Testament Against Non-Free Software

The idiots who claim non-Free software is somehow better because it’s non-Free should have been here last night. The Little Woman had her GNU/Linux PC crash. Traceback showed a bug in the “nouveau” driver. She was also using an older kernel that came with her distro. I thought I would upgrade her to a newer kernel hoping to leave that bug behind…

  • Linux 3.15 was difficult to build on her machine. She has very little RAM so it was “make -j 1” or nothing… (in retrospect, I could have copied the configs over and built on Beast, 4X faster, but I was “almost there”…)
  • After an hour or so I had a working kernel configured for her particular machine but the “nouveau” driver wouldn’t use the full resolution of the monitor… not even close.
  • I decided to use Nvidia’s non-Free driver. It works fine on two other PCs running this kernel…
  • Nope. It wouldn’t build without reconfiguring and rebuilding the kernel several times to get it right, but after many hours I got it working.
  • That’s a lot of hours but it was compounded by Nvidia not supporting her hardware with their latest driver, you know the one that’s a click or two from downloading? I had to hunt for it, looking at several “legacy” drivers all of which Nvidia said would work.

So, the result was I pulled an “all-nighter” just because Nvidia didn’t contribute full-heartedly to the Linux kernel at least in the old days… Their EULA essentially denies users the right to reverse-engineer their stuff so it’s on them. Now, 3.15 updates every week or so… I am sorely tempted to upgrade her machine to another motherboard with anything but Nvidia on it. 8-( If the next upgrade is not real smooth Nvidia is toast around here.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology. Bookmark the permalink.

55 Responses to Testament Against Non-Free Software

  1. dougman says:

    I always enjoy a good M$ bashing, but that guy’s website is full of half-facts. He should at least explain this as the Forbes article did.

    Once a Windows desktop operating system leaves mainstream support, it enters the extended support phase—the very same support phase that Windows XP found itself in from early 2009. On April 14, 2009, Windows XP exited mainstream support and entered the Extended Support phase; Microsoft continued to provide security updates every month for Windows XP; however, free technical support, warranty claims, and design changes were no longer being offered. Extended support ended on April 8, 2014 for XP.

    So this is a non-issue really, the deadline in 2020 for Windows 7 is the one everyone should be aware of.

    http://www.forbes.com/sites/gordonkelly/2014/07/10/microsoft-windows-7-mainstream-support/

  2. Good one, JD! “if you are on Windows 7 you have six months from yesterday before Microsoft unilaterally cuts off your access to security patches. You know, those holes that they are responsible for, the ones you pay them to not put in products in the first place, the ones they promised wouldn’t happen in this version, or the last, or the one before that. On January 13, 2015 you can’t get patches anymore unless you pay Microsoft a lot of money. Once again they will be making the patches until January 14, 2020 but they won’t let you have access to these vital security patches. Also note the subtle hint about paying up in the last column too, still don’t think this is extortion?”

    Yep. This is the Wintel Treadmill in action, with whips… I’m done with that. I figure M$ will maintain revenue more or less flat for a few more years by charging more to the last ones in the room after the light goes out.

  3. That Exploit Guy says:

    I just thought I would leave this here

    You know that guy that won’t stop talking you about, well, stuff, and you always try your best to nod and smile at him so he won’t flip out and start killing everyone in the office building?
    Well, apparently that guy now runs a website called “SemiAccurate.com”.

  4. That Exploit Guy says:

    To clarify on the “patent” “issue”:

    IEEE 802.11i specifies 2 modes of operation for AES encryption: CCM and OCB. The OCB mode is currently under patent protection and therefore, unsurprisingly, the only mode that is mandated by the standard is the CCM mode. In other words, there are no patent issues in Wi-Fi certification or 802.11i compliance.

  5. That Exploit Guy says:

    @ Ohio Ham

    In fact there are very good reasons not to implement AES in the wifi controller. Power usage.

    So, you’ve spent an entire wall of text babbling about that NDISWAN nonsense and you now just expect me ignore that and talk about, what, power consumption?

    OK, why not play along?

    Compared to the CPU, the difference a NIC chipset makes is between zero and nada. After all, we are still talking about average PCs, right? Or are we now talking about something made out of rubber bands and a box of paper clips?

    That Exploit Guy this happens because not every controller implements. Linux and OS X don’t need it implemented in the controller.

    Yep, rubber bands and a box of paper clips.

    Perhaps cheap Taiwanese Wi-Fi adapters just aren’t cheap enough.

    Intel and ARM(some models) and many other cpus have AES accelerator instructions.

    AES instruction sets in general-purpose CPUs are only recent introduction. AES support in Wi-Fi controller chipsets, on the other hand, has been around ever since WPA2 became a thing. Why do you think WPA2 support should be built around something that even the computer I am using right now doesn’t have instead of something that is as common as air?

    Atheros ar5523 chipset for example is as dumb as bricks wifi controller does not implement RC4 or AES in the wifi controller.

    Wrong.

    Basically a stack like this one. So you are implement aes-ccmp you do it one way you run patent 1 do it another patent 2 and so on.

    Again, you are drivelling on about this ridiculous notion of an AES “patent”. Let me get this straight: it doesn’t exist, nor am I going to pretend it exists in any way. Get over it.

    mac80211

    mac80211, as described in kernel.org, is a framework for implementing drivers for SoftMAC wireless devices.

    SoftMAC refers to Wi-Fi devices that implements MLME in host software. It has fundamentally nothing to do AES encryption or decryption, which is typically done in hardware.

  6. oiaohm says:

    http://answers.microsoft.com/en-us/windows/forum/windows_7-networking/windows-7-wireless-network-problem-windows-7-fails/ce399590-1c0d-482e-bc7e-bd4016e154b2
    http://support.microsoft.com/kb/893357
    KB893357
    That Exploit Guy this happens because not every controller implements. Linux and OS X don’t need it implemented in the controller.

    In fact there are very good reasons not to implement AES in the wifi controller. Power usage.

    Intel and ARM(some models) and many other cpus have AES accelerator instructions. So its more power effective to decode in the cpu than in the wifi controller. So reason for running dumb as bricks wifi controllers is longer battery life why because a slower and lower grade silicon in the controller would burn more power doing the decode.

    Atheros ar5523 chipset for example is as dumb as bricks wifi controller does not implement RC4 or AES in the wifi controller. Its all implemented driver side using your cpu. Only thing in the controller is tune the radio and provide a byte stream hopefully broken correctly into packets(yes I do stress hopefully). A few people have found out this the hard-way attempting to hook this type of dumb wifi device up to a PI as well and run out of CPU time(sometimes you want a wifi device with brains particularly when you have a cpu without accelerated AES instructions). There are quite a few wifi cards and controllers that are fairly much bare bones transceivers no brains.

    Atheros ar5523 chipset is still in mass production and found inside new like dell laptops.

    Cheap taiwanese wifi controller provides AES is correct. The issue here is a even cheaper taiwanese wifi controller does not. Stripping the encryption processing out is a small bit of silicon so less production cost also you don’t have to pay stupid patents.

