Slashdot Mirror
Red Hat CEO Szulik on Linux Distro Consolidation
Rob writes "Red Hat's CEO has rejected the idea that a reduction in the number of Linux
distributions would be good for the industry, and described Novell's acquisition of SUSE
Linux as "theatre". There are over 300 distributions listed on DistroWatch.com,
but Raleigh, North Carolina-based Red Hat's CEO, Matthew Szulik, maintained that choice and
specialization outweighed any advantage that might be gained by focusing customer
attention on a smaller number of offerings. He was particularly disdainful of acquiring
other distributions for the sake of protecting or expanding market share. "We have
zero ambition to do that," he said. "I think when
people approach the problem with an eye on consolidation it destroys the idea of natural
selection.""
How does Novell, aquiring Suse, consist of theatre. They needed a distro on which to build their OES/NLD products, and since they seem to be partly in bed with IBM - who also uses Suse - that distro was the natural choice.
= Grow a brain...
When it comes to natural selection, is consolidation banned from the game?
I read
OS/2 did better in Europe than America as well. Your statements in no way invalidate Szulik's statements.
Well he can't very well call for "distribution consolidation" as that is a very Microsoft-ish thing to call for. There's certain things that even Big Linux can't call for without losing their Linux-cred.
It's like having to be hazed to get in a fraternity. No one really likes it, but you don't get in without it. I can just hear him squirming as his natural business executive instinct is to consolidate, but he's selling a product whose culture won't let him do it (yet). So for now, he smiles and yells, "Thank you, sir! May I have another (distribution)?!"
News at 11.
Seriously, does anybody expect Redhat's CEO to announce that "Novell is a serious contender, and Redhat is about to lose market advantage"?
"There can be only one"
Good idea, but even if you talked the distros into doing it, 10 people would fork it after each duel off, resulting in 3,000 distros.
-everphilski-
I thought that the features from each distro had to be shared back to the community by virtue of the GPL. Any modifications are available to everyone. So, all the features are available to everyone, but "good features" is a relative term. There is no single Linux distro that can be everything to everyone, hence many of them. I think 300 is too many, but there are different markets that are totally different. They are using Linux on cell phones, routers, desktop computers, laptops, servers, etc.
There are not only two versions of Windows. There is Tablet PC edition, Home Media Edition, Windows CE, etc. And, as for versions of just Windows XP, there are many different versions. Many companies create their own standard version that includes the utilities and features they want to include. Granted, they all come from MS, but they are customized. I'm not arguing in favor of Windows, but simple does not work when you need to span many different realms of consumer devices.
But like everything in life moderation is key.
Of course RedHat don't mind fragmentation it helps them. By encouraging fragmentation they can sit at the top and say to people "look, we offer stability". That's why Debian does so well (although I have to say I believe stable is a little to stable - 18 month update cycles please :) they offer some stability. It's important to try now ideas out but it's just as important that the OSS community tries to pull together.
While it is great that I can choose from 300 different distributions I have to ask the question: how many of them don't suck? About 5 to 10 would probably be the answer. I just want to cry when I look at the amount of time and effort that has gone into some of these projects that get maybe a hand full of users and then die a slow death as the idologues that started the project realize they aren't going to caputre the market.
It's great that people want to help it's just a shame there are a lot of people that feel the only wheel they can use is the one they built themselves.
I'm sure this post will get moded as a troll in two seconds flat so I am going to stop wasting my time.
I used to have a better sig but it broke.
Insightful? Just plain wrong, I'd say.
How many versions of Windows XP are there? Really just two, Home Edition and Professional Edition. How many versions of Linux 2.6 are there? According to that article, around 300.
You're comparing apples with pears. You should compare Windows with Linux distributions.
And since my posts always seem to get modded as trolls no matter what I say, if you Linux people don't get off your mighty high horse and look at what could get people to migrate from Windows to Linux, it will never happen.
Because you are trolling.
I don't care if it is Ubuntu, or Suze, or Red Hat, or whatever.
So you say, that you know only two - the third is called SuSE - of the mentioned 300 distributions? You just don't have to care about the other 298 distributions, they're made for special purposes. A few of them (Familiar) are made for PDAs, just like Windows CE (yes, another Windows).
