Ubuntu's need to catch a wave

Submitted by dag on Sat, 2008/05/17 - 23:38

Let me play devil's advocate here. Mark Shuttleworth's recent pledge to join a synchronised release plan for Enterprise Linux distributions is no more than a wish to benefit from a lot of work that Novell and Red Hat are already doing in the Enterprise space.

Let me explain.

Red Hat's Enterprise Linux offering is a very important proposition to businesses, not only because Red Hat guarantees (and has proven) to support each RHEL version for 7 years after general availability, but mostly because each version is updated with new hardware support every 6 months.

Why is that important ?

When a version comes out all components that comprise a distribution have been frozen half a year before and have been tested very hard. But hardware does not freeze and companies buy new hardware constantly. Red Hat needs to provide support for newer (selected) hardware from vendors to make its solution acceptable to customers. This work is not taken lightly. Red Hat has to backport fixes and backport kernel infrastructure, update drivers, perform regression tests and provide QA together with hardware vendors.

Every 6 months a new update release is being made to ensure that newer hardware is ready to be deployed with a recent RHEL version. Red Hat guarantees that new boot media is available with newer hardware support for another 3 years. That means 2 RHEL releases at every point in time are being maintained to support new hardware.

This picture explains Red Hat's Enterprise Linux offering best.

Click the picture to see it in all its glory

So what are you getting at ?

The sheer manpower to do this, together with new development and bugfixing for 4 different RHEL versions is something that Canonical/Ubuntu simply cannot take upon it. Even Novell does not support that many released versions in the way Red Hat does.

So ?

So Mark's article is wishful thinking and hoping to ride the wave that Red Hat (and Novell) are funding. If he can use that same kernel, with the same backports, fixes and regressions tests, Ubuntu LTS does not need to do anything to support the same vendor hardware. Easy, but at the expense of both Novell and Red Hat.

That explains why Mark wants 2 out of 3 Enterprise Linux distributions to enter his circle. I could imagine Novell and Mandrake joining forces to align the release cycle to try and take on Red Hat's 80% Enterprise Linux market share (sales, not necessarily install-base). There is hardly any benefit in Mark's proposal for Red Hat.

Also, Mark's Ubuntu timeline also resembles CentOS's timeline. And I guess he must have been studying Red Hat's release cycle very hard to come to his conclusion. (CentOS' timeline obviously comes from Red Hat's release dates)

Let me add that CentOS is not really involved in this discussion as CentOS does not directly compete with Red Hat (RHEL) or Canonical (Ubuntu LTS) in this market segment. Both offer paid-for services and support, while CentOS is provided as-is.

Update: I updated the CentOS timeline to include the upcoming RHEL6/CentOS-6 release.

I think the problem with

I think the problem with Ubuntu is that it is trying to tie in its enterprise offerings with regular releases. This is completely different from the Fedora/RHEL combination.

Maybe Canonical, with all its money, should just sponsor CentOS. That way we get a free enterprise operating system near the release schedule of RHEL. ;-)

Maybe Ubuntu, with all its focus on the desktop, should create easy-to-use-yet-complete-enough desktop tools for server applications such as Apache, FTP, Subversion, etc. Then we may have the good and stable backend of RHEL, without the horrors of system-config-*. system-config-* sucks by the way. :-)

Dag -- I couldn't agree with

Dag -- I couldn't agree with you more. Red Hat provides an excellent production-class environment, and it's no wonder that so many enterprises use RHEL as opposed to Ubuntu or Debian for their mission-critical environments. Don't get me wrong, Ubtuntu and Debian provide a GREAT desktop release, and I love their release and support their work 110%.

Unrelated, I love your RHEL/CentOS/Fedora packages! Thanks so much for your contributions to the community!


Looks like you've been slashdotted.

I'm interested to see how the conversation proceeds. :)


Now, if the OpenSolaris gang were to join forces with Ubuntu, that would make life interesting.

"Now, if the OpenSolaris gang

"Now, if the OpenSolaris gang were to join forces with Ubuntu, that would make life interesting."

Sure. The OpenSolaris gang only needs to drop the Solaris kernel and the other Sun software, and they'll be welcome.


share and share alike, I think it is good for everyone if we cooperate, I may be wrong but I don't think shuttleworth is trying to 'ride the wave' but share the work. It seems silly to carry out several separate trains of development on an essentially open and equal system.

Ubuntu does not share the

Ubuntu does not share the work even with it's dekstop oriented product.
Just look at
As you can see there is nothing to share with Ubuntu because Kanonicals (Ubuntu producer) is not doing anything serious. They just use work of other companies - mainly of RedHat and Novell.

So you are totaly wrong and Shuttleworth is really trying to use work of other companies for his own needs. Period.

Looking only at _kernel_

Looking only at _kernel_ stats is not very helpfull.