    Yes another reason todo it in the CPU. A CPU with accelerated AES instructions has already paid the AES collection of patents once for implementing it in hardware.

    AES-CCMP is not implemented in the controller go read the kb link DrLoser brought in. kb893357 for Windows XP does not depend on the hardware driver or controller providing AES-CCMP. Depends on the card not to drop WPA2 traffic as invalid. Cards marked as WPA2 compatible back in the time of XP did not implement it in controller either.

    Yes That Exploit Guy Microsoft believed we would not have dumb as bricks wifi controllers so no need to pay off some of the parties that have stupid patents over implementing wifi. Please not the stupid patents don’t cover AES exactly.
    http://www.google.com/patents/US7263186
    Basically a stack like this one. So you are implement aes-ccmp you do it one way you run patent 1 do it another patent 2 and so on. Most are most likely invalid if someone takes them to court. Patent trolls harm end users basically. Its like AES is patent fine but then you want code something using it and there are patents all over the place. It what IBM calls barb-wiring.

    I did not say Microsoft used the most sane way over overloading wifi encryption in on XP.

    That Exploit Guy you are believing a fake story that wifi controller on the whole have encryption processing. Majority don’t have it. This is why loading the wrong driver magically ends up back at square one.

    mac80211 under Linux and net80211 under bsd relations(include iOS and OS X) that are particular for wifi cards are purely design that wifi cards are as dumb as bricks if they happen not to be that was more good luck than good management.

  7. That Exploit Guy says:

    That Exploit Guy anyone running Windows Vista or newer does not have a fall back if their wifi card maker did not implement the protocol they required.

    I am willing to forgive the ridiculous notion of an automatic protocol fall-back. I am also willing to forgive the more ridiculous notion of an AES patent. I am also willing to forgive the even more ridiculous of not being able to use third-party supplicants in Windows Vista and up.

    What I am not willing to forgive is this blatant lack of consistency.

    That Exploit Guy software fall back encryption support on wifi cards happens under XP SP2 to the last version of XP.

    Again, either your controller chipset supports AES-CCMP or there is no WPA2, XP or not.

    Again, if a cheap Taiwanese controller can provide AES-CCMP encryption on the chip, then what is the excuse for using a chipset that doesn’t do exactly that?

    Performing AES-CCMP encryption in the driver is such an unthinkable thing to do you might as well use a bare-bone 2.4GHz transceiver and implement the entirety of the 802.11 protocols in software. Seriously, are we even talking about Wi-Fi adapters at all?

    I used stackable instead of stack because inside networking you have too many things called stack.

    This is supposed to be a joke, isn’t it? Or are just really that clueless about basic technical terminology?

    NDISWAN

    RAS is for handling dialup connections to LAN, mate. The least you could do would be to understand the very thing you were trying to provide a hyperlink to, wouldn’t it?

    So technically WIndows XP SP2 and SP3 support WPA1 and WPA2 100 percent of the time and nothing else Windows does WPA1 and WPA2 100 percent of the time supports

    So technically your entire argument is based on the faulty assumption that AES-CCMP is generally implemented in the driver and the more faulty assumption that the completely unrelated NDISWAN IM driver is what’s responsible for Wi-Fi encryption.

    Who knew by simply asking you to “read the documentation” you would end up bringing your drivels to such a whole, new, ridiculous level?

  8. oiaohm says:

    Also That Exploit Guy you were just been over picking on my rough overview as well.

    [wifi-card]-[wifi driver]-[{[wpa supplement winpcap driver]-[wpa supplement userspace]-[wpa supplement winpcap driver]}/{MS XP SP2 NDISWIN WPA 1-2 driver}]-[OS/Applications]
    Why is it the wpa supplement winpcap driver is in fact it pushes the exclusive load option so its not a share winpcap.

    That Exploit Guy so either I could write the simplified with a stack of middle stuff compacted or I could write the more exact one that is also missing major amounts of detail and is way harder to read.

  9. oiaohm says:

    http://mrpogson.com/2014/07/09/testament-against-non-free-software/#comment-163471 If you go back first bit on this topic I mention XP has fall back and later does not.

    That Exploit Guy anyone running Windows Vista or newer does not have a fall back if their wifi card maker did not implement the protocol they required.

    That Exploit Guy software fall back encryption support on wifi cards happens under XP SP2 to the last version of XP. Does not happen under newer versions of Windows.

    I used stackable instead of stack because inside networking you have too many things called stack.

    http://msdn.microsoft.com/en-us/library/windows/hardware/ff554846%28v=vs.85%29.aspx

    Read this is 5.1 XP NDISWAN. Intermediate Drivers that are the new version of NDISWAN exist in Vista and later these are also stackable on top of vendor NDIS driver as well but there is no software wpa NDIS Intermediate Driver Vista and later. So sorry NDIS drivers are stackable by design stackable. Intermediate drivers/NDISWAN drivers can be more than 1 on top of card own driver. My rough overview matches how it hooks up.

    wpa suppilcant equal done in XP by Microsoft is using the NDISWAN interface. No user-space code required.

    NDISWAN drivers due to change in NDIS design between XP and Vista required a rewrite to become a Intermediate Driver and Microsoft never ported it so effectively reversing the XP SP2 patch.

    Current windows day windows you buy has no fall back WPA2. Windows XP before SP2 has no fall back. Windows XP SP2, SP3 have Microsoft provided software fall back for WPA1 and WPA2 and that is it. So only Windows XP SP2, SP3 are what you call officially supporting WPA1/2. Everything else allowed third parties to plug in what ever they like.

    There is a wpa suppilcant fork that I used that was a NDISWAN driver this was more stable than the libpcap version. Reason no packets could sneak around NDISWAN driver.

    “Each vendor card is its own encryption stack under Windows”
    This is true. Under XP SP2 and SP3 with Microsoft NDISWAN driver for WPA1/WPA2 provides software backup in case the card vendor does not implement something like WPA2 in their encryption stack. This is missing from Vista on.

    XP SP2,SP3 does not load the NDISWAN WPA driver on every wifi card. Instead looks at the NDIS report from the driver and decides if it has to or not.

    So technically WIndows XP SP2 and SP3 support WPA1 and WPA2 100 percent of the time and nothing else Windows does WPA1 and WPA2 100 percent of the time supports

    That Exploit Guy everything in the network stack design of windows means Microsoft could implement a fall-back driver. Why did wpa supplement open source ndis driver disappear answer is driver signing. libpcap version avoids driver signing and is horible.

    Does Microsoft in Vista + provide the moving parts to support WPA2? The answer is no.

    Does Microsoft provide the framework that third parties can choose to implement WPA2 or any other network encryption in Vista+ they choose? The answer is yes.

    The first one being no is why you have the issues appear.

    To use a Linus term here. Wifi card vendor WPA support is like trying to herd cats under Windows Vista +. Each of them doing their own implementation with no common implementation that can be applied to them all.

