Linux's Achilles Heel Apparently Revealed 1469
ahab_2001 writes "In Information Week's latest 'Langa Letter', Fred Langa points to something that he calls Linux's 'Achilles' heel': 'New Linux distros still fail a task that Windows 95 -- yes, 95! -- easily handles, namely working with mainstream sound cards.' After lamenting his difficulties in getting a particular sound card to work with nine Linux distros, he concludes that his experience 'empirically shows that, despite its many good points, Linux still has some huge, gaping holes--holes that Windows plugged almost a decade ago.' (Oddball note: Information Week prefaced the e-mail alert pointing to this article by saying 'Occasionally, we have news or analysis of such importance that it warrants a special alert to you.' Hmm...)"
Win95 sucks at sound (Score:0, Informative)
Sound cards?? (Score:4, Informative)
Is this true? (Score:5, Informative)
Interesting... (Score:3, Informative)
The article's tripe.
Goes against my experience... (Score:3, Informative)
1.) Ensoniq PCI sound card - detected by redhat/debian/slackware/SuSE and setup in the Install. Had to use the driver CD in windows 2000.
2.) Intel OnBoard/Laptop i810 audio (labeled Yahama XC-something under windows) -detected and setup by redhat/debian/slackwaare/SuSE install. Also works with ALSA. Windows: had to download drivers from notebook manufacturer website.
It was written by a Windows Fan... (Score:5, Informative)
Although, this doesn't exactly invalidate his point. Microsoft's got a deep driver library database included in Windows XP... containing many cards that there is no known Linux drivers for.
Re:Huh... (Score:2, Informative)
I don't like the use of the word 'boxen' at the best of times, but at least don't use 'a boxen'
sorry, but ermm ... no. (Score:2, Informative)
want easy audio in linux right now? get a usb sound card. yup, thats right. usb-audio works great, and paired up with jackd, you can quit 'worrying about some magic achilles heal' that may have just popped up out of somewhere
Re:This is a Joke, Right? (Score:4, Informative)
Wrap That Driver! (Score:3, Informative)
Re:WARNING! (Score:5, Informative)
Winders does devices well because that's where the market's been. Linux would smoke Winders boxes in all tests if it had better drivers.
You have the cause and effect backward. Windows has drivers because it's popular. Popularity came first, vendors bending over backward to help Windows work with their products came as a result. The technical framework for third-party drivers is there for Linux. But it's not being used by most vendors.
You *are* aware that Microsoft doesn't write the drivers for most devices that work with Windows, right? It's the hardware manufacturer that makes the devcice that does that work.
Re:It was written by a Windows Fan... (Score:3, Informative)
I'm not saying that he was solely responsible for what happened to Byte, but it was on his watch.
On the other hand, that might imply that his experience does extend beyond those used for Windows Magazine.
Re:Critical! (Score:3, Informative)
Re:Huh... (Score:3, Informative)
I think most people in this crowd will realize the author is trying to appear unbiased, but not doing a very good job.
Re:Notice... (Score:1, Informative)
Re:Win95 sucks at sound (Score:3, Informative)
Re:WARNING! (Score:1, Informative)
Too few people know that fact. What I find even funnier is that the hardware makers have to pay Microsoft for the privilege.
Re:Well I gotta agree... (Score:1, Informative)
spent some time on it. I havent had a sound-card from the last 5 years not work to some degree,
although some required editing the plugin play stuff to force an irq, and driver that mostly matched the chipsets used.
For youre dwl520e:
http://home.columbus.rr.com/andrewbarr/dwl520e1
I don't think he's telling the full story (Score:3, Informative)
NT is notorious for this problem because it won't tell you anything. The driver will install. You'll be prompted to reboot. And you'll get a error in the event log saying the driver couldn't start. That alone could lead to hours of frustration.
But there's also the issue of OEM compatibility or OEM pat on the back ability. Microsoft and Intel go together like white on rice. Those to have worked together for years. Of course an Intel board is going to be supported with the default drivers, let alone an intel soundcard. But for 95 to support a new board with a new sound card with no additional drivers is very hard to believe. 95 probably needs updated chipset drivers for the board alone. And he didn't mention what version of 95 he used, either. If any version of '95 could support a new sound card, which I doubt it would without a driver from the manufacturer, it would have to be 95 OSR 2.x. And that's still stretching it. Out of the box, 95 will support most ISA cards with Microsoft provided drivers. But PCI support is more dependent on support from the manufacturer.
