Is the Importance of Commercial Linux Distributions Waning?
Examining Conventional Wisdom
I have long held that only a commercial Linux distribution with a strong total product offering (product, support, training, reseller channel, etc.) could make the transition to the mainstream market. I'm not alone in this, of course--it's what led to Red Hat's success at their IPO. Investors clearly believed that a strong company would generate value in the Linux space. And commercial Linux certainly has generated that value.
The conventional wisdom goes something like this: in the early days of a technology market, users are willing to tinker to get what they want. These "early adoptors" don't expect a complete and perfect product. They want cutting edge technology, and are always on the lookout for ways to apply it to their scenario in order to be a hero and gain a competitive edge. They have the inclination and imagination to project how the new technology will fit into the mainstream IT landscape down the road. Early adoptors are excellent evangelists for new technology, and getting them on your side at the beginning of a movement like Linux can really make a difference. Developing loyal user communities was the most important thing the distributions, both commercial and non-commercial, could do in the early stages of the Linux wave.
The only problem is that the early adoptors are not really a viable market when it's time to make the sales. In exchange for a cutting-edge but incomplete product, they will provide feedback, bug fixes, and testimonials--but they won't pay.
A recent post on the mailing list for my local LUG reminded me of this truth. The poster in question has long been an advocate for Mandrake Linux, and I've seen him recommend it time and time again to the group. So I was surprised when he asked the list if anyone had a current paid membership with the company's club--he wanted to distribute copies at an InstallFest, but clearly wouldn't pay even for a membership program clearly designed for early adoptors like himself. He didn't seem to think that anyone else should pay, either.
To get to a paying market, you have to make the leap to a mainstream audience. I didn't invent this, of course, as all of you familiar with Geoffrey Moore's books will know (the rest of you should read "Inside the Tornado" by Moore for an eye-opening analysis of emerging technology markets). Mainstream users of technology have different needs from the first wave of early adoptors. Their objectives are different. Unlike the early adoptors, they shun cutting edge features as too risky. They want solid, stable technology that serves specific purposes in line with their standard requirements.
Moore calls making the shift from meeting early adoptor requirements to meeting mainstream requirements "crossing the chasm" (and it's the name of one of his books). It's not easy, and probably more technology companies have failed than succeeded at this stage in the game. Part of the trick is to recognize that the mainstream market needs a more complete product--sometimes called the "whole product." This usually means all the things that will make a mainstream audience feel secure in making a change. Support is usually the first consideration, followed by training availability, the perception of related skills in the marketplace, a third-party market of add-ons, and the existence of an entity that will take responsibility for the relationship.
The conventional wisdom asserts (and I have argued as well) that only a commercial Linux distribution can provide the "whole product" to customers, and make the shift to widespread mainstream adoption.
But now I'm not so sure.
In a recent study from Evans Data Corporation, developers are showing a preference for non-commercial versions of Linux for the first time. The study included developers of the Linux kernel and applications that run on Linux. If the developers are favoring non-commercial distributions now, then we should expect applications to be more numerous and work better for those versions in the future.
What about deployments now? When Jupiter Research did a study of open source use in the SMB market a couple of years ago, they found that adoption was faster than in the Fortune 500. But the perception has been that open source adoption in this space has been slower. Could this perception be the result of the difficulty in tracking non-commercial installations? There is plenty of anecdotal evidence that small to mid-sized companies are opting for non-commercial Linux at the same time that larger organizations are relying on commercial versions.
Consolidated Information Systems, a 50-person Arkansas-based company, has been migrating their Windows servers to Linux over the last five years. So far they have moved file, web, email, and proxy servers. Nick Shaver, head of IT, tried a couple of commercial distributions before making a final choice for Debian, a non-commercial distribution. "We settled on Debian, and it's worked great."
An important factor for Shaver in choosing a non-commercial distribution was freedom from any vendor. "We just felt like the features added to new commercial releases were partly driven by the companies' need to make sales. But with Debian, the only reason for a new release is that it's an improvement over the last one. Upgrades are a lot of work, and we just didn't want to be driven to implement them by a vendor."