    Every other non Microsoft platform has WPA implementation independent to card vendor. This allows all Linux/OS X/…. systems no matter the installed wifi card to talk the same protocols over wifi. Less headaches.

  10. That Exploit Guy says:

    @ Seriously, is this random bunch of letters supposed be a name?

    I did not confuse RC4 with RSA.

    Yet, everyone familiar with RC4 knows pretty instantly what exactly is wrong with your comparison between the algorithm and AES. I mean, did you not at least try reading your favourite source Wikipedia first?

    Both are not one way.

    You see, the “one-way” part was the giveaway about RSA, but that wasn’t the main part you have got wrong about RC4. After all, MathWorld has already told you that, though debatable, RSA is sometimes thought of as an one-way function and it is indeed a cryptography algorithm featuring a private key and a public key.

    That Exploit Guy the fact NDIS includes means to determine what driver provides is how come using NDIS drivers under Linux it possible to support WPA2 and full WPA2 Enterprise on a WPA1 or WEP only NDIS driver.

    That’s something I have not tried and have no interest in verifying on my own. However, as far as I am aware, wpa_supplicant and suchlike uses libpcap to capture and thereby manipulate network traffic at the link layer. This is why the Windows version of wpa_supplicant requires WinPcap in order to run properly.

    That Exploit Guy XP sp2 also includes a software support on WPA1 or WEP only wifi cards to make them do WPA2.

    Really? That’s not even what you said ten or so comments earlier. In fact, you said, “Each vendors card is its own encryption stack under Windows.”
    So, what is it now? Does Windows provide the moving parts to support WPA2 or not? Make up your mind.

    Of course their is absolutely no need to. the NDIS drivers are stackable. Stackable is how WPA supplement did it under XP and 2000.

    A stack, not “stackable”. What do you think this is? Dungeons and Dragons?

    [wifi card]-[wifi cards own driver]-[wpa supplement driver or XP sp2 patch adding all the extra functionality]-[OS and its applications]

    Wrong. Did you even read the documentation, for crying out loud?

    That Exploit Guy sorry there is very little excuse for Microsoft not to provide a software fall back the code is already done for them under a BSD license.

    Again, some comments ago you said, “Same NDIS driver loaded under XP last service pack will show and work WPA2 encryption due to the software fall back.”
    So, does “fall back” occur in Windows or not? Do you love lying just so much you don’t even care about consistency the slightest?

  11. oiaohm says:

    That Exploit Guy come on are you attempting to prove you are as big of idiot as DrLoser. I did not confuse RC4 with RSA. TKIP that is RC4 and CCMP that is AES are the two packet encryption items WIFI uses. Both are not one way. RSA is only used some Wifi Enterprise modes. Not all Wifi Enterprise modes. Containing CCMP in hardware of the wifi device does not mean it can perform more than a WPA1 handshake.

    That Exploit Guy the fact NDIS includes means to determine what driver provides is how come using NDIS drivers under Linux it possible to support WPA2 and full WPA2 Enterprise on a WPA1 or WEP only NDIS driver. That Exploit Guy XP sp2 also includes a software support on WPA1 or WEP only wifi cards to make them do WPA2.

    The issue here is just because a driver only provides X there is absolutely no requirement to limit the device to that yet this is what Windows Vista + does.

    Yes Exploit Guy is correct its not possible to make a NDIS driver accept an OID request it does not support. Of course their is absolutely no need to. the NDIS drivers are stackable. Stackable is how WPA supplement did it under XP and 2000. [wifi card]-[wifi cards own driver]-[wpa supplement driver or XP sp2 patch adding all the extra functionality]-[OS and its applications]

    Wifi drivers and cards include means to send and receive raw packets. This was built in from the start to allow new encryption protocols to be implemented.

    That Exploit Guy sorry there is very little excuse for Microsoft not to provide a software fall back the code is already done for them under a BSD license.

    Sorry That Exploit Guy Goggling an arguement will miss the fact NDIS drivers are stackable. Also miss the fact wifi drivers and devices allow raw packets by design to allow software fixing. Result there is absolutely no requirement to give users Wep only or even no encryption options. Only reason for this to be possible is a bad implementation by the OS maker. In this case Microsoft.

  12. That Exploit Guy says:

    @ Gene Ray the inventor of Time Cube
    Ok RC4 is not modern. AES is modern and its not one-way and does not use large prime numbers.
    Ha! You must have confused RC4 with the RSA algorithm, which incidentally is also conjectured to be a one-way function.
    Nice try there, I must say.

  13. That Exploit Guy says:

    One has to wonder why, despite what oiaohm says, no one has ever mentioned this mysterious “WEP failover” on the Internet.
    Of course, this is easily explained by how NDIS actually works, namely, that the framework has already provided a simple way to determine what the driver supports and what it doesn’t.
    Of course, oiaohm will always come up with some ridiculous story to explain why it is possible for a driver to accept an OID request that it doesn’t even recognise.

  14. That Exploit Guy says:

    @ Bob

    Chuckle. One of the huge advantages of FLOSS is the reuse of software so creators of applications don’t have to reinvent the wheel.

    Chuckle. Unless you want to MacGyver your own Wi-Fi adapter, I don’t see what such “reuse” is good for.

    Generally, either AES-CCMP is supported by the hardware or it’s not supported at all. After all, even cheapo Taiwanese WLAN controllers do AES-CCMP on the chip. If you have to resort to implementing the feature in the driver, then I doubt your controller can even handle 802.11 frames on its own. In fact, it might as well be a bare-bone 2.4GHz transceiver for all that matters.

    Also, seriously, out of all things you could possibly buy into, you picked oiaohm’s psychotic drivels?

  15. A Pesky Interloper wrote, “Just stop calling it “the beast,” having pitiful four cores does not make your computer a part of devil’s inventory”

    I designed Beast to run as a terminal server to demonstrate what is possible in schools. I’ve had 24 students run on it simultaneously and they all felt they were on the fastest machine in the planet. That’s what file-caching does. No waiting. Typical uses of desktop application need ~50-100MB of RAM and a few hundred MB for the OS leaves a gigabyte for file-caching. In such use all the RAM is put to good use and the CPU never reaches 50% utilization. It is a beast thanks to GNU/Linux. I get to use all of its capability when I build kernels these days. Lately, with custom kernels, it takes about 7m to build one and 1m to install it. There are things I could do to speed up Beast but why should I bother? Faster drives or networking would give me a tiny performance boost. Indeed, I am looking for a reduction in performance just to reduce heat/noise/cost of operation. Beast has a 45nm CPU. Moore’s Law works for AMD as well as Intel. An optimal solution for me would use 28nm and just dual cores. Faster RAM would offset most of the performance reduction and with faster clock there would not be much, if any reduction. It’s the performance of the whole system that matters, not how much I spend on each component. I am seriously considering an ARMed replacement which in 2014 are good enough:2gHz, 8 cores, 4gB RAM, gigabit/s… They are releasing versions for servers this year and desktops sooner or later. Power consumption is expected to be a bit over 2W per core. Beast uses about 25W per core.