I've seen Linux kernels with a module under sound that says,"AC'97". And if there's one thing to learn about drivers is that especially in Microsoft's case, the manufacturer's drivers should be used first, if they're available.
pay attention to the source... (Score:2, Informative)
The person (Fred Langa) is on Bill's side... Just look at he article track record... I have read a few already.
The unfortunate thing is that he is published in Information Week and is obviously NOT interested in accuracy...
I have wasted time reading the article.
-GO
CM8738 hated. (Score:3, Informative)
I had a hell of a time with the CM8738 drivers (ALSA and non-ALSA) working with the sound card built into my IWill KA-266 Plus motherboard. Interrupt problems, no sound, choppy sound, computer locking.
I modified just about every setting known to man (BIOS and OS). I finally decided that my time was better spend buying a Creative Labs PCI card, sticking that in and using it, than to mess around any more with the horrible sound drivers.
Almost plug and play. It was a shame that (even after seeking so much help and reading so much documentation) that I had to go buy even more common sound hardware to get my sound working right.
But yes, I'm just an unfortunate example of something similar.
Comment removed (Score:3, Informative)
Re:I reckon he tried 9 version of Mandrake (Score:4, Informative)
will save the current mixer settings for the next b oot. You can add the following lines in your modules.conf file:
post-install snd-card-0
pre-remove snd-card-0
Re:WARNING! (Score:5, Informative)
And it isn't for lack of trying either. I've tried several dists, I've patched the 2.4 kernel with ALSA, I've built the 2.6.5 kernel but NOTHING works. ALSA sees the card, but it is muted even if you run the mixer and unmute everything and stick on the max. Yes, I have the speakers plugged into the right connection and yes I'm certain I've double and triple checked everything. It still doesn't work. I'm not alone in this - the internet is filled with people in the same boat as me.
At least 2.6.x comes with ALSA out of the box which is a blessing. But even so, if it takes major kernel surgery (and in my case it still doesn't work) there is something seriously screwed with the model.
On Windows or OS X, at most you stick a disk into the machine or click an exe. That's assuming it doesn't just work automatically. On Linux you could waste a day applying patches and rebuilding to do the same.
Linux really, really needs to sort out the whole driver issue because it throws a wet blanket over widespread adoption. Expecting people to rebuild kernels, or be in possession of a toolchain to build a module is unacceptable.
A single unified ABI for drivers would be a good start. I can understand if Linus doesn't care to support such a thing, but I can't fathom why the dist vendors wouldn't.
Re:Win95 sucks at sound (Score:3, Informative)
The Live does mixing in hardware. It wasn't until the era of the Live, Aureal Vortex, Yamaha PCI, and a few others, that cards were doing hardware mixing. Thus, cards like the SB16 and the Ensoniq/Creative AudioPCI don't. Windows 2000 introduced software mixing through DirectX. Afterwards, cheapie chips went back into not having hardware mixing again. This is why some people have problems with sound in Linux. They have a cheapie, integrated POS sound chip, like the C-Media, i810, nVidia nForce APU, Realtek, etc, and they cannot do hardware mixing. Creative Labs is fortunately one manufacturer that is still making chips with hardware mixing. Audigy series seems to do this. The CS46xx cards (like the Turtle Beach Santa Cruz) are great alternatives as well.
I'm betting that this was the real problem with the author of the article. If anyone wants a high-quality and CHEAP soundcard that works great with the Linux ALSA drivers, they should buy a $5 Aureal Vortex or Yamaha PCI card from Ebay. The Aureal cards do hardware mixing and also have a hardware graphic equalizer. The Soundblaster Live Value cards are also good choices, and can be purchased for $10-$15.
For your Windows Audigy needs: (Score:2, Informative)
Also, its technically possible to have multiple outputs out of your soundcard (read this like in "i got some music playing from the speakers, and also game sounds from earphones plugged in the front panel"). But you know what? Creative drivers makes this thing impossible. But the hardware admit it!! Sucks, isnt it?
Heres your savior: The KXProject [lugosoft.com].
If you dont mind going into complicated stuff (you use Linux, right? it shouldnt be a problem then), you can control [lugosoft.com] how the soundcard should behave when it got some audio input. For example you can shoot the line-in to the front earphone plug, normal (aka WAV/mp3) sounds to the main speakers, so on. that picture [lugosoft.com] speaks for itself.