But I think you're still right.


Is that graphic above a slide from a presentation? If so, do you have a link to the whole presentation? Thanks!

Yes, my presentation was

Yes, my presentation was linked from the article as well. Here is the link again:


Although I am revising the information and the comparison slide contain rough numbers and (now) incomplete information. For one, I have to revise the Ubuntu LTS information.

Feedback welcome.


Very nice presentation (and blog post). At least I finally know what is CentOS about.

And I love that Gentoo exclusion from Enterprise system :D

Very nice presentation!


Very nice presentation! Great way of positioning CentOS linux specifically, and Enterprise Linux in general.

(BTW, I never said it earlier, but I agree with your comments on the above 100%)

Couple of things

First time I've seen this blog, very good. Think you're running Drupal 6 too, which is quite impressive (since many modules haven't been ported yet).

Secondly: love the Leonard Cohen reference in the header. Am I the first to notice that?

Drupal adventures

I blogged about some of my Drupal adventures:

but I have to say I'm quite the novice. The reason why the transition to Drupal 6 was so easy was because this blog is not using that much extra functionality from modules. And because some of the things I enabled I redid using a different solution.

And yes, you are the first to comment on the Leonard Cohen reference :-) Although I gave that away recently after using it on my website since almost 1998. I am a big Leonard Cohen fan.

Had a look at all those posts

Had a look at all those posts you linked to, excellent stuff. I notice some important Drupal people have been commenting on this blog too.

I assure you I guessed the Cohen reference without seeing that post. :-)

Er, not so sure...

Firstly, I've learnt something about what Red Hat does thanks to your article, and it sounds very nice.

However, if all Shuttleworth wants is Red Hats code, then why hasn't he aligned Ubuntu to it already???

There are so many markets currently where the market share of linux is miniscule. In such a market, any thing done to increase share by any _linux_ benefits all distributions. Therefore the question should be: "Will doing this significantly increase the market share of linux?"

If the answer is yes, then do it. End of story.

Good question

Good question. I can imagine that my opinion, if true, is not something Shuttleworth would want to say in public. Because it could easily be seen as Ubuntu being a cheapskate.

Furthermore the Ubuntu community would probably not follow that logic since that would mean they agree Red Hat is doing much better than Ubuntu LTS in that regard. Most people in the Ubuntu community are not large enterprise sysadmins.

If what I say is what is going on, Shuttleworth's article is a first cautious notice to everyone that they want more of the Enterprise Linx pie. I am sure that Canonical needs a bigger share to pay their bills without relying on Shuttleworths money.

Enlarge the pie

My point was that making the pie bigger will massively help RHEL!

You need to get out of the 'static market' mindset. Yes Red Hat is doing some great work, but worrying about Ubuntu taking it is foolish.

By the time Ubuntu have enterprise sales volume comparable to the Red Hat figures you have mentioned elsewhere, the linux enterprise market will be many times the size it is now.

The result of this is that Red Hat will get a large increase in sales volume, although their percentage share of the _linux_ 'pie' will probably go down.

That the pie is getting

That the pie is getting bigger is irrelevant to the subject of the opinion piece.

Whether Canonical at the moment is actually helping to make the pie bigger (which is what you are trying to say) is unsure, at least if the pie is the (paying) Enterprise Linux market and not just the install base in itself.

And I doubt that a better orchestration of releases will help much in that regard. The only benefit I can see is that Canonical can just cherry-pick the patches they want with less effort. But the recent openssl debacle shows that that is not sufficient to create an Enterprise class operating system. At least if the patches are coming from Red Hat, they would have had good QA :-)

RHEL and FC have fallen by

RHEL and FC have fallen by the wayside, along with all other unices. Ubuntu has the most development muscle and the most mindshare. By far. Shuttleworth will continue to play attach-and-extend with the other distros, getting them to work for him. Ubuntu is by far the dominant distro and continues to gain momentum.

The other distros have been condescending and scornful of Joe User, the guy who is interested in this linux thing is afraid of the command line. By catering to that demographic, Shuttleworth has developed a very large and devoted userbase. Ubuntu is not about to replace RHEL on the corporate server, not today or even in the next two years, but it's carving quite merrily into the desktop share previously owned by XP.

The current gasp of buzz about RHEL is a dead cat bounce. The distro is painfully outdated and awkward, and has already begun fading into obscurity.

Ubuntu is doing fine in the

Ubuntu is doing fine in the desktop market. That is not what this article is about. But Ubuntu is definitely not the biggest contributor to Open Source development. A big userbase has not been turned into a big development force (yet).

It is also unknown how big the Ubuntu's userbase is. And only Canonical can give figures of how many Ubuntu LTS users are paying for support.