Just have one damn version and make the damn thing work for the latest technology, make it fast, and make it easy to understand for even the dumbest american.
What do you mean when you say "latest technology"? There are more cases of Windows not supporting the latest technology.
I use Gentoo; how does this affect me?
"I think when people approach the problem with an eye on consolidation it destroys the idea of natural selection."
Corporate mergers, buyouts, and bancrupties are part of natrual selection. Consumers migrating to one company's offering can lead to 'natural selection'. One company having a big bank roll and buying out weaker competitors is also a form of selection.
In the 1930's there were hundreds of car companies. By the 1980's there were the big three and a few non-US companies. Over those 50 years a lot of 'natural selection' occured, and companies merging was just one option. General Motor's many brands of automobiles are not due to GM's internal innovation, but really are due to GM buying weaker competitors.
Let's watch to see what company will be the GM of Linux distros.
I want alternatives to Windows, but we need the serious distros to stand up and fight for an "expanded core" that doesn't comprise a constantly shifting codebase. Developers can't effectively write code for Linux because the subsequent support is a nightmare! Even RedHat's code changes too frequently for people to keep up.
I believe RedHat's on the right track, but the rest of the Linux world continues to evolve ... something needs to direct the evolutionary process. Both Apple and Microsoft spend TONS of time on UI, which is the specific direction that Linux needs. Obscure the complexity, unify support for drivers, and support software that people use and you'll find Linux in everyone's home.
No disrespect to Crux, whatever it may be, but it and Arch and Xandros and the rest aren't "divergence" in any significant sense that affects Red Hat.
Bonus inflammatory opinion: Debian is about to become the dog wagged by the Ubuntu tail. They're looking more and more like the pre-Linux GNU Project.
What I'm listening to now on Pandora...
No it wouldn't. It would slow, stumble, trip, but it would keep going. Red Hat's disappearance would be an enormous blow to the OSS community. It would take us years to recover. But OSS disappearing would destroy Red Hat entirely.
I am trolling
And how will you solve that by dropping RPM and using deb?
The ONE showstopper which makes impossible to make software installable between different distros is the per-distro "package namespace". In redhat X.org is called "xorg-foo", in debian it's called "xserver-xorg". No matter how good your packing system and how good your "dependency solver" is, if every distro names every package differently THINGS ARE NOT GOING TO WORK.
There's no point in redhat adopting deb. Fedora X.org package would not work in debian because fedora's x.org package "provides" xorg not xserver-xorg. Now apply this same logic to all the 15000 libraries in debian.
The one way to solve that compatibility problem is to make programmers to package things instead of distros. If every project would package things and tell distros how the package is named and set the dependencies (builds with libc x.y.z, optional feature depends on libfoobar, etc) AND all distros would use the work provided by the programmers instead of redoing everything, renaming the package etc. The format (deb, rpm) would be irrelevant
Completely impossible. For example, the ingredients that make a great rescue distro (like Damn Small Linux -- live distro w/ gui and important apps: 50 mb) are not necessarily the ones that make a great desktop system. Besides, why should there be only one distro when people have different preferences? What is wrong with choice? Sure it might be more difficult to chose between steak, crab, ho-hos, or shrimp, but would you rather your choice be "hotdogs"? Period. Be happy with it cause that's all there is.
What changed under Obama? Nothing Good
Red Hat may be right that we will not see a massive fold in of distros, but we likely are going to see more and more distros building on Ubuntu instead of Debian, because Debian is just dog slow. Ubuntu is exciting.
I run Ubuntu on my laptop, and my desktop will switch from Sarge/Sid to Ubuntu too, at the next reinstall (a reinstall is the best way for me to get rid of old cruft).
But for a server installation, I'd prefer the "dog-slow", conservative, well-tested standard Debian distros over Ubuntu. "Exciting" just isn't part of the vocabulary I'd like to use when my boss or customers comes up with the question "How would you describe the stability of the system you are about to install?".