  16. A Pesky Interloper wrote, “Robert really should move himself to a thin client; that would probably end his fascination with the idea in relatively short order.”

    Nonsense. I’ve used thin clients for years at many of the schools where I worked. I made the most powerful machine on the LAN the terminal server and used thin clients for just about everything else. I have a box of 10 thin clients floating around from the old days. Unfortunately, they don’t do 1920×1080 so I don’t use them much here but I often use an application on one of our PCs from others using X. It’s all good. The only thing they don’t do well is full-screen video. I can route audio to them as needed. Same with USB devices. GNU/Linux is very flexible and efficient. I use X all the time except for software installation and the only difference for a thin client is the network lag which is minimal here at 100mbits/s and insignificant at 1000mbits/s.

    Really, if I use ssh -Y user@someserver "some application" the window for that application pops up on my screen just the same as if I were using the application locally and it’s all transparent if I set that into my icon on my panel. Typically, in a school, I would also share the users’ files over NFS so that any network lag would be similar for local or thin applications. Do the maths. At 10mbits/s it does take a perceptible time to draw a screen. At 100mbits/s even the fastest typist is only a few characters ahead of the screen. At gigabit/s very few users would have any clue they weren’t on a normal thick client. It’s all in the configuration. About the only time I ever had a problem with thin clients was a class of grade 1 students whose favourite application wanted to redraw the screen several times per second. After a few users, it was pretty sad but I doubled up the NIC and it was acceptable to them, my harshest critics.

    see 20 ancient notebooks used as thin clients. No one complained about performance once they were tuned up. It’s all about configuration. Done right, performance is great for the usual browsing/office suite stuff. It’s very efficient, too, with the good hardware working hard and the old/slow stuff just idling. After all, X worked on ancient machines just fine. It still does. The only concern is screen resolution. Lots of old machines had video cards that can’t do 1920×1080. These guys were using the terminal server by wireless back in 2005. Resolution was probably 1024×768 in those days. Today, that would be inadequate to run a big wide monitor like I use.

    If you want to see thin clients done well in a medium sized organization check out what Dave Richards does in Largo, FL. He uses several large servers to run sessions and various applications for hundreds of users. Each thin client has icons for applications on any server and it’s all transparent to the user. He only has to maintain a few servers instead of hundreds of clients. That’s the right way to do IT. He used to use X straight but now uses NX so the users have stateless clients. They can turn off a client they are using, walk elsewhere and resume the session anywhere in the city as if nothing happened. No one complains of performance because it is superior to anything the usual thick client can offer. Everything is in RAM almost all the time so there’s no waiting on a hard drive.

    Pesky Interloper also wrote, “his wife deserves a PC that actually works for a change, methinks.”

    I did a thorough cleaning of her PC yesterday. It had dirty fans and clogged inlet filter… so the problem might have been heating. Sensors did not indicate high temperatures but it didn’t hurt. With the latest upgrade to Linux 3.15.5 and Nouveau driver, her machine is working reliably overnight… Temperatures are about 6C lower than they were before the cleaning and nowhere near the danger levels above 70C.

  17. oiaohm says:

    A Pesky Interloper a beast system is normally a system built from parts of other systems. One of my workstations is officially a beast. Not because its powerful but because its built from parts that were all officially written off as systems were declare scrap. In fact not one of the systems it was built from was without a major hardware defects.

  18. A Pesky Interloper says:

    Actually, Robert really should move himself to a thin client; that would probably end his fascination with the idea in relatively short order.

    Besides, his wife deserves a PC that actually works for a change, methinks.
    …..
    @Robert:
    Just stop calling it “the beast,” having pitiful four cores does not make your computer a part of devil’s inventory, no amount of cores would do that, (don’t take my word for it, ask your personal religious zealot, he’ll tell you).

  19. DrLoser wrote, “What’s wrong with storing the 10K images on the Beast? (Buying a new terabyte drive if necessary.”

    I think I have another 500gb/1TB drive laying around unused. It was hiding in the dark one day when I installed Debian on one of our PCs so I just moved one over from Beast. So, that’s a $0 expense except for my time and discomfort (poor eyesight). There’s even some stuff on Beast I could delete like my local copy of Wikipedia. I have a backup on CD somewhere and I may never use it again since I’ve retired. I could just set up a fresh local wiki with my notes/research/etc., and perhaps a backup of this site. That’s just a few gB. I still think Beast is way too powerful for anything I do. Building kernels is about all that stresses Beast at all. Load average is 0.12 for the last 15 minutes and I have been browsing and typing like mad all that time. Beast could easily run everything in this house except for the full-screen video on the media-centres.

    The little woman hates change. It would be extra work for me to set up Beast to look exactly like her setup. I have to move her servers as well as her desktop settings and files. Since we have exactly the same distro and kernel, it is doable. I could bump up Beast’s RAM a bit too for a small sum. He uses ECC stuff but even that’s affordable.

  20. DrLoser says:

    Here’s a silly suggestion to consider, Robert: Dougman is absolutely correct.

    (Told you it was silly.)

    What’s wrong with storing the 10K images on the Beast? (Buying a new terabyte drive if necessary. What’s that, $50?) Just add DropBox for the ~1K that your better half needs on a regular basis, and round it off with a Chromebook.

    Seems like an unusually sound solution, considering the snake-oil source.

  21. dougman wrote, “Get her a Chrome device to replace the existing machine. “

    That would mean moving her 10K images to the cloud, printing problems, etc. I could of course install GNU/Linux on it but she already has a good monitor and keyboard. At our age, a small display is not optimal. She has a modest amount of storage locally but she also has her own FTP server to deal with the scanner and smart thingies. Still, it’s a valid option. We discussed several options today. All she really cares about is reliability so almost anything is better than what she’s had the last few days. Decisions…Decisions…

  22. dougman says:

    Get her a Chrome device to replace the existing machine.

    I have a Slackware UNRAID NAS in a 4U Rosewill case in the rack downstairs, maybe you could use something like that for storage.

  23. One option I am considering is upgrading her motherboard. This would improve her performance, reduce power consumption and get rid of Nvidia. She and I have both been swapping a little so we could move Beast to her machine with this upgrade and I could work from a thin client. We don’t need 4 cores except kernel-building so I’m thinking dual-core:
    AMD A4-5300 APU Dual Core Processor Socket FM2 3.4GHZ 1MB 65W Retail Box $39
    Gigabyte F2A55M-DS2 mATX FM2 A55 DDR3 1PCI-E16 1PCI-E1 1PCI VGA DVI RAID SATA2 USB2.0 *IR-$7* $63
    Kingston KVR16N11S8/4 4GB DDR3 1600MHz 240PIN DIMM Unbuff CL11 1.5V Gold Memory X2 = $90

    So, for about $200 we could have a better system without Nvidia. Both Beast and her machine have noisy fans. This would be a win all around. I could fire up my old Beast occasionally for building kernels. I checked. This mobo can work with or without UEFI enabled. Will have to discuss options with the little woman, including going back to previously used kernels… Performance is still an issue though. Her old machine is a bit slow even for her modest requirements. Going from 1gB to 8gB RAM, 2.2gHz to 3.4gHz, 32-bit to 64-bit, and DDR2-400 to DDR3-1600 would probably please her. We could also consider shifting my current Beast to her office and I switch to a thin client but Beast is short of storage at the moment. I like the upgrade best because it solves several problems existing and will solve future problems too.