I can give you some rough numbers for RHEL and CentOS:

* RHEL has about 1.5 million entitlements (paid-for supported systems) apparently comprising 80% of Enterprise Linux sales

* CentOS has a lower mark of 2.2 million systems (pulling updates directly) and an estimated minimum of 8.8 million servers worldwide (probably a lot more given companies do not pull updates from public mirrors) comprising of 0% of Enterprise Linux sales

What I am interested in is how much Ubuntu LTS and Ubuntu systems are out there. And how much actually have a Canonical support contract.

Why ? Because backporting kernel fixes, drivers and kernel infrastructure is not something volunteers prefer to do and doing it every 6 months for several versions costs money. Money that Canonical does not want to spend, if they do not have to. Canonical/Ubuntu focuses mostly on the integration work and the desktop aspects.

ubuntu is trivial

...in the larger scheme of things. centos runs the web hosting industry, red hat runs gov't and wall st. novell gets a few european firms due to the old suse network, but it's dying.

Fedora is annoying because its yum network is slow, and its packages change so often you pretty much have to reformat every 9-12 mos.

Ubuntu is great as a desktop distribution but that apparently isn't a profitable business. Trying to get into the server market is like tunneling through the floor of the bank after-hours. It might work, or you might have 4 inches of steel to cut through first.

Whats in ubuntu anyway

i'm with you on this one. I'm still not sure why every loves ubuntu. I wrote something about it:
Their marketing engine is pretty dirty too, first use of Novell's deal with ms to try and get them more dev's, now this. And the community sometimes fall for it too. tsk tsk Mr Shuttleworth, my compatriot.

"i'm with you on this one.

"i'm with you on this one. I'm still not sure why every loves ubuntu."

Because it works quite well, is easy to use, and provides no-hassles on-line updates.

I just got a laptop with SuSE pre-installed, and it simply doesn't compare.

Use the source

Hi Dag,

So Mark's article is wishful thinking and hoping to ride the wave that Red Hat (and Novell) are funding.

I don't really buy this ;). If he wanted to ride the RHEL wave, he could align LTS releases with RHEL releases. And since all sources are available from Red Hat's FTP server, one could freely ride the RHEL wave, including driver and feature backports.

Aligning releases benefits everyone:

  • Releases of major components can be oriented towards these enterprise releases.
  • Bug reports and patches can be shared.
  • The systems would become more compatible. This is good for software vendors, and users (makes community support easier).

Maybe I am naive about the motives involved, but for the whole community, I think this would be a good thing.

What about Debian

I constantly hear other RHEL/Debian sysadmins complain about the time it takes RedHat to plug a leak in the kernel, while the Debian patch has already been released.
If that is the case, that would mean that they are only using RHEL for its extra drivers...

From what I have noticed, a lot of said drivers are published by the vendor of the hardware, so technically they can be copied from the vendor into Ubuntu. (Like they do with ATi and nVidia)

That is the first I hear

That is the first I hear about it. Do you have numbers or references for this ?

BTW my experience with large companies is that they try to stay away from 3rd party drivers. Sometimes for support reasons this is (temporarily) impossible until the next update release, but in general we are not running proprietary kernel extensions for a very good reason.

We want to have predictable behaviour out of machines, behaviour that matches with what all other RHEL/CentOS users are seeing so that our technical problems are not unique problems and we have a better supported environment. 3rd party drivers bring quite the opposite.

must be the cowboys at redhat!

you know,

the ones who modified the openssh code to squeeze it into the redhat distribution.

so it wouldnt pester for passwords in public keys?

oh wait... that was debian!

Using the same kernel would make sense

Canonical has all the right in the world to use Red Hats work. And I think they should. It is a fundamental strength of Open Source that you can just take the work of others.

You have to free yourself from the notion of "stealing" work others did or paid for.

I think Mark has a strong sense where things are heading and as Linux matures there will be more colaboration.
I see a future where Novell, Canonical, Oracle and Red Hat ship the same Kernel, gcc etc. in their Enterprise offerings.

Thing is as long as Red Hat will do most of the heavy lifting they will make most of the decisions and really conserative customers will still buy Red Hat.

It might hurt Red Hats bottom line but it will be good for Linux (-User) in gerneral. ( Although it might just make the whole market for all Linux vendors bigger .. thus still adding to value to Red Hat )

With Open Source there can be no Microsoft making 10+ billion US$ in profits ( like 50 Euro :P ) each year with software alone.
Just get used to it.

Of course does

Of course does Canonical/Ubuntu have the right to use Red Hat's work. You are coining the word stealing (and maybe Slashdot's headline is), I never said they were stealing. And being part of the CentOS community I can defend that what CentOS is doing is not stealing either.

But the problem is that using Red Hat's patches is a lot of work unless Ubuntu LTS is using the exact same software versions. Especially for the kernel, and that is what my opinion piece is all about.

