![](https://seccdn.libravatar.org/avatar/6d198f8c8f1c94ccef873cebcf4f5dfa.jpg?s=120&d=mm&r=g)
Kevanf1 wrote:
On 5/10/05, Sid Boyce <sboyce@blueyonder.co.uk> wrote:
I very much doubt you'd get the various distros to agree a standard way of doing things, UnitedLinux was the one hope, but now it's every distro for itself, trying to relive the Unix experience for sure -- lemmings are destined to forever walk off cliffs and die. The kernel is the one strand that's stopping a proper set of forks. What we need is a set of interoperability checks and certification that a standard application set will run across all distros and a distro is not shipped until it's certified. Some weeks ago I downloaded a RedHat/Fedora source rpm which built and installed fine on SuSE 9.2, it just didn't run. You also saw my example of trying to install the Mandriva version of cooledit on SuSE 9.3. Look at http://www.skype.com/products/skype/linux/ to see the nonsense that Skype has to go through for each distro, at least a number of distros had the good sense to base themselves on Debian, possibly one reason why Munich went Debian after SuSE/IBM had done all the hard work. Regards Sid. --
It must also have the effect of stopping hardware manufacturers from producing drivers for Linux. This is something that I hadn't really thought of before but it makes a bit of sense. But, just to contradict myself the same manufacturers have had to produce very different drivers for the various flavours of M$ software over the years. Some have drivers for 9* series, Win2k and XP. I'm sure Linux drivers would not have to be that much different. Or am I wrong in this assumption?
The drivers problem is different as they are an adjunct to the kernel. The real problem there is that the manufacturers are afraid they would be letting too many secrets out of the bag so that a competitor could use their code to build a similar product lots cheaper and easier, they do the work and someone else uses it to put them out of business. Companies like NVidia have opted for a half way house that delivers useful drivers for free use with Linux and at the same time protecting their hardware development secrets, to me, an acceptible compromise. Some others are not prepared to go that far and would like to contribute binary only drivers a-la-Windows which break with at times just a slight change in the kernel. Binary-only drivers are also disliked becuase the kernel developers are unable to troubleshoot any possible problems that those drivers may introduce or kernel problems that are exposed by such drivers. When manufacturers have been willing to disclose their driver secrets, the kernel developers are able to make changes with them in mind so that things continue to work and the kernel is not "tainted", if you submit a kernel problem with say, NVidia video driver module active, the kernel reports "tainted", so they won't look at it. What I do in these cases is to use "nv" instead of "nvidia" in /etc/X11/xorg.conf and send them the untainted report. Overall, we are not too badly off with drivers as thankfully the range of hardware to choose from sees to that, also many manufacturers are using essentially the same hardware in their products and the problem seems to be of less significance now. Just last week I came across an ORIGO 10/100 Fast Ethernet card I had around for years that I never got going when I bought it, the chip says Davicom, now I popped it into the Mandriva box and it's recognised, driver loaded and configured. When I bought it, DEC Tulip cards were common and recognised, but this one wasn't. 00:09.0 Ethernet controller: Davicom Semiconductor, Inc. 21x4x DEC-Tulip compatible 10/100 Ethernet (rev 40) Regards Sid. -- Sid Boyce ... Hamradio License G3VBV, Keen licensed Private Pilot Retired IBM Mainframes and Sun Servers Tech Support Specialist Microsoft Windows Free Zone - Linux for all Computing Tasks