Did I mention free, too?
So there. Have a nice day!
Re:Huh... (Score:5, Informative)
Re:Damn (Score:2, Informative)
Well, Win95 was not initially released with USB support. So that's where that comes from, because it is partially based in reality. Mods: don't hurt me, I'm not posting any flamebait, or anything else bad. I'm trying to be informative.
Re:Notice... (Score:2, Informative)
Most soundcards will gladly accept the plain SoundBlaster driver, assuming your soundcard is ISA. Otherwise, what you've observed in Windows is *NOT* the soundcard emulating a SoundBlaster, but a SOFTWARE INTERFACE emulating a SoundBlaster. PCI and all that. I see these emulation drivers in autoexec.bat files on friend's PCs when I fix them.
If his testing environment was truly a VM setup to emulate a basic SoundBlaster, he should have known enough from the VM documentation to install the soundcard. After all, the "plain-vanilla SoundBlaster" existed before the days of PnP. If PnP is out of the question, tell me how a distro should be able to detect a generic SB out of the blue? Just randomly initialize 0x220 and IRQ 5 while crossing your fingers?
Honestly, the more I read into this article, the more it sounds like one of those narrow situations that distro makers don't have the telepathic abilities to forsee. And I find it just impossible to believe that every distro on the market right now doesn't include a sound configuration utility that gives you the chance to say "8-bit SoundBlaster on this memory address and this interrupt."
Re:WARNING! (Score:5, Informative)
Re:Not necessarily... (Score:3, Informative)
Linux has trouble with bleeding edge stuff and stuff that uses almost, but not quite, compatible hardware. The later seems to be a problem with cheap hardware and is usually fixed as soon as some developer gets a chance to look at it and spend a few minutes adding the needed changes to the normal drivers.
My question is. Why shouldn't devices come with drivers installed on the device themselves in a platform-indepedant language? Let Windows, Linux, FreeBSD, MacOS, or whatever compile the real driver from that abstract driver when first ran. Instead of updating a driver on the OS you could update it in the flash mem of the device and then let it recompile and run the new driver from the device. Then all OS's would have better driver support - even Windows. This wouldn't be to hard to implement as a standard for new hardware so why isn't it done? Legacy hardware could still have the drivers written in this abstract driver language.. you'd just obviously have to keep a legacy driver cache for your OS to compile when it found those devices. You'd also get the benefit that the drivers could always be compiled to get the best use out of your hardware while being a transparent operation to the user.
Re:WARNING! (Score:4, Informative)
Solution: get a volume control program for X, and leave it running at all times, thereby blocking other programs' attempts to control the volume.
Missing an essential part of the article... (Score:4, Informative)
Well, right there in the article it says it DID work on SOME Linux distros. Why would it work on one and not all? Why isn't there a centralized LINUX device driver database that every distribution uses in it's install? Why should we depend on HW manufacturers to write umpteen odd versions of their drivrs for umpteen odd flavors of Linux? One centralized repository, one way to handle devices and drivers. If someone doesn't want to use this DB, they are welcome to try a DriverDB-less distro.
Re:Huh... (Score:1, Informative)
Well, the first problem is with you. You just completely changed the terms of discussion.
Do sports-car enthusiasts think it's a problem that I never learned to drive a standard transmission? Are the going to redesign their cars for me? Of course not.
"Devs" are not just "enthusiasts". Even ignoring for a second the reality of ideas/bits being separate from cars/atoms, your analogy is critically flawed. A better comparison might be between car engineers and devs. Now, do professional engineers think it's a problem that most people can't drive a stick shift (no longer "standard", sorry) transmission? Obviously - look at modern car designs.
Re:RH and MDK testing..... (Score:3, Informative)
Got that with mkd10/SBLive! too. Turned out to be a positive feedback from the mic input (too close to the speakers). Muting it fixed the problem.
I'm sorry. I take it back (Score:3, Informative)
Fred was editor-in-chief, but I think I have the time period wrong. According to this web page [langa.com] Fred was editor-in-chief from 1988 to 1991. This was after the change from Robert Tinney paintings on the cover to photographs, but still while Steve Ciarcia had his Circuit Cellar [circuitcellar.com] column there.
