“There are plenty Linux reviews, though almost none of them are about Linux. You see, Linux is only a kernel, but most Linux reviews are about KDE, SuSE, the desktop and so on. Those reviews aren’t about Linux, strictly speaking. In contrary, this review is about Linux, made for the people who normally don’t configure and compile their own kernel.”
Maybe just my own personal zealotry but thinks it could be a bit improved by leaving out the subtle digs and comparisons to another OS’s lack of features and fragilities but rather to let linux stand on it’s own merits and features and my world improved to be able to read an article on linux without mention of it. Just a thought.
What you said made little sense.
It made sense to me. What he’s saying is that he thinks the article should concentrate solely on Linux and its merits, and not make stabs at Windows. (Which the article does do, and inaccurately at that, since in my experience Windows has very good out of the box hardware support)
The bag of drivers take care of the hardware support. Not windows.
Yeah, and windows has a nice bag of drivers in part becuase it has a stable driver API. You’ll be lucky to get a driver from 2.6.11 working in 2.6.13. There’s no stable ABI. Constants change, function names change, everything changes! Even if you had the source code for the driver, you’ll most likely still have to tear your hear out making it compile with a later kernel version.
But he does have a point, Windows doesn’t have hardware support “out of the box” like the Linux kernel has. No nForce/AGP/Sound/ethernet, this is true for alot of chipsets including VIA, nor does it have full speed USB2 suport.
In the UK we have NTL ISP and there modems both ethernet and USB work out of the box with the Linux kernel, it just works.
In the UK we have NTL ISP and there modems both ethernet and USB work out of the box with the Linux kernel, it just works.
Pardon, and if you hooked up that to the Windows box using enternet, it would be merely a networked machine accessing a router; no drivers required.
Pardon, and if you hooked up that to the Windows box using enternet, it would be merely a networked machine accessing a router; no drivers required.
Technically speaking, drivers are required. Both are functionally equivalent in this respect
Technically speaking, drivers are required. Both are functionally equivalent in this respect
True, all the USB function is, is ethernet over USB, which is in the linux kernel, and the other is straight enthernet; having heard the reliability horror stories with USB, I’d go for straight ethernet anyday.
I can see what you mean, why state the obvious
–
But really, I agree, either write the article to compare features between many OSes (OS X (Mach/BSD/Apple), Linux, Windows and Solaris might be a good mix), maybe throw in some benchmarks (now that you can run all those on the same hardware), or make is about the features of one OS. I personally not bothered by this article since I want Linux to have my babies, but I can definably see how it might bother some people.
Just trying to help with a little honest criticism. My idea of a good linux article is one that helps me forget windows exists. b-)
make gconfig for a nice GTK UI for the kernel config, you can use make xconfig for QT/KDE. Just to add the foredeth onboard ethernet driver(nForce chipsets) is not experimental anymore in 2.6.17-rc1.
I do think he talked about other OS’s to much rather than consentrating on the Linux kernel features.
I like gconfig too.
Every time I start I think there should be a label to generate a minimal modules set checked .config for the current hardware.
No offense, but, I doubt GNU/Linux/Whatever had these drivers when Windows XP was first released.
I think the fact that the kernel that is Linux is updated more frequently helps in this area.
Granted, I found it hard to concentrate on the article (probably not its fault, im just a bit exhausted)…
Where’s the innovation? He talks about it at the beginning and the title, but I don’t see what specifically he considers innovation?
Or is he going by the Microsoft definition of innovation..?
Where’s the innovation? He talks about it at the beginning and the title, but I don’t see what specifically he considers innovation?
I don’t think he can talk about it like that. This is a kernel, so innovations in it are highly technical things.
He talks about choice of load management though, and talks about udev, which is not in the kernel (as udev is userspace device management), but uses lots of Linux kernel features and innovations, the powerful dm layer, the fact that you can configure every aspect of your kernel (like compile for embedded, remove things like filesystem handling, …), NPTL/TLS, schedulers, … He can’t talk about the schedulers for example (memory and I/O).
The schedulers are some powerful things on Linux. I think you can’t appreciate them as long as you didn’t try a process eating all the CPUs on Windows : the PC is so slow it becomes unusable, if you have a CD/DVD burning going, you’d be pretty sure of losing the CD/DVD (even with burnproof).
That’s the kind of load I was putting on Windows at home, and it would not last long.
On Linux, I have to launch a gkrellm process to know if a process is eating all my CPU (like with some compilations), as I can’t notice it most of the time.
Recently, I updated my glibc live (never do that, unless you know the consequences), and a daemon was eating all the CPU for 20+ hours before I noticed it.
And I burned numerous CD during that time.
Some innovative schedulers are tested in the Linux kernel (stair-case), a lot of innovative features are tested too, and a lot are scraped for different reasons.