However I do not agree that synchronising the distribution releases is going to benefit the community at large. Like what I wrote I think it mostly benefits Ubuntu LTS and Canonical. Not existing Enterprise Linux users, nor the community.

Kernel sync

OK, good to know.
Have you read Marks latest post ( http://www.markshuttleworth.com/archives/150 ).

I think there will be more streamlining between EntDistros. As far as I know they said so with the anouncement of the driver backporting stuff at the Linux fountation collaboration summit.
Makes a lot of sense to me. It would make designing and maintaining EntDistros much cheaper and It would help everybody.
Lets say 2.6.30 will be the first "driver backporting kernel" (DBK), which will be announced before all major distros release. Using that kernel would make a lot of sense for a lot of distros. When that happens using the same toolchain just comes natural .. might be the next DBK 2.6.35 that even the toolchain gets synched, but I think it will happen.
Canonical, Mandriva etc. and even Red Hat need to improve their products without spending more money and this is the way to do it.

FOSS is all about using what is already out there and improving upon that.
"Survival of the laziest ;)"

Everyone puts a little, or a lot in; that's FOSS.

What Mr. Shuttleworth is suggesting makes a lot of sense for all Linux users.

Red Hat's focus is not on the desktop, instead they choose to spend their time and energy on high-quality server support and development. Canonical, on the other hand, is a relatively small company, who have struck a chord with Ubuntu and have found themselves with a huge user-base including many new-to-Linux users who appreciate the great community forums and focus on Desktop ease-of-use.

While Red Hat's business model is very different from Canonical's, it behooves Linux communities to not act as a greater, united community. If we do, Red Hat will get more attention paid to their Desktop, and Ubuntu will get better server support. Linux will become less scattered, more coherent, and more supportable.

Perhaps you don't think that Canonical's small staff can or do add any real value to the Desktop Experience, but consider the massive number of users of Ubuntu. With such a huge base of _potential_ bug reporters, and at least a portion of those reporting, we end up with a rich harvest of bug information providing an avenue for superior hardware support, and software robustness. This supports Red Hat because it supports Open Source Software.

Keeping Linux users on the same page in terms of software release makes sense to me, because we all get the more-immediate benefit of everyone else's work, which is one of our biggest advantages.

ubuntu wishing they used rpms?


recently heard a radio interview from the linux.conf.au conference on ( melbourne, australia ) radio three triple r ( www.rrr.org.au ), and an interesting discussion with one of the main debian maintainers, martin shaw ( or was it martin short ? )

havent been able to track the podcast down, i think they run a couple of days behind the broadcast, but i'm sure a link will appear soon enough ( it would have been for may 14, 2008 ). try here for starters: http://byteintoit.wordpress.com

anyway, the same proposition was being put forward by this guy, albeit at a much lower level, in terms of common source repository infrastructure, and i presume some nice 'n easy way to pull per-distro build out of such infrastructure.

the interviewee repeated ad-nauseum a desire to 'work with' the redhat/fedora people, but curiously also claimed not to know too much about how they organised things. perhaps he tries to 'clean room' as much as possible? dunno. anyway, aside from my conclusion that it was all a pipe dream, he did come across as very balanced ( avoid flame wars, etc ), and articulate. clearly had been smoking crack, but presented the idea well enough :)

the notion that mr. shuttleworth seems to want to run with of 'linux version 2009' will comprise the same basic core components, wired up the way a particular distro wants to do it is of course the logical end result.

and on this front, i completely agree with the devil ^h^h^h dag here - just far too impractical, and points more toward canonical coming to terms with just what is involved in providing stable, long term support for enterprise customers.

and yeah, all up it really does sound to me like canonical has realised they have hitched their wagon to the wrong horse - debian runs a fine distro if you know what you're doing, but redhat rules the enterprise roost for good reason .

( full disclosure, i run fedora on desktop, centos on dozens of my company machines, and recommend redhat to our customers for support, stability, and ease of management )

Dag, Red Hat is a


Red Hat is a organization that works very efficiently doing one thing: Red Hat Linux maintenance. Red Hats user experience has not really changed that much since the introduction of Red Hat 6.0. RHs organization produces very little new software that is much bigger that what one engineer can cook up within their cubicle.

It takes brave newcomers such as Ubuntu and SUSE to drive the Linux usability to levels that are acceptable by real life end users. This is reflected by their exploding userspace. Please appreciate this effort: it's not an easy task. At the moment Red Hat found this effort too big for them to handle.

So, in the end of the day Red Hat will benefit from this work greatly as this work is really something Red Hats organization was never geared to do.

jmk, time to try CentOS 5

jmk, time to try CentOS 5 out. The only way I can explain your comment is that you left Red Hat in the 6.0 era and missed out on anything new :)

Plus, you are ignoring that Red Hat is the biggest single contributor to the kernel and Gnome, not Ubuntu. Red Hat leads by contribution most of the projects that comprise a Linux distribution and while I don't want to belittle all other contributions, which is vastly more, your comment does not give credit where credit is due, quite the opposite.