Re:Holes that windows plugged a decade ago... (Score:5, Informative)
Unless you're looking to do pro audio. Then you'd want a Mac-compatible card from CreamWare, Alesis, Digidesign, Event, Lucid, Ensoniq, Opcode, Lexicon, RME, Lucid, Sonorus, Echo, or M-Audio, among others.
What, did you think that all those Macs in recording studios were using the built-in audio to run ProTools?
k.
Binary-only modules. (Score:4, Informative)
You know, there's a reason Linux doesn't work well with binary-only drivers. And that's because binary-only drivers are a bad idea for Linux.
--grendel drago
He was using VIRTUAL PC - Hardware is not relevant (Score:2, Informative)
The problem most likely comes from the author's confusion between a real and virtual computer. His REAL PC might have had a perfectly ordinary mainstream video card. That does not mean that the emulated PC has the same features, or could use the same driver.
If he had manually configured his virtual linux installation as sound-blaster compatible it probably would have worked, but then again who knows what kind of sound hardware the latest version of MS VPC likes to emulate.
This also explains why he was able to run Windows 95 and Gentoo on the same computer - imagine trying to build a real computer that will happily run both.
Re:They're Getting Desperate (Score:3, Informative)
andy
Re:WARNING! (Score:2, Informative)
- SPDIF output will almost certainly give you trouble, because the chip only seems to be really
happy with 48KHz. I found no way to get 44.1KHz audio to come out the SPDIF port, except to use something like sox to resample it at the higher rate.
- Some programs may play back audio "too fast" because again most ripped MP3 tracks are based on 44.1KHz audio and the chip spits it out at 48KHz for some reason.
Not all programs do this (xmms is okay, for example) but some do.
- At least one version of ALSA I worked with had a bug where you simply could not get it to unmute if the volume was currently set to zero. It'd pretend to unmute, but you wouldn't hear anything. Solution was to use aumix instead of alsamixer to initially set the volume. Once it was truly set nonzero, alsamixer could then deal with it.
In general, you will probably be much happier if you pick up a real sound card that's known to work well in Linux. The VIA stuff can be made to work, but I wouldn't consider it to be working well.Oh, and the USB driver for some recent VIA chipsets apparently has major problems as well. Just a heads-up.
Re:RH and MDK testing..... (Score:2, Informative)
Maybe you have another device that could be mistaken for a soundard, and hasn't been picked up until recent distros? Just thought it might be worth mentioning. Hope you get it working.
Re:Huh... (Score:3, Informative)
Replace the "put in a driver CD" step with "click the K menu, go in 'System Setting' sub-menu, click 'Printer Configuration' and answer a few simple questions", and you pretty much have my Linux experience of installing a new printer under Fedora. I know nothing about CUPS, yet I print. How come ?
Most digicam today are USB Mass Storage Device, just like your thumb drive. You do not need drivers for these. For the rest, GPhoto (now FLPhoto) come installed on just about every modern "desktop" distro and work with all the camera supported by Linux.
In the bottom right of my screen, a big, red flashing "!" tell me when update are available. I just click it, answer a few simple questions, then my system get updated. Just like Windows Update, except you don't have to reboot.
Also, if you want to stay informed about security update, there are better channel than Bugtraq. Most (all ?) distribution today have mailing list specifically for their security advisory.
As I demonstrated earlier, this is irrevelant anyway as Linux update does not require knowledge of these tools (if you are proficient enough to click a flashing red "!", that is). Instead, go with Windows and teach them about anti-virus, how to safely use email, spyware removal and other user-friendly concept.
I heartily agree that Linux have it's flaws and do not want to paint a too rosy picture of the situation. However, I see many armchair critics around here who make a lot of uninformed claims about the state of Linux usuability. Welcome to 2004; nobody use Slackware 3.0 anymore.
obviously you were never a hardware tech (Score:5, Informative)
i spent HUNDREDS of hours searching for drivers and changing default settings trying to get soundcards (from turtle beach to via to sound blaster compatible...) working in windows 95. as another poster said, it's not because of windows that these worked (or didn't work) it's because the drivers were well designed (or sucked ass).
it's the manufacturers fault for not providing linux drivers. but we have to remedy the situation by picking up their slack.
that said, i've configured around 8 computers with linux. i never checked the HCL first. and i got the sound to work (even on board sound) to work every time. maybe i'm just lucky but it seems that if you know what you are doing you'll get it to work. i didn't say it's easy.