Cerient Technologies, a Linux consulting company based in Raleigh, NC, delivers mostly non-commercial Linux solutions to its small to mid-sized customers. Over 50% of the installations are the free version of Mandriva Linux (formerly Mandrake). This non-commercial version is used for file servers, web servers, mail servers, and even desktops in a few locations. According to Cerient President Jason Tower, "I don't see a compelling reason to use a paid-for version. The free version is stable, has all the packages available that we need, and it simply does the job. I have yet to see a case where a commercial version offers a technical benefit." The only benefit Tower sees in the commercial versions is having a "neck to wring," but he acknowledges that that value is overrated. "Sometimes the choice of distro is dictated by the customer or a third party application, otherwise we recommend the non-commercial versions."
Global Knowledge, an IT and management training company, has been gradually adopting Debian along with Red Hat. They migrated a significant portion of their infrastructure to Red Hat Linux about three years ago, and the systems are still in production and performing well. Recently the company began a build-out in a new data center.
A key part of the new build-out was the move to a single sign-on system using LDAP and Kerberos. The new services are being built with open source and running on Debian. According to Ryan Leathers, Manager of Platform Development, the decision to adopt Debian was based not on price, but on application updates.
"With Red Hat, we know that the packages installed are working as expected, but we know we're not going to get a lot of application upgrades. Take OpenLDAP, for example. With Debian, we'll get more updates earlier. Yes, we could go get the RPMs for Red Hat, but it's not automated and tested. With Debian, we have good confidence that everything will work out of the box." Leathers expects to use Red Hat for some third party applications like Oracle, but his preference now is for Debian.
Web Performance, Inc, makers of web load testing and web stress testing software, made a shift from Red Hat to CentOS, a non-commercial version based on Red Hat. The company maintains a testing lab with a collection of servers running Linux. Product Manager Michael Czeiszperger made the decision to switch when it was time for a wide-scale hardware upgrade in the test lab. Pricing had increased for Red Hat since their last deployment, and this prompted him to investigate other options.
"Instead of paying Red Hat, we paid a local consultant to give us personalized service. CentOS is rock solid, and we get free updates. It was just what we needed." Czeiszperger believes their experience is fairly typical. They had some Linux experience in-house, so they were comfortable going with a non-commercial version and relying on a local Linux consultant for the initial set-up.
Larger, better known organizations have also received attention lately for choosing non-commercial distributions. Duke University has chosen CentOS as part of a push to replace Sun hardware. And the city of Munich, ever in the news related to Linux, has settled on Debian. Ubuntu, a newer non-commercial version of Linux, is now being offered in Europe as a companion to HP laptops. HP is distributing CDs of a special version of the distribution for some laptops on request.
How did a non-commercial Linux upstart get the attention of a major vendor like HP, when there were more established--and commercial - desktop versions to choose from? And what does this signal for the future?
In the face of the Evans Data Corporation study and the anecdotal evidence, I've begun to doubt the conventional wisdom. Maybe it won't be a commercial Linux that truly breaks the barrier to mainstream. Of course, it could be argued that once HP created a special version of Ubuntu, it became a commercial distribution. But that can only be half true, however, since HP is only offering the OS separately, and not installing it. The situation seems really to be more of a certification on their hardware.
Maybe the HP-Ubuntu relationship is a window into the future. Maybe not. It would be foolish to speculate any more than this, but I'll do it anyway: maybe a general-purpose operating system is too ubiquitous to be pinned down by any one or two commercial entities in the long run. In the new freer market of software driven open by the Linux phenomenon, it is possible that non-commercial Linux will ultimately be favored, not because of the price, but because of the true freedom it provides.
As an open source business analyst, Ms. Winslow assists clients in understanding the technical and budgetary impact open source software will have on their computing environments. Her recent book, The Practical Manager's Guide to Open Source, guides IT directors and system administrators through the process of finding practical uses for open source that will integrate seamlessly into existing infrastructures, as well as understanding the costs and savings. Ms. Winslow is a frequent speaker and author on the topic of open source, and is the contributing editor of open source applications at LinuxPlanet and Linux Today. She can be reached at firstname.lastname@example.org.