  24. I’ve upgraded her PC to 3.15.5 and using the nouveau driver. It works fine up until it crashes. There are still lots of bugs in that driver despite Nvidia’s help.

    I’ve dug into LKML and found this: Machine freeze on latest Linus kernel, seems related to nouveau. They even found a website that triggers the freeze in two different browsers.

  25. oiaohm says:

    http://answers.microsoft.com/en-us/windows/forum/windows_7-networking/windows-7-wireless-network-problem-windows-7-fails/ce399590-1c0d-482e-bc7e-bd4016e154b2
    DrLoser you knowledge is wrong. WPA2 was part of XP SP2 and last XP. Never came part of the official NT kernel. That update was never applied to Vista none of the latter versions of Windows have it. So like first version XP if the wrong drivers is loaded under Windows 7 back to WEP encryption that is broken.

    The fault I linked to only exists in Windows 7 because the Windows XP fix to the issue implementing software fall back is missing. Same NDIS driver loaded under XP last service pack will show and work WPA2 encryption due to the software fall back. Yes this is a case where you can load the identical driver into XP, Vista, 7, 8 and 8,1 and only 1 has WPA2. Yes the one that is the end of life OS.

    DrLoser sorry I have provided evidence in this case so shut the hell up you are wrong.

    The patch in XP SP2 came out of the fact a lot of people were installing WPA supplement on SP1 and before and sometimes this would cause issues.

    DrLoser the Sans controller and the UTF-8 is off topic and I am not getting into it here.

    How wifi is implemented in some closed source operating systems is a Testament Against Non-Free Software or at least against allowing fragmentation in particular areas.

  26. oiaohm says:

    http://en.wikipedia.org/wiki/Wi-Fi_Protected_Access#Encryption_protocol

    I was clearly talking about wifi encryption stack. Of course DrLoser cannot admit for one min I was correct.

    I came very aware how windows network cards did thing in the year 2000 when I had a IBM and a White box machine with 100 speed network cards that would not talk to each other. Why the IBM network card driver forbid netbeui and the other network card would only talk to cards that did talk netbeui.

    This is the nasty reality about the Windows networking stack not only is your wifi driver encryption stack unique per vendor but the driver it self can be the reason why your network packets for particular applications just magically disappear. Linux, QNX, OS X and most other OS the network driver role is quite a minor one.

    DrLoser does get something correct entropy. The big problem with each wifi card implementing its own solution when it requires entropy values and we already know from the openssl cases that at times NSA and others interfere with stuff to make items use bad entropy.

    Of course DrLoser has to prove he is more of a idiot. What encryptions do wifi use. AES and RC4.
    but I believe that all modern encryption algorithms are one-way.
    Ok RC4 is not modern. AES is modern and its not one-way and does not use large prime numbers. Remember the Pre-shared key thing this is not based on primes.

    WIFI WPA2 enterprise most versions us TLS for the handshake but all packets after handshake go back to AES. So you are in fact dependent on how good the wifi entropy generation is for the AES key changes over time.

    Please note I said most WIFI WPA2 enterprise uses TLS not all. Now it gets worse this is where windows drives businesses nuts using WPA2 Enterprise. There are 8 approved modes of WPA2 Enterprise. Android, IOS, OS X, Linux and Open Source BSD support them all. One of the big limitations to WPA2 Enterprise roll out is the fact Windows wifi card drivers are picky on what WPA2 protocols they support with no way to force full conformance on them. Why are they picky Microsoft left this up to the vendors to implement.

    You need uniformity between hardware vendors the OS kernel maker has to step up failure todo so leads to mess.

    By the way lot of websites in TLS handshare use large prime encryption then switch to AES. The dominate volume of encrypted packets are send using two-way encryption.

  27. DrLoser says:

    DrLoser, it seems to me like you urgently need some sleep, since you clearly have no idea whatsoever what oiaohm was trying to say.

    An interesting observation. Does anybody, ever, anywhere, have the faintest clue what oiaohm is trying to say?

    Perhaps you could try paraphrasing his wisdom. Good luck with that.

    No-one is talking about encypted code

    Very true. I jumped the gun. I admit I was wrong to do so. Admitting that you were wrong is a practically unique skill round here.

    Rather like using a spell-corrector. “encrypted code?” Dear Lord.

    Anyhow. From oiaohm:

    Now DrLoser there is a big security issue due to the way Windows does wifi secuirty. How are you sure that the wifi implementation is upto date. Each vendors card is it own encryption stack under Windows.

    Is there any particular reason why this (assuming it is true. It is so banal that I can’t even be bothered to check) is “a security issue,” oiaohm?

    I repeat. To the best of my knowledge, and I am always looking for an agricultural worker with a sheep fetish to improve and extend my knowledge, WPA2 encryption has actually been part of the M$ kernel since XP Service Pack 2.

    Any new thoughts, oiaohm?

    And whilst you’re at it, could you explain to us all how you came to know everything there is to know about the IBM SAN Volume Controller?

    Any version of the IBM SAN Volume Controller would do, oiaohm. In fact, any small detail that demonstrates that you do not just spend your time fiddling with ovines would be appreciated.

    Got one? A detail to do with IBM SAN Volume Controllers, of course. Not an ovine.

  28. DrLoser wrote, “Now, (1) a potty little PC driver, for any potty little purpose, is not written off the back of a white noise generator. It is, in fact, written (to all intents and purposes, since it has to deal with the interface to the card) in assembler. Assembler is not white noise.”

    I think DrLoser neglects the fact that the software driver is controlling a device in hardware, out there, not otherwise controlled by the OS. Hardware devices can and do contain their own white noise generators. It’s rather simple to generate a random stream of bits by pushing electrons through a resistor and Zener diode. A triggering device can periodically sample the voltage/current and decide whether to spit out a ‘1’ or a ‘0’. Many modern devices do have a microcontroller out there that executes the firmware we often debate about. In that firmware can be almost any algorithm known to man. So, it comes down to where the WPA protocol is generated, out there in hardware or in the PC. Since this stuff is mostly closed-source, one cannot know what the manufacturer does there.

  29. luvr says:

    DrLoser wrote:
    “Unfortunately, it is blatant drivel.
    […]
    You still sound like an utter buffoon.”

    Wow… I mean, “Wow!” Talk about drivel… Talk about an utter buffoon…

    DrLoser, it seems to me like you urgently need some sleep, since you clearly have no idea whatsoever what oiaohm was trying to say. No-one is talking about encypted code, but simply about code that implements wireless encryption. Why assembly code would or would not have to have anything to do with white noise in this regard, escapes me.

