Linux lacks any clear-cut system for determining which is the most popular or the best distribution, or which desktop environment is used more than others. That may be one of the major frustrations among Linux developers trying to spread the word about adopting the Linux desktop instead of Microsoft Windows or Apple’s OS X.
Linux, like the countless denominations of religions, is a product of technology, ideology and endless options. Unlike its competitors, the Linux OS is not sold. It just is available to those who know about it.
One reason for the apparent uncertainty surrounding Linux is that no single company markets Linux as a product. The open source operating system lacks traditional sales and distribution channels, so there are no warranty or registration numbers to tally. Likewise, no store receipts exist for off-the-shelf purchases of boxed Linux operating systems.
Similarly, little if any tracking of actual computer purchases with a particular Linux distro preinstalled occurs. As a result, potential Linux converts are left to their own initiative in learning about options and usability issues.
“Measuring Linux adoption, especially on the desktop, has always been — and will likely always be — a difficult task, due to the lack of empirical data,” Jeremy Garcia, founder of LinuxQuestions.org, told LinuxInsider. “While a couple distributions have attempted census-like projects in the past, they typically didn’t end up being as useful as originally hoped due to their opt-in nature.”
Many publications claim to track the most popular Linux distro by frequency of downloads each distribution receives month-to-month. This causes the most popular rankings to change frequently.
However, counting downloads is not an accurate indicator of actual use. An alternative method — counting responses to user surveys — lacks precision and credibility.
Even a successful census-like project would only give a partial view of the overall Linux ecosystem. The reality is that a single download might be used to install Linux on thousands of machines or on none, Garcia explained. That typically leaves other adoption statistics to be used, such as Web analytics.
“These have historically likely undercounted Linux adoption, perhaps dramatically, and are anecdotal at best,” he warned.
Some Linux distros that “sell” their desktop version using a subscription business model may have a firmer count on the number of active users its brand has. That is the case with Red Hat Enterprise Linux (RHEL).
Since Red Hat is a publicly traded company reporting quarterly financial results, it has a very clear understanding of how many customers it serves and how many subscriptions they are using, according to Mark Coggin, senior director for product marketing at Red Hat.
“While we do not publicly release volume numbers, our consistent growth, revenue and inclusion in the S&P 500 speaks to our rapidly growing base,”Coggin told LinuxInsider.
Red Hat also gleans its market share from technology research and advisory companies. For example, IDC reports market size and vendor share in unit shipments and revenue, then ties those numbers to its tracking of several operating environments, he said.
“We conduct our own market research every year that provides usage estimates for Red Hat Enterprise Linux compared to other commercial and community Linux distributions. These numbers are kept confidential for our own strategic and business planning efforts,” explained Coggin.
Download Stats Meaningless
Statistics on how many people download a Linux distro amount to nothing useful. The Fedora Project community sees lots of first-time users connecting with older versions of its well-known Linux desktop trying to update the OS. The Fedora Project is sponsored by Red Hat and is sometimes thought of as a breeding ground for new features in RHEL releases.
“Download numbers by themselves are not representative,” Robyn Bergeron, Fedora Project leader, told LinuxInsider. “One download could burn an installation disk that spawns 500 installations. So there is no correlation between download numbers and actual users.”
Even counting users who subscribe for paid support gives false results on how popular a distribution is. That only provides a limited number of users.
Knowing that limited number of known support subscribers does little to assess real user numbers. You can not count them as using a particular desktop day-to-day, Bergeron explained.
No Value Added in the Mix
Popularity and download rankings speak to specific markets, typically defined by free users, noted Red Hat’s Coggin. Popularity based on downloads is a measure that fails to take into account the value delivered to the customer.
In other words, who uses your product can shout volumes about a Linux distro’s stature in the pecking order of most popular desktops.
“We would regard Red Hat Enterprise Linux’s use by the New York Stock Exchange for its trading systems — and at many other stock exchanges — and by hundreds of other global entities to run their core business as powerful indications of the product’s merits and suitability for the enterprise market,” Coggin declared.
Another factor is the implied damage that ranking Linux distro popularity does to the concept of a unified professional operating system. Counting downloads or trying to measure popularity harkens back to Linux’s hobbyist background in the 1990s through 2000s, Coggin added.
“It trivializes the markets where Linux is now operating,” he said.
Too Many Variables
Another fallacy in trying to determine distro rankings is relying on the number of downloads at uncountable outlets for installation disks. The various websites that track Linux usage apply different methods of accumulating their statistics — and far too many websites offer one or more versions of different OS distributions.
People get installation disks for Linux in such a large variety of ways. The increasing use of virtualization to run one operating system within another makes counting downloads even more inaccurate, noted Fedora’s Bergeron.
Then developers factor in differently. Some developers run numerous distros because they are developing applications for different distros. A developer can download one copy and make 10 more, she said.
“So you really don’t know an accurate installed user base number,” Bergeron concluded.
No Phone Home
Websites and mobile apps are notorious for placing hidden cookies to track user locations, surfing activities and buying habits, but that approach does not exist for operating systems. Web analytics can determine which platform a site visitor uses, but that does not reveal the specific variety or version of the operating system.
A couple of Linux distributions have tried embedding identifiers in the OS coding that would send a user confirmation to the distro developer, but no long-term, large-scale successful projects of this nature exist, said LinuxQuestions.org’s Garcia.