Red Hat does not directly focus the desktop, but indirectly does by means of Fedora. Saying that Red Hat has not done any improvements since the Red Hat 6.0 era is ridiculing all businesses that pay Red Hat for their services.

I can only advise you to work for a big company that is deploying Linux and look at their requirements, problems and solutions.

And yes, this article is about Enterprise Linux, not about Ubuntu desktop. I know that the dynamics are very different.

Wow..Deja Vu

This all sounds so familiar...The thing I've never understood about RHEL is it's insistence on focusing on the server vs. the desktop. Everyone I know who runs linux on a desktop runs Ubuntu (or Kubuntu). Just like years ago when everyone ran Windows on their desktops and Unix as a server platform. Once MS injected themselves into the server space it was over. People could take their existing desktop paradigm and shift it to their servers. That's what gave MS the advantage in the server market. We all know it certainly wasn't the strength of their product.
Now we see RHEL making the same mistake Unix made all those years ago. Laughing at this upstart desktop distro maker while touting their server license sales. When the worm turns, it's going to turn quickly. All those desktop Ubuntu users are going to start questioning the validity of RHEL's dominance in the server market and it's going to be over. The way to win market share is to be in your potential markets face every day. If RHEL cedes the desktop market to Ubuntu (which it looks like they may be doing) it will haunt them later.

I agree. From that

I agree. From that perspective Red Hat is risking their long-term business. Once all these Ubuntu users enter the Enterprise market they unmistakenly will bring in Ubuntu in the Enterprise, maybe to Red Hat's demise. Maybe.

What I do not agree with is that RHEL is the solution to this, nor that Red Hat is touting the desktop market. I think the biggest problem is that from an economic point-of-view, there is no desktop market. There may be an Enterprise Desktop market, but that is one where support is required and Red Hat is already providing that.

By the way, my opinion was always that Fedora, Ubuntu, OpenSUSE and others are not the solution for the desktop for most users. Even when home users don't need the support, they do need something dependable and something that does not change every few months. So my believe is that CentOS, RHEL, Ubuntu LTS and SLES are a much better fit for desktop and appliances in general because they do not require the same level as maintenance as most bleeding-edge distributions.

Simply think of your mother. Are you going to update her computer every 6 months or do you prefer that she can use the same environment as long as possible (+5 years). Possibly the same old OpenOffice she is accustomed to and works.

Sadly, the people promoting Linux are usually the people that get away with updates and upgrades and non-technical people are influenced by the people that do not necessarily understand their needs.

If you use your computer for email and web-browsing, spending a day to upgrade your computer every year is a big big waste of time. If you can reduce this to once every 5 years, possibly when you buy newer hardware anyway, you win.

The Linux desktop is almost there. Newer versions of the desktop offer less and less new reasons to upgrade my desktop. In fact I am running CentOS 5 since last year with little need to upgrade. Maybe I go with CentOS 6 and possibly I buy myself a new laptop for this purpose.

Still one problem

So my believe is that CentOS, RHEL, Ubuntu LTS and SLES are a much better fit for desktop and appliances in general because they do not require the same level as maintenance as most bleeding-edge distributions.

Exactly. Look at the main competition: one can install Windows or OS X once, and use it for many years. People want to surf, e-mail, and do word processing. They are not interested in the latest compiz or Rhythmbox, as long as it works well.

Though, I think there is one problem with the current enterprise distributions: the driver backports are server-oriented. You'll mostly see new drivers for NICs and disk controllers. The enterprise distributions do not work well with, say, the latest MacBook.

Of course, this is a gap that CentOS can partly fill through driver packages.

Exactly. Look at the main

Exactly. Look at the main competition: one can install Windows or OS X once, and use it for many years. People want to surf, e-mail, and do word processing. They are not interested in the latest compiz or Rhythmbox, as long as it works well.

They can do the same with any Linux desktop distro as well. I'd also like to point out Ubuntu is easily upgradeable from major release to major release. As far as I know, Windows comes closer to that than OS X. OS X also has frequent updates, and sometimes their software requires the absolute latest, forcing everybody to make the move frequently.

But the main point is: nobody's twisting your arm to upgrade your Linux desktop. It just happens that updates are more often while retaining support for older versions for quite awhile.

RHEL doesn't have to worry about Canonical....

If we just take the latest Canonical LTS release (8.04) I think it's safe to say RedHat doesn't have to worry about Canonical's offerings in the server market just yet.