    You might want to reread oiaohm‘s post and make an honest attempt to understand it (which, admittedly, may require some effort, but really, it is doable) before you decide to spout any more nonsense.

  30. DrLoser says:

    Each vendors card is its own encryption stack under Windows.

    Unusually deep for you, oiaohm. It almost approaches the level of a Zen koan.

    Unfortunately, it is blatant drivel.

    There are two basic elements to encryption, oiaohm. Just as there are only ever four octets in a standards-compliant implementation of UTF-8 … although in this case, the two basic elements are more fundamental.

    1) For any encryption that is worth more than spit, you need a very good source of entropy. Normally, this translates into a separate bit of hardware that generates suitable white noise. (Sometimes it translates into that hilarious Debian bug of a few years back, but let’s not distract ourselves.)
    2) Given a suitable number of bits of entropy, we proceed to the algorithm. I’m prepared to be wrong on this one, but I believe that all modern encryption algorithms are one-way. In the canonical case, this involves a number of very large co-primes.

    (I encourage the eminent mathematicians out there on this blog to take me to task for omitting the latest theoretical developments in the field.)

    Now, (1) a potty little PC driver, for any potty little purpose, is not written off the back of a white noise generator. It is, in fact, written (to all intents and purposes, since it has to deal with the interface to the card) in assembler. Assembler is not white noise.

    And (2) there is absolutely no evidence whatsoever that NVidia or anybody else uses one-way mapping functions to “encrypt” their code.

    You know why there is no evidence for this, oiaohm?

    There is no evidence for this, because you would need some form of Public/Private Key Certificate before you could even use the driver.

    On Windows, MacOS, Linux, Solaris, anything at all.

    Each vendors card is its own encryption stack under every single operating system out there, including the ones that demand the source code and compile that source code within the kernel.

    I tried to make that sound more realistic, oiaohm. I really tried. But you know what?

    You still sound like an utter buffoon.

  31. oiaohm wrote, “Each vendors card is it own encryption stack under Windows.”

    Chuckle. One of the huge advantages of FLOSS is the reuse of software so creators of applications don’t have to reinvent the wheel. The reinvention of the wheel has plagued non-free software since the days of OS/360 where there were dozens of implementations of the “time” protocol… No kidding. If IBM wanted to change anything about time, they had to find all the places where time was used and check that it was done right. Now replace the mainframe with hundreds of millions of PCs and tens of thousands of ISVs (Independent Software Developers). Compare that with a world where FLOSS is universal and GNU/Linux underlies everything… Instant standardization instead of chaos. The “problem” with FLOSS having a few different standards for everything pales into insignificance to M$ and “partners” each doing their own thing for everything.

  32. luvr says:

    Blech… I hate driver problems under Windows. Actually, one of the events that led me to take Linux serious, was the horrors of wireless networking under Windows XP back in the days. I posted my experiences on the Techspot forum (http://www.techspot.com/community/topics/getting-atheros-ar5212-wireless-cards-to-work-with-windows-xp.39651/#post-278876), and it soon became clear that I wasn’t the only one for whom wireless networking under Windows was a disaster.

    Back in those days, I still assumed that hardware would work under Windows by default, and that some luck was needed to get it to work under Linux. No so!

  33. oiaohm says:

    DrLoser to be truthful why windows users think Windows has WPA support at all is that they did not install the OS themselves to find out the reality.

    A percentage of course get to work and due to OEM installing wrong driver(like the Microsoft one instead of hardware makers one) or not providing a WPA 2 driver find out their laptop and it don’t work. Some are forced to buy and insert extra network card so it works.

    Now DrLoser there is a big security issue due to the way Windows does wifi secuirty. How are you sure that the wifi implementation is upto date. Each vendors card is it own encryption stack under Windows.

    DrLoser the reality not everything is roses with Windows. WIfi with WPA2 can be one of those things. The pot luck has more people coming up with working combinations than failing ones. But those stuck with Windows Vista, 7, 8 or 8.1 in the failed pile are in trouble. In fact that KB artical for XP would dig XP users out the hole. Issue is there is nothing like that KB patch for Vista, 7, 8 or 8.1 and it was not built into to everything including past Vista.

    This is one of these horrible flaws. Lot of people want to believe that if something was fixed in XP it fixed in latter versions of Windows. Reality some of the faults of early XP are back in Vista on.

  34. oiaohm says:

    http://answers.microsoft.com/en-us/windows/forum/windows_7-networking/windows-7-wireless-network-problem-windows-7-fails/ce399590-1c0d-482e-bc7e-bd4016e154b2

    DrLoser excuse me Windows 7 and 8 truly don’t have WPA2 support built in. All wpa support that appears in Windows is part of the driver. Under Linux WPA support is part of WPA supplement so if the card cannot do it software fall over exists.

    Windows Vista, 7, 8 and 8.1 only has software fail over for wep just like XP. Not even WPA1.

    DrLoser only a incompetent idiot like you would go and claim a OS has support for something it does not.

    I said clearly Windows it is up to the driver. More fun is the Microsoft provided driver for a wifi card will be missing WPA2 and some cases WPA1. So go to manufacture site download their drivers manage to uninstall the Microsoft drivers and magically your Windows Vista, 7, 8 and 8.1 now has WPA2 support.

    DrLoser there are enough google searchable posts of people mentioning they don’t have WPA or WPA2 support under Windows Vista, WIndows 7, WIndows 8 that you should have became clear that its not a one off.

    Linux can use WPA2 over any wifi card it has drivers for. Windows vista on can only use WPA2 over a wifi card that it driver support WPA2. Windows XP could use WPA supplement to fix a card up. One of the trap to those updating from XP your wifi network card may have to go in bin. Yet upgrading to Linux everything would be fine.

    By the way there is still new cards being made for Windows 7 and 8 that don’t have WPA2 Enterprise support in their drivers. Some of these are in laptops. This is good fun when you are running a radius controlled wifi network and it will not connect. Same card running Linux or OS X and everything is golden. Why both Linux and OS X has software fall over in case driver does not support something from the wifi encryption standards.

  35. DrLoser says:

    Oh, and whilst we’re talking about your tenuous grasp on reality, oiaohm?

    Remember that IBM SAN set-up you promised us?

    It’s very noticeable by its absence.

    Care to share, in the traditional FLOSS way? You’ve had the best part of a year to get your perfect configuration set up. We all need these things. It is what FLOSS is all about.

    Share and care, oiaohom, share and care. Particularly when it comes to proprietary crap like IBM SANs.

  36. DrLoser says:

    You’re really losing what little grasp on reality you ever had, aren’t you, oiaohm? No WPA2 on Windows 7 or 8? Some lunatic assumption about patent costs?

    Try this.

    Sadly, it’s only Windows XP SP2, and therefore a bit ancient … but it beats your pointless links to ISO 10486:2000, doesn’t it?