That is an area that could infringe on distributions that are freedom-focused. Those communities, including Fedora, tend to shy away from embedding tracking code in Linux distros, Bergeron insisted.
“We cannot embed any tracking features in a desktop. It’s not going to phone home with a user tally. I think some people think that approach would be useful. Besides, people do not necessarily want it known what they are using. It becomes a slippery slope,” she added.
Another type of slippery slope is the appearance of usage rankings tracked by various industry watchers. It is not uncommon for product reviewers to refer to one or more of these so-called distro rankings as statements of fact. Perhaps one of the more credible Linux distro ranking sources is DistroWatch.
Is it possible to accurately track real user numbers for Linux distributions? When LinuxInsider posed that question to DistroWatch officials, the answer was to the point: In a word, No.
“There is no way to know. Unless there is a complete break of privacy on the Internet and we are all required to submit data to a central database, we’ll never ever know,” Ladislav Bodnar, founder and executive editor of DistroWatch.org, told LinuxInsider. “It’s as simple as that, and anybody who tries to determine the usage is really just guessing based on various metrics. Some might be more accurate than others, but none of them is perfect or accurate.”
LinuxQuestions.org makes no real effort to track Linux distribution usage, according to Garcia, but his website lets users add which distributions they use to their profile — and he releases general OS statistics for the site once per year.
Range of Metrics
Linux communities and distro developer websites watch the soft statistics for clues on installed user bases. The amount of activity on forums and other factors hold some measure of waning or growing popularity among various Linux distributions. Canonical’s Ubuntu developers, for example, use a range of metrics to understand how Ubuntu is used, where its users are, how many there are, and what their overall needs are.
“This ranges from direct numbers such as the number of downloads from the Ubuntu site, through to feedback from OEM partners or subscribers to our Ubuntu Advantage support services,” Steve George, vice president of products and communications for Ubuntu, told LinuxInsider.
Ubuntu also monitors input from its community of users through their comments and uses metrics like bug reports and user answers generated on the help forums, he said.
“Our experience is that talking to organizations and people that actually use the software for different use cases is still the best way to get real user insight,” offered George.
Business Considerations Count
Does the Linux OS really need an individual distro ranking? Perhaps just for non-enterprise users.
“The download market is important to a small subset of Linux distributors because it is how their business model operates, but these days, it is a pin prick in the overall Linux universe,” said Red Hat’s Coggin.
Enterprises choose a particular Linux product to meet their essential needs. Part of the value of Linux is that there is a wide choice of distributions. The best choice depends on the planned usage, he suggested. Google uses its own Linux on millions of servers. Companies like Facebook and Amazon are running businesses, not popularity competitions.
Today’s enterprises have workloads that demand the high computing power that Linux offers, and as such, Linux adoption is accelerating across the board. As enterprises look toward deployment in the cloud, they turn to Linux for its ability to support next-generation workloads. In fact, eight out of 10 major public clouds are built on Linux. Many of them use Red Hat, according to Coggin.
Industry distro usage stats based on downloads represent little more than guessing games. So Linux users have to do the same kind of consumer research they would apply to any product purchase.
That involves reading the descriptions and reviews and then trying out different distributions. After a few trials and errors, users are likely to find a distro that is perfectly compatible with their hardware and comfort level, suggested DistroWatch’s Bodnar.
The major factors are going to be the focus of their use and their experience with Linux. Other factors users should consider are the strength of community support, the hardware compatibility, regularity of releases and security updates reliability, offered Ubuntu’s George.
One of the great strengths of Linux is that there is not one end-all and be-all distribution. That means most distributions have a fairly specific focus. For some, it’s ease of use; for others, it may be security, noted LinuxQuestions.org’s Garcia.
All the usual things matter — price, support, performance, capabilities, security, application availability — but with different emphases, concluded Coggin. Every distribution offers a particular mix aimed at a specific market and use case.
I think this difficulty in tracking Linux users really only matters if there is some active push to tip the power balance in the OS world.
I do not use Linux in hopes that someday it will unseat another operating system; rather, I use it because I have the freedom to manipulate my implementation of it to suit my needs. I do appreciate the fact that it is free software, but if efforts to displace Windows or OSX were to become serious, that would require dollars for marketing, mass distribution, and licensing. And with so many distributions to choose from, not many individual Linux developers would be able to raise enough capital to endorse their implementation above all of the others. This could lead to mergers of several different Linuxes, if you pardon the term, to strengthen fundraising efforts. One of the cornerstones of Linux is variety, a benefit which then might begin to fade.
In our connected world where most people use the Internet probably the best metric we could get would be which distros are accessing web sites. Ideally we could get those numbers for the most popular web sites and that would give a pretty good reflection on what is being used on the desktop. We do have those numbers from Wikimedia, which includes Wikipedia, the fifth most popular website:
I think that provides a really good indication of what platforms are being used in the real world. Sure, some users spoof their User Agent string, but I would guess that would be a statistally insignificant number that has an equal effect across desktops.
Obviously everyone doesn’t use Wikipedia, and ideally we could get an even more accurate number if we could get similar statistics from Google and Yahoo.
Even so, I think these numbers are a lot closer to the true platform use than either download numbers or Distrowatch page-hits.