Linked by John Munsch on Mon 30th Dec 2002 19:05 UTC
Why do it? I am asked this question more often than I expected, even by existing Linux users who I expected to know as well as I the reasons for building a next-generation desktop Linux for the home user. So here are some of my reasons for thinking that we must spend the effort to create a better desktop on Linux than any existing version now has. Editor's Note: Due to a technical glitch, the first segment of this article was ommitted for some readers. If you missed the "why" section, before, you can read it now.
Permalink for comment
To read all comments associated with this story, please click here.
To read all comments associated with this story, please click here.
It sounds like the author wants BeOS. I don't have the time to go into this articles' contradictions in detail, so I'll just pick one -device drivers:
HW Manufacturers should be able to write a single device driver for all Linuxes (paraphrased)
This implies that the ABI (because you don't mean the API, surely; users don't want to have to compile the thing!) has to stay still for everyone - 2.2, 2.4, and 2.6, not to mention 2.8, 3.0, ad infinitum, because HW manufacturers come and go, with no change or improvement.
They have to cope with developing drivers for (or losing support for) DOS, Win3.1, WinNT, Win9x, WinME, WinXP, from the Microsoft stable alone. If they want to sell to MacOS users, they need to develop a MacOS 9 and MacOS X driver, at the minimum. If they want those remaining OS/2 users to buy their hardware, they have to write an OS/2 driver.
This is a choice on the part of the hardware manufacturer.
If they also want Linux support, you seem to be proposing that I go to www.my-hw-vendor.com/linux, click "install" and get a driver which will run on my Linux distribution.
Do you care to explain how this could be done, so that I could use, say, a USB Scanner, on my Intel laptop and my SPARC desktop? How exactly would that binary be composed?
Maybe the Linux developers should redesign all the architectures Linux supports, so that they will all accept a single binary format?
We already have one major vendor taking this approach - nVidia. I was *so* thankful when my nVidia card died, and I had an excuse to use my i815's onboard SVGA. With the nVidia binary driver, the machine kept crashing unexpectedly (hmm, similar to Windows, where drivers are the most likely culprit for a crash. Strange you're advocating a move towards the Windows model). Of course, I couldn't ask LKML for any support - neither they nor I know what is in that module. Now I'm back to Linux's own i815 driver, it works perfectly, but if for some reason it did not, I would be able to get support for it.
Linux (and this article is a great example of why phrases like GNU/Linux should be more common - Linux is the kernel, RedHat, Debian, etc, are Linux+GNU+Other Stuff but the author only sees the Linux buzzword) does not need to kowtow to hardware manufacturers for support. Linux is not sales-driven, HW manufacturers are sales-driven. Therefore it is the HW vendors who need to embrace Linux (if they want Linux users to buy their hardware), rather than Linux crippling itself for the convenience of HW vendors.
As I said, I don't have much time, but this is a long enough response to a single point mentioned in the article ... If you want, say, Lindows, to take this approach, then write to them, and convince them of your ideas.