    Once again, I repeat. Hassle free.

    Nobody claims that “non-free” is inherently superior. That would be stupid. And nobody claims that “free” is superior, well, actually, some stupid people do, but then again they don’t consider the costs.

    And only really, really stupid people … I will go so far as to define them as “hopelessly ignorant,” would insist that any Microsoft operating system since XP Service Pack 2 has not had full support for WPA2.

    This is possibly the most absurd thing that oiaohm has ever claimed. And I have to admit, it has a lot of competition.

    Tell me, oiaohm. How do you suppose that the 80-90% of us out there (I’m not going to argue with the precise figures) who use a Microsoft desktop at work deal with WPA2?

    Show, show us your Pasa-Doble, little pointless twerp.

  37. oiaohm says:

    JD Wifi no WPA2 under windows 7 and 8 is a huge pain in ass. It not that Microsoft allergic to strong encryption its that Microsoft leaves wifi encryption up to the driver maker to implement. Wait Microsoft is kinda allergic they don’t want to risk to pay patents on it.

    WPA Supplicant was an option under XP to fix up poor drivers but driver signing and other cost making drivers for Windows saw the end of that. So yes Chrome OS, Android, IOS, OS X ….. Basically everyone bar microsoft has proper Wifi support. Microsoft Windows is pot luck.

  38. dougman says:

    LOL…..Win-Dohs 8 having driver problems??

    Nothing has changed in the realm of M$, all they ever do is change the lipstick on the bloated pig of an OS that they sell.

    I use Linux Mint myself, and never really had any troubles from using the Nvidia drivers.

  39. JD says:

    @Loser
    You signed up to FLOSS. You get to experience the results of this contract, good or bad.

    Just yesterday, I had to spend a couple of hours helping a friend connect his daughter’s brand new iPod to his Windows 8 laptop. iTunes kept saying “There is a problem with this device”, and then ejecting it.
    1. The local Apple store’s recommendation – get a new USB cable (even though this iPod is brand new).
    2. iTunes’ recommendation – reinstall iTunes. That didn’t work either.
    3. Apple support pointed us here – https://www.apple.com/ca/support/ipodnano/assistant/itunes/ and here – http://support.apple.com/kb/ts3218. Neither of them helped.

    We finally figured out that Windows 8 was too stupid to load the correct driver. We had to go into device manager, then hunt around in the iTunes folder for the correct .inf file, and point Windows 8 to it. So much for a no-hassle solution.

    Later yesterday evening a friend brings a laptop running Win 7 Ent to my home and the darn thing won’t connect to the WiFi. No proper error messages, nothing, just says Windows can’t connect to this network. The same thing works fine with a WAP running WEP, but won’t connect to mine with WPA2. M$ must be allergic to strong encryption. We never did manage to get the thing to work.

    At least with FLOSS, I have a shot at troubleshooting properly, and perhaps even fixing it, depending on how badly I need it. With M$, the best I can do is pray.

  40. lpbbear says:

    I use NVidia video cards in my systems. I have tried to use the “nouveau” driver. Every time I have tried to use it I have found it to be quite buggy. Since I do occasionally play 3d games on my Linux system I have stayed with the proprietary NVidia driver. I haven’t had problems with the NVidia driver for quite a while. My distro of choice is PCLinuxOS. PCLinuxOS has always auto installed the NVidia driver for me and occasionally an NVidia driver update comes in Synaptic. I have never had a problem with the update process. I also update the kernel through Synaptic and again never had a problem with any of this in PCLinuxOS.

  41. oiaohm says:

    Robert Pogson last year I got rid of about 6 geforce 6 series because it failed color calibration. So its not just that the are old a lot of that series are mechanically dieing.

    I have a old TNT2 PCI that can still pass color calibration tests. Getting drivers to run it is hell.

  42. ram wrote, “That must have been a really old NVidia graphics card.”

    “lspci|grep VGA
    00:05.0 VGA compatible controller: NVIDIA Corporation C51PV [GeForce 6150] (rev a2)”

    According to Wikipedia, it’s an “entry-level” device from 2005. Maybe Nvidia is right to think folks don’t use it much…

    Her CPU is dual-core 2.2gHz, about Atom speed. An upgrade might be welcome. She certainly needs more RAM, the way she wastes it… Since we use the same software, more or less, she could be a thin client of Beast and do all right. She rarely uses suspend to RAM… She rarely does full-screen video. We have TVs for that.

  43. oe wrote, “it is Maytag computing compared to the closed source alternatives…hassle free and stable, obviously I not running bleeding edge distro’s.”

    Exactly. In the schools where I worked I only once installed the testing flavour of Debian and that was after I had used it for months on my own PC. It was stable in use and was not dissimilar from Ubuntu GNU/Linux in bug-count. Except for the OpenSSL problem, I’ve not had any trouble with it. That’s one serious problem in a decade of use on hundreds of PCs. OTOH that other OS gave me pain every month, year after year, for updates and weekly for malware.

  44. oe says:

    Interesting sentiment, while I’ll initially used Linux in the late 90’sin grad school for flow modeling work and then in work for machine tool control due to it’s low level coding capabilities (machine tool or lathe control with windows quite frankly scares the crap out of me); nowadays I like that it is Maytag computing compared to the closed source alternatives…hassle free and stable, obviously I not running bleeding edge distro’s. Right now slowing migrating from 10.04 to 12.04 but no rush. BTW closed sources OS migration NEVER worked without disk wiping and clean reinstall. Ubuntu has nailed the seamless upgrade in place from pretty much 8.04 LTS onward (first stated using 4.04)….

  45. ram says:

    That must have been a really old NVidia graphics card. I have had fairly recent (months) NVidia drivers running on cards over five years old. A recent upgrade of the cards, keeping the very same driver, also worked without issues.

    As a general rule, running bleeding edge software on older legacy equipment is begging for trouble. One would expect some “issues”.

  46. DrLoser wrote, “I don’t waste a weekend farting around.
    I rely on “hassle free” software.”

    My time wasn’t wasted. I love solving problems and fixing what’s broken. I chose beta software for our distro of choice because of its features and I expect a few problems. This was one. The software is beta because it’s mostly from the Debian “testing” flavour and 3.15.4 is a very recently released kernel. It’s been tried by only a few thousand people, I guess, some of the Linux developers and other people who think as I do. I could have taken the current release of Debian and been OK but there would be lots of features missing that I know about and want. Ignorance is bliss. If you don’t want to know what’s coming in the next release of Debian, don’t use the “testing” flavour. I do. Despite all my software except a few applications being beta, this is one of the few bugs that I found and the little woman was running for months without it being an issue. Her PC is light on RAM and she rarely closes a window, so the system was under stress, a great way to test stuff. She’s never lost any important stuff due to GNU/Linux which she regularly did with that other OS, you know, the one made by paid professionals working in office buildings…

  47. DrLoser wrote, ” Nobody ever claimed that “non-free” is better per say. That is an absurd allegation.”

    DrLoser wrote, previously, “It’s about time to admit this to yourself, Koz: the amount of effort spent on FLOSS is, at best, a cost-benefit deficit. If not, what have we got? A diseased and broken desktop, which can’t even decide which audio stack it works off, for God’s sake, and a simplistic LAMP gateway model (which might as well be WAMP or LNMP or ****, because you could start from scratch and come up with the same thing in six months).”