Honestly, that sounds like Microsoft’s definition of innovation. Though, I think they have 2.
I’m still not sure where the innovation is, as you were kinda vague (not a knock, I appreciate the info you did give).
As far as Windows and cd/dvd burning. I’ve never had a load/lag/whatever problem when burning a cd or dvd. Now, watching an HD movie, browsing and trying to burn something, that’s a different story. I think that has a lot more to do with the HD that anything, and I don’t think HD movie playback is great on any OS yet.
I’m sure there are some innovative things at the lower-level in the kernel. But I think that can be said for all the major operating systems. Except Microsoft doesn’t get recognition for theirs, since it’s usually just the internal people that know. Or people that watch channel9 maybe.
Honestly I thought that was a great intro…I would say more than a review into Linux…this would be something great for a person like me for example who has been used to using XP all the time and wants to install a Linux distro and learn more about how to get the most out of it. One of the rare few articles that are really really good.
“I found it stunning, Windows XP without any service packs doesn’t offer USB 2.0 support.”
XP came out in 2002, and he’s comparing it to a modern Linux kernel as far as supported features? Ok…
and people still point out “flaws” in linux based distros that was fixed long before XP was released. the blame goes both ways…
Its only fair to compare the latest stable release of the Linux kernel to the latest stable release of the Windows kernel. It doesn’t matter what they have in development, its vaporware until it is released. The day Vista is released, it will be fair to compare it to the latest stable release of the Linux kernel, which may very well be 4 months out of date at that point.
True, but just because Windows’ distribution method is less agile than an Open Source OS’s distribution method doesn’t mean we should overlook the fact that one method provides better support than another (out of the box).
A similar argument can be made that it’s harder to make a profit from Open Source than closed source. This corelates closely with the distribution methods I mentioned above.
Bottom lone, no one way of doing things is best in every way.
True, but just because Windows’ distribution method is less agile than an Open Source OS’s distribution method doesn’t mean we should overlook the fact that one method provides better support than another (out of the box)
No method provides better support than the other. The distribution methods are not what provides support.
Human beings provides support, some of them helped, some of them not helped.
You could have perfect support with the Linux method of distribution if it was not for political problems.
Just look at what commercial distros are doing.
when Linux drivers support is so bad. At least with Windows, once you download a binary driver, it is almost certainly guaranteed to work for the life cycle of the product (7 years.) With Linux, unless it comes with the kernel, be prepared to constantly update your drivers everytime you update your kernel. How many times have you downloaded those Windows XP nvidia drivers?
it realy depends. they keep releasing new versions
and often the only thing you need to do with the linux driver is recompile the kernel interface, not redownload the whole thing again.
still, its not a perfect solution, i give you that. it would be nice if someone could come up with a way for the drivers to have their kernel interface be recompiled transparently when a new kernel got installed.
like say one could agree on a default place to put the makefiles and other stuff so that distros could have a script in the package go over and compile them when needed. and if failed, tell the user to go download a new version of the drivers that had problems.
“still, its not a perfect solution, i give you that. it would be nice if someone could come up with a way for the drivers to have their kernel interface be recompiled transparently when a new kernel got installed.”
Debian does this with LIRC if I recall correctly, when there is a kernel update, the LIRC package will re-compile the modules for it. Though I saw it do this one time and it failed since I hadn’t installed the proper linux-headers package as well. But the capability is already there with Debian via the module-assistant program.
Leech
but then debian is often a no-go zone for non-free drivers…
what we need, is the same kinda system but for rpm-powerd distros…
It would be more interesting to see a comparison of innovated features to another Unix like operating system than that *Other* OS.
There is no need to configure or build kernel as root. Only need root to install.
This article is really a brief tour of the main components that go to make up a Linux kernel, and how to configure and compile it. OK, to thank extent it is very useful to have something demystified and explained a little.
Just my 2 cents, but I thought he could have widened the article’s scope a little. For example, what are the main differences between, say, Linux, BSD and Solaris kernels? How do you patch a kernel? Is it a good or a bad thing that the Linux kernel lacks a stable api for closed/proprietory drivers to bind to?
Another thing that can make quite a difference are your compiler settings. Are they optimized for your processor rather than generic default settings? If so, recompiling the kernel and modules can deliver a nice speed bump.
Oh well, big subject. Distros seem to have their own patches and settings for the kernel, and their own “recipes” for compiling it, so it is quite hard to write how-to stuff about it in general terms. Kudos to the author for making a stab at it, at least.
If you haven’t delved into the details of how the thing actually boots, you haven’t told us much.
The reason for the low amount of reviews towards linux kernel is because it’s mature enough not to review it, whereas the GUI interfaces and distros lacks alot of wished features that users want namely (performance and stability).