'll probably use more colorful language... but you get the gist).

anyway, it's disappointing to have all this rigmarole reduced down to something as trivial as drive jumpers. if you've got some time to kill, you should try repeating the experience with xp and rh7.2 on a non-dell desktop.

cheers,
--corigan



Thanks for the good work with the Red Hat 7.2 distro. You've reminded me why I won't buy Dell, IBM, HP, Compaq (only Alphas), or Gateway. They all use funny hardware. They tell you you're getting a Voodoo 3 3000, but you're really getting some cut-down OEM version of the card. You thought you were getting an FIC PA-2013 motherboard, but again it's a cut down version with a bizarre BIOS made by Phoenix instead of AWARD (like the normal 2013 has). If you need to swap out the motherboard, you better have some metal working tools because even the form factor is weird. Anyway, you've taught a lot of people, including me, one more helpful tip. Hopefully Red Hat will find a way to deal with this issue.

Harry Homeowner will be better off buying a preconfigured Linux machine. He won't have to install Linux (at least the first time), it will perform better, and it will have replaceable commodity hardware. If Harry does decide to upgrade, his components are guaranteed to work with Linux. Only Microsoft has enough muscle to get manufacturers to put those stupid "Designed for Windows Whatever" stickers on tested products (and they seem to be losing out with the Win2K and XP driver issues).

One explanatory bit of trivia: the Linux kernel eschews the BIOS. Only the bootloader uses the BIOS. While it may not be obvious to Harry Homeowner, most of the time this is a good thing. BIOS implementations are horribly buggy and limited. In fact, there are reports of BIOS
implementers adding or withholding features at Microsoft's request, which has turned BIOS implementations into something of a Windows extension. Once you've booted Linux, you might as well forget about your BIOS. I'm guessing the cable select problem is a Linux kernel problem with your IDE controller, and that Red Hat's installer exposed this problem. Of course, Red Hat (like all vendors) often distributes a highly modified kernel (one that's been heavily tested and patched), so they could have screwed things up from that angle.

--Paul Komarek



It's *obviously* Dell's fault for not setting Master/Slave settings on your drives by default. As a matter of fact I am impressed and stunned that everything else worked till now. Cable select? You must be joking!

I am not Harry Homeowner so I would obviously never buy any crap box from <insert big brand name here>, but I guess it's little thingies like this one that separate good brand names from bad ones. Dell just got a yellow card in my book.

btw I am a Windows user myself, tried various linux distros from time to time, including the redhat 7.2 you are trying to install. Red Hat works OK for now, I have some lilo problems for the moment, and the compiler seems buggy but it's OK.

Also, if you are looking for a linux distro for the average Joe, Red Hat is not the way to go imho. Red Hat does better at servers. Mandrake, SuSE are A LOT more user friendly from the ones I have seen.

SuSE can only be had from a shop, Mandrake and other distros can be freely downloaded from http://www.linuxiso.org if you have a lot of bandwidth. Hope this helps....

--Thomas Venieris



For about 6 months I did 3rd party onsite Dell hardware support. I ran into a similar problem on at least 5 machines. Cable select drops and regains drives. The actual problem that was occurring is the ZIP drive would disappear and reappear, when the computer went to write to the CD burner it would also write to the undetected ZIP drive and permanently? trash the disk.

First time of course it had been diagnosed as a failed ZIP drive. Installing the new drive of course seemed to fix the problem. Only to fail again in a week. Switching from cable select solved the problem every time.

--not signed



It's my view that this is probably Dell's fault. I work on computers and used to build them and I have seen many problems with cable select. I don't believe the BIOS or the cable should ever be left To choice and that it should ALWAYS be set manually on the drive. Manually setting jumpers saves on end user problems. I once had the same cable select problem with WIN2000. Also It would keep you from making the mistake of not paying attention and formatting the wrong Drive.

--Richard Shade



Whilst it might be simple to blame Red Hat for failing to cope well with Cable Select on IDE devices I would urge you not to. The problem lies more with Dell's decision to actually use it, and more with the fact the cable select doesn't work 100% of the time.

I've had numerous problems with Cable Select. I've tried it dozens of times and my mileage has varied, but one truth has always remained - setting your drives to master and slave always works more reliably. Whilst I use Linux almost exclusively these days, I've had the same problems with cable select in Windows (9.x) too. The ideas a good one, but it just doesn't seem to cut muster.

Cable select should work fine in the Dell box if used as Dell intended, but remember that Dell designed that box to run Windows and they tested extensively before saying it was okay for Windows. That's why people pay big dollars to Dell (and IBM, Compaq, HP and others). Dell have taken the time to make sure that computer you're buying works with the Operating System your buying it with. Keep in mind that the hardware Dell have chosen works with Cable Select and Windows. The particular combination of hardware has been specifically chosen because it works. Take out the hard-drive and replace it with another and reinstall Windows and chances are it won't work. It's not that Windows is bad (it is ;-] ), it's that Cable Select is very picky.

So why Dell use cable select I don't know. It might make their manufacturing process easier, but it can cause all sorts of problems upgrading. Given that almost every IDE device comes set to Master, if you're adding hardware, you need to know what you're doing. More than likely, you're going to have to change the selector on the back of the new device. Going with Master/Slave instead of Cable Select wouldn't have changed the way the system worked with Windows one bit, but it will make you life a lot easier when adding hardware Dell hasn't approved, or installing a new OS.

regards
--Rodd Clarkson



The fault rests in both their hands. In an age of convenience, Redhat should know that people buy cheap packaged machines. It's cheaper in cost for them, and they are cheaper for the company to produce. Cable select is used by the inexperienced, and those companies that are too lazy to make a machine right. Redhat should have predicted cable select, and had that built into the install program if they had not already, and if they have, it needs improvement. Also Dell is at fault for being so lazy and not setting up the drives properly. They should use Master/Slave, that's what it is there for. In all they are both to blame, Cheap manufacturing, and failed understanding of the market. If linux wants more markets, they need to understand the owners who don't have servers and clones. Dell needs to improve their computer building methods.

--Stephen Mirowski

Related Stories

Win-XP vs Red Hat 7.2
Red Hat Hell continued
Red Hat redemption

More about

TIP US OFF

Send us news


Other stories you might like