What Red Hat ought to worry about isn't SUSE, but rather somebody who comes along, takes Ubuntu, tags on support, precompiles it with plugins, extensions, some valuie-added management tools, and takes it to the server. It WILL happen, it's just a matter of whether they have the means and polish to make it worthwhile.
Since Ubuntu is Debian-based, and Debian usually goes through rigorous (hence "dog-slow") testing before release, wouldn't this long work be sacrificed if a distro based upon it switches some of the packages for newer ones? And if the (server-) relevant packages are left unchanged, what would be the difference between the two distros, and the incentive to switch from original Debian on the server side?
As I said, you're right. And no one else would step in to maintain or develop it. Oh poor GCC, the orphan child of OSS, no one but RedHat could do it justice. Oh what ever did GNU/Linux do before RedHat?
[sarcasm off]
--- Grow a pair, liberals... stop letting the Republicans bully you!
That aside, of course, Red Hat would hope that the number of non-Red Hat distros would stay high, since that tends to increase the gap between Red Hat, the only Linux distro that most ITers know about, and the rest of the pack. In addition, the confusion, or perception of confusion, drives corporate Linux users to Red Hat.
Disparaging Novel/SuSE is also to be expected from Red Hat, since SuSE was and is the only competitor to Red Hat.
(My own opinion is that the proliferation of distros is a serious problem that wastes effort by Linux distro developers, complicates support, makes life difficult for application developers, and gives many potential users, both corporate and consumer, the impression that Linux is immature. If whatever-we-mean-by-Linux were a complete system, like FreeBSD, we wouldn't need packaging to be a separate operation, and the number of distro outfits would be very small.)
Several of the responses to this article and every other claim for consolidation revolve around "linux is about choice." I agree, choice is an important mainstay in the linux mindframe. I want to be able to choose how my UI looks, how my mouse behaves, what web browser I use, what permissions I have, etc. But a simple assesment of the current Linux situation is uncomfortably jumbled with too many distributions striving to achieve the same thing but through different means. If just half of the distributions had decided years ago to work on ONE installer, where would we be? If they had decided to work on ONE set of configuration utilities, where would we be? The end user wants choices when it comes to how the system interactively reacts doing every day tasks. But when it comes to system maintenance, software installation, package managements, etc, choice is not productive. When a support team has to know 15 different ways to change the default IP for a wireless network card based on the plethora of distributions, that is not productive. When a package has to be compiled and released in 4 package formats with 10 different sets of libraries to support the majority of linux distributions, that is not productive. I love choice, I love that there are 10+ GUI's for me to switch between. I lvoe that I don't have to pay another company to change my theme in those window managers. But I also love that fact that there is really only one X-windows. But what I don't love is that a dos to unix conversion utility is called "dos2unix" on redhat/suse/mandrake and "flip" on ubuntu. There are several more examples and I'll leave is as an exercise to the commenters to flame me for my criticisms and critiques of the state of GNU/Linux. To summarize, choice is wonderful, but so many of the problems and complaints that linux users have could have been effectively solved by now if the 200+ developers working on 40+ projects decided to work together instead of trying to invent forty versions of the wheel.
No. There is a huge amount of examples where parts of, say, the human body are not designed intelligently at all, but kinda work and thus have been retained in the course of evolution.
Often quoted are the blind spot of the human eye (which is not present in the octopus eye, although otherwise the 2 versions are very similar), or the fact that the birth canal runs right through the only bone ring in the human body that can not expand.
Other examples revolve around the fact that the human is bipedal, and many things would have to be designed differently if this constraint would have been known beforehand. E.g., humans are the only known species to suffer from hemorrhoids, which comes from the fact that the sphincter is located in the same direction as gravity pull.
The spine would have to be designed completely differently for a bipedal - the frequent damage to intervertebral discs in humans also comes from the fact that a quadruped design has been adapted to bipedal. Huge parts of the curriculum when learning martial arts are devoted to teach how to handle the limitations of the bipedal spine, and its fixed conjuction to the pelvis, under the influence of gravity (esp. in Tai Chi Chuan).