Re:Was it really Soundblaster? (Score:3, Informative)
This guy is lying: here's the proof (Score:5, Informative)
His point is thus moot and shown for what it really is: FUD. Big, stinking, FUD of the worst kind.
Couple this with the fact that he does not give out the chipset model of the built-in sound card and I do not believe a word he wrote and neither should you.
Re:RH and MDK testing..... (Score:1, Informative)
http://opensource.creative.com/
How does Red Hat make this work on install that few others can? Have they closed off the source to the GPL project(s)?
Creative releases Linux drivers. Everyone includes them in their distros. Most are at least partially broken. This is Creative's fault, how?
Dumb fuck. Gets your facts straight.
Re:They're Getting Desperate (Score:2, Informative)
<nitpick>
Hmm, a quick google [google.com] suggests otherwise, win2k actually includes a driver for the AWE64, and I can confirm this by having one working in a win2k machine.
</nitpick>
Re:Huh... (Score:2, Informative)
Yeah actually. My PSC-2210 worked perfectly out of the box for all the mentioned features. It acutally does a lot of stuff much better than running under windows. For instance, I can set a cron to sync the time on the printer with a timeserver so the time is always up to date. It may be that you have to install scanning software, just as you would with windows, but that can't take more than 2 minutes with an RPM or whatever you use.
Most "desktop" distros come with just about everything included as a kernel module, so it is very rare that you actually have to install or download a driver or recompile a kernel. I happen to run Gentoo, and I generally enable as much as I can under USB/firewire/gamepads as a module. This gives me the ability to just buy something and plug it in without having to recompile.
If everyone hasn't checked out samsung lately for printers, take a look. I picked up a new ML-2152W for my office printer and it came with a Linux disk. It was very slick and everything installed at least as fast and smooth as it did under windows.
If you take a step back, you can really see the difference between a good vendor and a not so good one. When it comes to drivers as mentioned in the article, it surely is a vendor issue. MS didn't write all the parts of all those drivers it uses for autodetection itself. This is the work of the vendor. You can't broadly blame "Linux" for this. It just doesn't make sense, for two reasons:
a) The vendor should have but didn't port their driver to linux/BSD. If they didn't make a driver for windows, it wouldn't have one either.
b) The vendors generally block the efforts of open source developers by keeping their specs a secret. Basically you are asking the OS community to reverse engineer the hardware, which may be illegal in some places.
For now, you just have to be conscious about the hardware you buy for a linux desktop. Ultimately it is going to be the vendors that bring the support to linux. Look at the efforts of samsung, nvidia, ibm, high point, etc. It is going to be the vendors! Not SuSE, Redhat or Mandrake, just like it isn't Microsoft that makes Windows driver support so broad.
I haven't had hardware support problems under linux since I was running redhat 5.0 on my desktop and couldn't get my webcam to work. Maybe I am just lucky. More likely, I know what to shop for.
People who write articles like this and make some similar comments I have seen are the ones who just don't get it. Eventually some distribution of linux will get it right for them, but for now, they should probably look elsewhere. Those of us that do get it, are happy now and have little concern for those that don't.
daniel
Re:Huh... (Score:2, Informative)
Usually for any given car the manual transmission has one or two more forward ratios than the automatic transmission version. Even current model vehicles with five speed autos are competing with six speed manuals. There's a reason why race car run manual boxes with as many gears as possible. Note that I'm referring to race cars such as rally cars that are setup to essentially run on normal roads. Drag transmission like air shifted Lencos are more different to regular auto transmissions than regular manual gearboxes, and they're usually manually shifted anyway, they just don't require clutching between gearchanges. Looking at drag cars to decide what is appropriate for a road car is ludicrous as you'd wind up trying to run things like wrinkle wall carcass slick rear tyres and ladderbar rear suspension which are extremely dangerous and in most cases illegal on public roads. Note that I'm not saying that all this equipment is mutually inclusive, but that equipment focussed on making a race car go very fast in a straight line is generally totally inconsistant with useful function in a car meant for use on public roads.
Add to this the fact that the heavier gearsets in auto transmission (big chunky planetary gearsets with integral clutches) and the hydraulic pumps to provide the hydraulic pressure (to run those integral clutches) AND the torque converter all suck horsepower to function and you're quickly running out of good reasons to run an auto for performance purposes. Computer controlled sequentials along the lines of BMW's SMG dodge a LOT of these issues, and shift faster than most if not all drivers with a full manual gearbox, but then they let you select your own gears as well.