    OK, I agree, DrLoser, your opinions are sometimes absurd, certainly your opinions about FLOSS are. FLOSS is a very natural way to create software. If you have an organization of a bunch of people who are creating software, it is natural and helpful that they share just so more eyes see the bugs, and more minds solve bigger problems faster. FLOSS extends those principles to the world and FLOSS scales.

  48. oiaohm wrote, “Robert Pogson you are a little old school rebuilding the complete kernel to update 1 drivers. Yes with DRM drivers I know this is simpler since they those are not auto generated.”

    I was bringing all my clients to the same kernel, because I like the faster resume. It was a rare crash that brought the two ideas together, getting faster resume and replacing a buggy driver.

  49. oiaohm says:

    http://drvbp1.linux-foundation.org/~mcgrof/rel-html/backports/

    DrLoser by the way its worse. People reinstalling windows also run into Nvidia legacy hell. Nvidia drivers bad documentation on what it works on it also not new.

    The issue I missed is since Nvidia makes the generic drivers and the OEM are allowed to make custom versions you end up in horible where only X version generic works on X motherboard/card. This X version version generic is at least identical between Windows Linux…..

    Nvidia its way safer to stick to Gigabyte and ASUS brand cards and boards these makers don’t go into the custom driver thing so increasing you odds that legacy drivers will work a lot.

    To be truthful most of the driver compatibility updates done by the Linux kernel development are in fact done by automated scripts. This is including the auto-generation of new modules for older versions of the Linux kernel that project back-ports make.

    Robert Pogson you are a little old school rebuilding the complete kernel to update 1 drivers. Yes with DRM drivers I know this is simpler since they those are not auto generated.

    The core Linux kernel development avoids a lot of work.

  50. dougman says:

    Since when is Win-Dohs hassle free??

    Lets count the ways:
    Forced upgrades is a hassle
    Dealing with malware is a hassle.
    Continuous restarting and rebooting is a hassle.
    Windows 8 is a hassle.
    Activation for Windows is a hassle.
    Reinstalling Windows is a hassle.

    Feel free to add more to the Windows Hassle title.

    Eh.

  51. oiaohm says:

    Robert Pogson video cards do wear out. Something I have found with Nvidia is most of the time once you are using legacy the card is already not generating properly.

    DrLoser there is a set bugs in Nvidia closed source driver for Linux that is driving users nuts using particular set of legacy.

    In fact the bug starts at the kernel driver. 1 the driver file in kernel space no matter the series is called nvidia.ko and it has nothing open spec to pull what version it is after its loaded.

    Installing legacy closed nvidia drivers on Linux no matter how you install you can end up with a mix of newer and old resulting in complete failure as well.

    update-alternatives –config nvidia exists on debian to allow selecting between many nvidia drivers manually. Yes problem nvidia does not support this properly in all legacy drivers either.

    DrLoser be truthful if you under windows had to manually align the userspace with the kernel space so Nvidia worked you would not be happy when every other driver graphical driver under Linux you don’t have to because the system can access the kernel module version and use the right group.

    Final issue is the stack of options Nvidia demards in a kernel that they don’t document. Part of the reason why Nvidia drivers have been broken by kernel development so much is lack of documentation on what their module uses.

  52. DrLoser says:

    Let’s ignore that you missed my points, Robert.
    (Whilst we can leave them all to one side, they are as follows:
    1) Nobody ever claimed that “non-free” is better per say. That is an absurd allegation.
    2) 99% of the population would rather pay $50 for a no-hassle solution than stay up all night, with results that are not even guaranteed.
    3) Even given (1) and (2), you did a pretty rotten job by your own admission, didn’t you? How long have you been working with the Gnu Toolchain? Ten years? And you compile a kernel every week or so, for no doubt very valid reasons?
    4) One would have thought that you would have learned something over the course of those ten years. Use a mobile phone, man! No, scratch that, download the kernel. Or, if you have to, build it on a server and target it to a specific device.

    Not too difficult, thanks to the Miracle of the Four Freedoms.)

    Now, this beef you have against Nvidia?

    You signed up to FLOSS. You get to experience the results of this contract, good or bad. You are not capable (and neither am I) of sorting out an Nvidia driver problem, even given a totally open and FLOSS source code tree.

    And don’t you think that it contravenes the entire spirit of “Free and Open Source Software” to sit back on your fat ass and just hope that “the kernel boys and girls” will magically sort this stuff out for you?

    I’m in the same position as you are, Robert. I rely on “the kernel boys and girls” at Microsoft (or Apple, or Google) to sort this stuff out for me.

    The difference is, I don’t waste a weekend farting around.

    I rely on “hassle free” software.

  53. DrLoser wrote, “Nothing is better just because it’s “non-free,” Robert. Nobody has ever claimed that.”

    Then he/she wrote, “Let’s assume that Nvidia provided you, as a valued customer … this is a stretch, but let’s assume … with the full source tree of the relevant driver.
    You really wouldn’t have had a clue what to do with it, would you? At best, you’d have turned a fun-packed, educational all-nighter into a dreary fortnight of unpaid slog.”

    I build kernels and drivers almost every week. Why wouldn’t I have a clue what to do with it. The problem was they did not make their current driver backwards compatible with my PC a few years old. If they had issued a FLOSS driver it would have been in the kernel being updated by the kernel boys and girls and it would have been no hassle at all. As it was, the stuff they supplied would not build very well at all. I had pages of warning/error messages and stuff like that that I have not to deal with in the much larger kernel. It’s keeping their stuff to themselves that is the problem, nothing else. That’s the essence of non-Free.

  54. DrLoser says:

    Nothing is better just because it’s “non-free,” Robert. Nobody has ever claimed that.

    Many things are better because they are “no-hassle,” though. Leaving aside your usual apocrypha concerning M$ distributions on … what was it again … punched tape? … there are a couple of things I’d like to point out.

    1) It is indeed less hassle if you target a kernel build on a server. Perhaps, as an interesting experiment, you might try targeting it on a mobile phone next time? Small smart powerful thingies are the way of the future!
    2) It’s even easier if you let somebody else build the kernel for you. This is generally what people do, these days. I’m given to believe that practically 100% of M$ and Apple and indeed Android customers do this.
    3) Let’s assume that Nvidia provided you, as a valued customer … this is a stretch, but let’s assume … with the full source tree of the relevant driver.

    You really wouldn’t have had a clue what to do with it, would you? At best, you’d have turned a fun-packed, educational all-nighter into a dreary fortnight of unpaid slog.

    “No hassle,” Robert. No hassle. It is the way of the future. And indeed of the recent past, as in anything this millennium.

Leave a Reply