Ubuntu and it's deratives are great for desktop users, new to Linux. However many seasoned *NIX veterans will quickly be put off by some of the practices Canonical takes (using loads of beta-quality packages in a LTS, breaking things midway in a release cycle etc etc) Canonical's QA department is still seriously lacking compared to RedHat. In all the time I've worked with RedHat software I've only seen a single serious goof (bind package which broke the DNS service) and a few minor ones (RHEL4 net-snmp packages, 2.6.9-42 kernel in RHEL4 wrt WiFi)

Also, Canonical doesn't (yet?) have the tools for people to manage large installations, such as RedHat's Satellite package. Such a tool is a must-have, imagine having to manually update hundreds of servers, containing 3 or 4 different releases, after issues have come up.

As for the desktop, I prefer my desktop to "just work" for a long time, so I use RHEL/CentOS there as well. Which means I only have to worry about upgrading every 3 years or so.... having tried Kubuntu for 2 years, constantly having to "fix" things after even minor upgrades was getting quite tiresome. Now, thanks to Dag and others like him I can still get most of the "bling" that's in Ubuntu on my CentOS/RHEL machines without sacrificing stability.

RedHat in decline

A fundamental problem RedHat, Ubuntu and others will need to face is that Linux distributions are maturing and becoming a commodity. That means that as the hardware/software becomes more reliable and more stable, there is less and less reason to pay for support for one distribution, when the next offers a perfectly stable (and free) alternative.

Don't take my word for it - check out SecuritySpace's latest Linux/OS Hosting Report Compare the top distributions, and you consistenly see that RedHat, SuSE and Fedora are losing market share, while Debian, CentOS and FreeBSD are gaining. Month over month, without fail.

Add to that that hosting providers used to offer almost exclusively either Windows, RedHat or FreeBSD, and suddenly Debian and Ubuntu are becoming a much more popular alternative.

This all points to a situation where RedHat's days are numbered. And yes, ultimately, if distributions are synchronized, (which RedHat will never agree to), it would allow other distributions to steal even more quickly RedHat's advantage, what little they have left, because it would allow for a more direct comparison. When you're losing market share, the last thing you want to do is to make it easier for your customers to see that your competitor is offering the same or better.

I find it interesting that a

I find it interesting that a company with a $20 million profit and a $140 million revenue is compared to a tiny company with probably no profits at all, on equal grounds.

Like, where are you looking, guy? You can't see through all the hype?

You completely missed the point

Read the opinion piece again. Canonical is trying to compete "on equal grounds" with RH. The point dag was making, as I read it, is that by syncing up with other more established distros it can reap the benefits of work others have done. I can't say whether this is actually the case (I spent 14 years actively involved in the Linux/Open Source world but had to retire in 2004 due to health reasons; these last 4 years I've been just a user).

Ubuntu vs RedHat vs Blah Blah

As a systems analyst for a University in British Columbia I find myself always playing catch up with the latest enterprise releases. It takes us an average of about 1 month to build the scripts, test them and overall certify an environment before deployment. We only support long term supported distributions for just this reason. From that perspective I could see how it would be nice to have coordinated releases. It would give people a change to test and certify all the versions nearly simultaneously. We could track compatibility issues between the distributions and work around them prior to them being released and report back to the "upstream provider". This could benefit others extensively.

On the flip side, Ubuntu has a *long* way to go for enterprise deployment. RedHat and Novell have been far ahead of Ubuntu in this space. It seems that Ubuntu suffers from what I call "Windows syndrome". The Ubuntu team has focused so much time and energy on the *user* Desktop that the *enterprise* desktop deployment experience is very difficult. Preseed is no where near autoyast or kickstart in terms of ease of use, and Ubuntu only seems to support a subset of Debian's preseed.

It would be nice if there was more collaboration between them to ensure a better experience for us administrators too!

BTW: If anyone knows how to tell preseed to only delete GNU/Linux partition types please let me know. Autoyast and Kickstart do this incredibly easy but I can' find it anywhere in the preseed documentation.


We use CentOS for our servers, Ubu/Kbu for user desktops. It's a good mix. CentOS provides us with stability and reliability (yes, I know Debian, Ubuntu makes a good server but CentOS is by far more predictable), and the 'buntus provide the bleeding edge (i.e. multimedia mainly) packages users want. Simple to me really. chkconfig is much more efficient (I know it's heritage, thank you) for enabling/disabling services than anything a Debian based distro is offering. Install software, run chkconfig, it's done. Server side, that is indeed a thing of beauty. My hope is that there can be some co-existence between the two camps as each has compelling reasons to be used.

And thanks Dag, really, for all the work you do in making my life easier, the rpmforge repos are a mainstay for things like Munin and countless other packages you port for the unworthy masses like myself.