"When I first heard Daydream Nation it quite frankly scared the living shit out of me." -- Matthew Stearns
It's not just the programmers: Debian has a very strange way of naming source tarballs with "_" in the names that doesn't actually reflect the content of the tarball. SuSE stuffs the Wacom kernel modules and configurarion tools in the xorg packages, for no good reason whatsoever. RedHat's actually been pretty good about naming packages what the author wanted it to be called, but when you look at the lack of conventions in the Perl crowd for the naming of source packages compared to the modules that are actually in the package, you find an endless source of confusion and conflicts.
.spec file and RPM structure, but then you have SuSE go and commit absolute absurdities in it, like adding tarballs of patches instead of the individual patch files and selecting which patches to apply in a script in the source files, rather than in the .SPEC file itself, and you have absurdity piled on absurdity.
/usr/sbin or /etc like the rest of us.
The Linux kernel authors are doing a good job of creating what you describe, a basic
Unfortunately, I've worked with open source authors on other projects who absolutely refused to accomodate the most basic public standards, such as Dan Bernstein's refusing to include the documentation for qmail or his other tools in any easily mirrored or downloaded way but only maintained intermingled with other tools on his website, and insisting on his *OWN* hard-coded top-level directories for things instead of using
Um, yes yum works, but I would never agree with someone saying it works well.
It is very slow and hogs memory like only beta versions of mozilla. It sets its pace at the time it takes to check to see if any of the repositories have changed, even if it checked 30 seconds ago.
A "yum search" not only hits the network, but it takes over 50MB of ram to do that.
Yum is only tolerable when called from cron IMO.
There: Something at a specific location.
Their: Owned by someone.
Please make sure your english compiles.
...that if there were fewer distributions which development energy were focused on that greater strides could be made in the technology. While I think that the number of distributions currently "in the wild" isn't completely ridiculous, it seems to be heading rapidly in that direction. If we could, for instance, gather all of developers of the "networking utility distributions" together and let them focus all of their efforts on one single "product", we would have the best features from the best distributions. Then again, choice is good.
> If linux ever hopes to gain a fat chunk of the market share
Linux doesn't hope anything. Linux is kernel; a bundle of software. It lacks hopes, dreams, fears. Moreover, Linux is not developed by a single company with a (supposedly) single purpose. It is developed by a community of independent developers, eash with their own goals, hopes, dreams and fears. "Gaining a fat chunk of market share" is way down on the list of interests of many (probably most) Linux developers. Making a good, reliable, flexible system that does what it's supposed to and is free to be used for any purpose is the main priority. If people want to use it--great. If not--their loss.
The purpose of Linux is to fill a need, not to be a need! Anyone can make a distribution if they have the time, talent, and the need to fill a need. Linux is a tool, not a product. Moreover, it's a general purpose tool that can be adapted for many needs. If flexibility and freedom scare some people away, well, guess what? They're not the intended market! If some people feel the need to challenge MS dominance, well, fine, but that's a need, not the need, and they're just going to have to figure out how to do it while others continue to work on their own needs which are often better filled by creating specialized distibutions. Trying to create "one-size-fits-all" distributions is just as bad, because then you have 10 million choices inside the distribution. Unless you can persuade everyone in the world to need exactly the same thing (or to put up with tools that are a bad fit for their needs), you're either going to have to have many distributions, or many confusing choices within a distribution, or both. There is no way to "dumb it down" and keep the power and flexibility that is currently available.
Szulik seems to get this, and more power to him! Trying to challenge MS head-on is a mug's game. Ignoring MS and concentrating on making good systems that fill people's needs is the only way Linux will ever get anywhere. And guess what? That's pretty much what's happening.
Hmmn, this really comes over as a senior suit dissing the competition and engaging in a little preening. Lunch with Shadowman would surely have been more entertaining though it would probably be a couple of days before the hangover subsided.
Novell/SUSE have an increasingly strong product and it's very, very far from "theater". And besides, the ultimo, leading Linux distro may not even have been launched yet. A major corporation could enter the Linux world tomorrow with a brand-new distro and turn the entire place upside down.
I guess Red Hat had better keep running because there could be some really hungry bears after them.
Las qué passoun
tournoun pas maï