Distrowatch is one of the most famous websites about Linux distributions (and other Unix-like operating systems) that has been in service since 2001. It covers the new releases of a huge number of Linux distributions in its database, and also has a special “ranking” algorithm for those distributions.
What we unfortunately notice in a lot of discussions online about Linux distributions is that people tend to use that ranking in an attempt to figure out which distribution is more popular than the other. But this is totally wrong, as the Distrowatch’s ranking algorithm does not have any base that links it to the distribution’s users in reality. The website itself doesn’t claim that it is an actual correct ranking of Linux distributions.
In today’s article, we’ll quickly learn why this is the case.
How Does Distrowatch Rank Distributions?
Everything about the ranking algorithm of Distrowatch is written in their “Page hit ranking” page. The algorithm simply depends on the number of hits the distribution’s page receives per day, and it calculates only 1 hit per IP address per day (to prevent duplicates and spam).
This means that if you were searching about “Fedora” in Google, and came across the Fedora’s page on Distrowatch, the counter will simply get increased by 1.
Same goes if you were browsing the site yourself and clicked on the Fedora page, but you can contribute to the counter in a maximum of 1 hits per day as we previously said. No matter the other ways you access the Fedora page in the same day, it will be increased only 1 time (per your IP address).
You can see some possible options on the Distrowatch’s homepage to rank distributions in the latest 6 months, or 3 months, or 30 days or 7 days. All of these options are depending on the measured hits-per-day counter above.
Is This a Good Ranking?
Now, Distrowatch itself doesn’t claim that this is a reflective result of users of that distribution:
The DistroWatch Page Hit Ranking statistics are a light-hearted way of measuring the popularity of Linux distributions and other free operating systems among the visitors of this website. They correlate neither to usage nor to quality and should not be used to measure the market share of distributions. They simply show the number of times a distribution page on DistroWatch.com was accessed each day, nothing more.
And they are fully honest and correct. This ranking does not represent how popular a Linux distribution is at all, for many reasons:
- There’s no connection between the size of the user-base of a distribution, and the hits it gets on a 3rd-party Internet website.
- The number of hits itself goes up each time there’s a new release of that distribution, and likely stays still otherwise. This happens because Distrowatch publishes distributions’ new releases on its homepage, so if you are a daily Distrowatch reader, you would click on whatever is published today. So the distributions that produce more releases of them each few month are more likely to rank more in the Distrowatch ranking. This favors rolling-release distributions for example, while creating the illusion that the distributions which get released only each 6 months or 1 year are not that common.
- Again, this is only about Distrowatch readers, which probably do not exceed 0.000001% of all Linux users. So thinking that this metric can be considered a somehow correct ranking mechanism for Linux distributions’ popularity is far-fetched.
That’s why we urge you to never use such ranking in any discussion related to Linux distributions and their popularity. And if you see anyone trying to use it to rank distributions, just throw this article in their face and call out their BS.
What’s a Better Alternative to rank Distributions?
Right now, there doesn’t exist an easy way to measure how popular a Linux distribution is (from the users side). A good start would be to ask distributions to release the statistics of the number of downloads they are getting, but most Linux distributions simply won’t do it. The downloads number is also not that good to measure popularity, as a lot of people may download the same distribution many times, and even those who downloaded it one time, can possibly install it on tens of machines.
Another alternative could be releasing the hit statistics for the official distribution’s repositories. Almost every user may need to download a certain package or an update from the repositories at least once every few weeks, so if we could access the logs of how many unique IP addresses are accessing the distribution’s repositories mirrors per month for example, we may gain a good vision on how popular that distribution is.
While this alternative is theoretically good, the issue about it is that it won’t count offline installations. People from both sides can argue with strong reasons why offline installations are important or not important, but it leaves us in an issue anyway. Additionally, this would count Linux Mint users, Kubuntu users and Ubuntu MATE users all as Ubuntu users, simply because they are using Ubuntu’s official repositories, which is not a nice thing to have.
At the end, it sounds like each methodology has its own issues, but some are way more better than the other. Still, do not get tricked by people who try to use Distrowatch’s visitor statistics to rank all the Linux distributions out there.
Hanny is a computer science & engineering graduate with a master degree, and an open source software developer. He has created a lot of open source programs over the years, and maintains separate online platforms for promoting open source in his local communities.
Hanny is the founder of FOSS Post.
People reacted to this story.
Show comments Hide commentsStill refusing to call the distributions (not operating system) by their full name “GNU/Linux”, eh?
Still refusing to call it by the full name:
“GNU/Redhat/Canonical/Suse/Slackware/Debian/ . . . AMD/Intel/IBM/HP/Dell/ . . . Jan Doe/ John Doe/ . . ./Linux
Go fondle RMS’s ballsack zealot.
Does it really matter? No. It doesn’t.
I’ve always said that Distrowatch was a poor measure of popularity. Not only for the reason if the article but also because Distrowatch still counts different official Ubuntu builds separate (Ubuntu, Xubuntu, Kubuntu, etc) while combining the builds of other distros into a single count. Not a fair comparison. I’m not saying that distrowatch is doing this intentionally or that they are trying to make this unfair comparison. Although, I have seen this data used to make such an unfair comparison many times. This was especially used in the pass when a certain distro finally knocked ubuntu out of the top rank. The Truth was that if you added up the count of the official builds Ubuntu had more then twice the counts of the top ranked distro.
Well, then create a tool that measures distro downloads, or something of that nature. I don’t know. People are interested in that information.
DistroWatch has always been honest about what it’s ranking is about, page hits. There is no way DistroWatch can know if you go to the Distribution’s site, download the Distribution, and run it. They used to have a page where you could register what Distribution you were running, but I don’t think that is active anymore. I mostly use DistroWatch when I want to change Distros, to read the weekly newsletter, and to keep up on when possible Distros I might use are released. It is great for someone new to Linux, as it has a search function to find suitable Distros for yourself, and a page with package management and commands for newbies. If you believe page hits equal popularity, that is on you. DistroWatch is great for distro hoppers. I started on DistroWatch when you had to send to for the discs to a company, I forget the name now, something like OSD discs. My first disk I sent for was Xandros!
I wrote DistroWatch with the suggestion that the kernel could have an applet in it that each time it is installed would send data to some on-line database then the applet would self destruct or be removed during the cleanup after installation. Mr Smith wrote back and expressed the opinion that this would be an infringement on the users rights [or something to that effect]. The code, or applet, would only send a ping, or something else, that would be received and the database would count it.
It could also be interesting to know additional information such as CPU. I would not mind letting the Linux database know information on what type of computer I have and even my general location. I would not consider such information sharing to be an infringement on my privacy. Such information might reveal that Linux share of the total OS market is much higher than assumed.
I think that the DistroWatch team does an outstanding job. I notice that some new distros don’t get added to the DistroWatch website. It would be good if all new created distros folks had a organization that they could register their product.