Engineering is paramount

Red Hat's first foray into the "Enterprise Linux" market was 6.2EE which was essentialy Red Hat Linux 6.2 with a support contract for those that wanted it. It wasn't built differently, or tested any more (or less) than any oter previous version of Red Hat Linux. At the time, running a Red Hat distrubution seemed to me as though it was a constant treadmill. Build a farm of servers this month, and next month or a few months later, there is a new version released. A choice had to be made whether to move ahead with the new release or to stick with the old release based upon the relation of this farm to the previous one. Software vendors were in a worse position. They would test their software for a given Red Hat release, and would keep testing the same version of their software on subsequent versions of RHL. At the time Red Hat released patches for a given RHL version for quite a long time .. although I don't believe there were too many hard and fast timelines until the end of RHL, all I know is that I never had to reinstall a newer version of RHL because that version was being EOL'd. Once Red Hat became focused on the enterprise market, they changed their thinking, and more importantly changed their engineering process. Consider this. RHEL 2.1 uses a 2.4.9 kernel foundation. Although RHEL 2.1 will be out of support in 2009, Red Hat is still maintaining a kernel that has *7 years* of backported patches. Did I mention that they guarantee API/ABI compatibility for *7 years*? Who does that? Red Hat does. Any other takers? Didn't think so. I know that certain Ubuntu relases such as the most recent releases are dubbed LTS. Is there anything different about these besides the name and how long patches are released? Were considerations made as to what would be supportable and maintainable for the next several years, or did Canonical just say "This is the kernel we are using because it is current." If I recall, Red Hat used the 2.4.9 kernel as its foundation for RHEL 2.1 (aka Red Hat Advanced Server) because there was a significant change in memory management in the 2.4.10 kernel that Red Hat was not interested in using. These are the types of decisions that *MUST* be considered when choosing software to include in an "Enterprise Linux" distribution, most notably the kernel. For everyone that complains that Red Hat and subsequently CentOS are outdated, I have 2 words, centosplus, and rpmforge. Both centoplus and rpmforge as well as others I'm sure fill the gap that isn't viable for Red Hat to fill with updated versions of software. Enterprise Linux is all about stability, support contracts and having no surprises. If you want newer versions of software on any Linux distribution, knock yourself out, install whatever you like, but don't expect Red Hat or anyone else besides yourself to support it. If you buy a Ford Mustang, and replace the engine and transmission that's great, assuming you know how to do these things, but if your engine causes the car to explode, you cannot blame Ford, especially if you had to replace Ford parts with aftermarket parts to accomodate your new engine/transmission. The same applies to third party software and Red Hat support contracts. I think Ubuntu has done great things for the desktop, but until they start thinking like an enterprise company with an enterprise Linux distribution and those changes are reflected in their distribution, they will not be a major player in the enterprise space.

its really a honor to be

its really a honor to be writing to u dag..

most of the kids and people getting introduced to linux nowadays, r playing with a version of the 'buntus..
its a similar case like how it was while windows was gaining popularity... the kids back then who used windows went ahead in life and started implementing more of windows because they were more familiar with it.
its very obvious the way things r going that redhat will loose more of their market share.. the smartest move for redhat is to collaborate with cannonical.
If not anything else, atleast a lot of standardization will come into the common linux distros..

We use Ubuntu for everything

Hi Dag and others,

I completely understand your viewpoint and where you are coming from. You don't want Redhat to invest a lot of money on kernel drivers and for Ubuntu to simply come along and benefit from it, even potentially taking market share away as it does so. That's not fair.

However, your comment about a new install taking a whole day surprises me and shows that you may have other things to worry about. In our company, we use Ubuntu for everything simply because it's a release that doesn't take a day to install. Updates also work _extremely_ well. I know on my desktop with Hardy Heron I typically update my machine once a week or every few days, depending on how serious the updates are. It usually takes about 5 minutes and I've got all the security patches and latest bug fixes. When updates are that easy, it tends to happen more often. Server updates are also easy and "just work". Because our staff are familiar with Ubuntu from the desktop and updates are so easy, we decided to use Ubuntu for all of our servers. We haven't regretted that decision in the slightest.

Now, if other people have the same experiences we've had, RedHat may have to rethink its strategy a bit, focusing more on the desktop and easy package management. Otherwise, it will lose market share to Ubuntu and will have less money to contribute to the kernel we all love. To that end, I would propose RedHat switch to apt and use as much of Canonicals packaging technology as possible. There is no reason that RedHat can't benefit from Canonicals work as Canonical has benefited from Redhat's. It may, in fact, be necessary.

Ubuntu is not Enterprise, only LTS is !


I think you missed the point.

It is not about Canonical/Ubuntu stealing drivers. All of the stuff shipped is in the kernel. It is also not about Canonical/Ubuntu stealing patches, everything is open source. It is not unfair, they can do that. Red Hat has no lock-in like proprietary vendors.

It is just very inconvenient today for Canonical/Ubuntu to do the work and test with their own version of the kernel. They cannot afford to spend the money to do the work like Red Hat and Novell do.

Also, new installs with RHEL and CentOS don't take more than 5 minutes, I don't know what you are referring to. Applying updates and managing your systems is also a no-brainer with RHEL and CentOS. I can only conclude from your statements that you have litle or no experience with RHEL and CentOS.

And in fact, you won't get bug-fixes and enhancements in an *real* Enterprise distribution simply because enterprises do not want those being applied in production. You don't want to introduce new bugs when in production and only update in a very controlled fashion.

So only security fixes are available during a normal release, and maintenance releases every 6 months provide new hardware support, bug-fixes and feature enhancements.

Ack. My shot at it.

Hey Dag, sup.

My shot at it, from a more technical POV: http://dev-loki.blogspot.com/2008/05/re-ubuntus-pipe-dream-true-free.html

All non-technical motivations aside, what's striking me is the sheer cluelessness of Mark's idea.
Sync kernel, KDE, GNOME, OO.o and Firefox code freeze and release schedules ? Yeah, right.

Totally missing the point of Open Source

As someone who uses Linux both for enterprise and desktop purposes I am constantly frustrated by the fact that there is not more coordination between distributions across the board. For the most part Redhat's and Novell's contributions combined account for no more than 20% of the Linux kernel work contributed. Are they just leaching from the community since they sell a product mostly built upon the work of others? No. The OSS community is about contributing what you want, and making money where you can (and want) while implicitly sharing the source for what you have created. CentOS would be the ultimate example of leaching and contributing nothing back, but remember that every version of RHEL before RHEL3.5 was a race condition laden forked piece of garbage (especially RHEL2.1 which didn't even have the same scheduler as mainline). Having coordination across distros will help the entire Linux community be much more responsive to bugs since fixes should be far better tested and far more constant across the board.

It's hard to predict so far, but it seems as if the package management and updating is far simpler in Ubuntu 8.04-LTS and in RHEL or SLES (which require annoying registration stuff for each machine). My thinking is that this may make Ubuntu the best choice for tracking fixes for internet exposed servers. Like RHEL and SLES the Ubuntu 8.04-LTS release will benefit from the combined work done by IBM, EMC, Oracle, HP, SGI, Google, XEN, EMC, QLogic, LSI, Adaptec, Windriver, et. al.

Coordination of efforts would benefit all of these players by making all of the Linux solutions more robust and consistent.

The big difference between Ubuntu and ALL of the other major *commercial* players is that the base product is FREE (as in beer) including base product maintenance, and this may create vast new opportunities in the corporate market.

The way I see it, the previous Ubuntu 6.06-LTS was sort of a test run without all of the important core server support (BTW provided by HW vendors more than by RedHat or Novell), and they had not yet come up with the point release scheme they now have scheduled. As a result there is not yet ANY fact based information regarding their enterprise penetration since the product was JUST released, and NO enterprise based product by any company will ever reach product before its first point release (and probably the third would be a better bet).

IOW give it a year and then get back to me. :-)

Re: Totally missing the point of Open Source

You MUST be joking...
Ubuntu managing might be nice for 5-10 servers, but imagine having 500+ servers.

Do you want to log onto every box that needs updating and do a sudo apt-get update (of if you want everything bar a few packages updates having to specify exactly which packages to update)

Once you've worked with RedHat Satellite, you don't WANT to go back : At work my team runs around 600 servers. In RedHat we've got channels like this :


Normal patch routine : first wednesday we do all the test channels, week after all acc channels etc.

I just have to select a system group, select which patches/errata/updates I want to push (both default RedHat and our custom RPM's), select which targets I want to patch and select a scheduled time and forget about it. With a few clicks I've patched somewhere between 10 and 300 machines.

The registration stuff you find frustrating to us is a plus : the rhn daemon checks in every 5 minutes with satellite. Through satellite I can issue remote commands to multiple hosts, I can see which hosts might (still) have a nasty security issue, which hosts aren't checking in at all (and thus not getting patched) etc etc.

Now that RedHat has opened up the code to satellite I expect Novell and maybe Ubuntu to include it within a year, as this is VITAL to the bigger enterprises.

Ubuntu desktop

It's interesting to note here that many posts mention redhat is targetting the server market whilst ubuntu is targetting the desktop market.
Why then does the ubuntu desktop appear hardly any different from other (imho less buggy) offerings. I think they must have a v. good marketting dept as they don't appear to offer anything different to other distros, but make it appear that they are.


Ubuntu was a easy install on my laptop and everything worked.
CentOS not so good.Had to go through the rpm maze for certain applications for CentOS on my desktop, Ubuntu has them available for download and install no problem.
RedHat and CentOS may be great for servers but Ubuntu certainly got my attention as a user.