Submitted by The traditional market share numbers would say that Linux is currently at less than 1%, but some more recent numbers suggest that it might, in fact, be almost even with the Mac. This all brings the question of how many Linux users are there really? Unfortunately, we may never know. Certainly, there is no way of knowing currently, but it should be possible to at least get a rough estimate.
Another consideration is that almost all confirmed Linux installs would be taking the place of a Windows install; a piece of information that, alone, could double market share estimates.
I have a suggestion on how we can get a more accurate description of how many Linux users there are. Traffic statistics on the software repository servers. Count the number of unique IP addresses accessing the software repositories each day, average it out over about six months, and you’ll have a reliable, though probably underestimated, statistic. It would count how many people are either getting updates, or downloading new software. Perhaps not accurate in a business situation, where updates would be cached to a local server, and the software set stays the same all the time, but a good estimate of home market share.
then there are those that dual-boot or similar…
hmm, do a VM install count at install/use?
Counting IPs wouldn’t work because a lot of hosts hide behind a NAT firewall. However you could count MAC addresses.
Unfortunately you cannot. MAC Addresses are not transmitted past the first router. After that only IP is transmitted (not counting special cases like packet encapsulation for VPN traffic and such).
Once on the local segment ARP is used to match up IP address (or whatever protocol is being used if not IP) to MAC address.
Well, I guess my networking skills are lacking, thanks for the info
Exactly. But this wouldn’t stop firefox from sending your Mac address as an anonymous usage statistic if you chose to do so. It could send the mac address of the default network interface, along with the operating system and other stats. Obtaining the mac address of a network interface is trivial enough for a program running locally.
Edited 2008-08-15 18:05 UTC
Just look at a security update in the base install (or a very frequently used package), the number of times that gets downloaded is probably a good indication.
Counting MACs wouldn’t work either since the MAC address that shows up is the one facing the internet behind NAT… Same issue as counting the IP addresses.
There are some distros that allow new installs to send anonymous info about the new install to servers. This would be a much better method.
Edit.. redunant. Never looked at thread view before posting.
Edited 2008-08-18 20:25 UTC
Only one of the many Linux installs I have done are taking the place of a Windows install. Normally I buy an “update kit” (motherboard, case with power supply, memory and CPU) and I add a blank hard disk, optical drive, and video card (if I didn’t already have these) … then I install Linux from a liveCD. It costs about half of the price of an equivalent Vista box from the store, it takes no longer after purchase to assemble and set up with all your required applications, and it is three or four times as functional.
The only install I have done to “replace” a Windows install is on a refurbished (second hand) laptop … even then I got the supplier to wipe the disk of Windows prior to delivery and I got $50 off the price!
A similar approach sort-of works for firefox update … except that many Linux users get firefox (or perhaps iceweasle) via their distribution’s repository, and hence cannot use “firefox update”. Firefox’s own figures would then tend to show a far heavier use of the Windows version of firefox than any other version as a result. Perhaps unique IP addresses accessing the firefox extensions page?
Even then, in just my domestic situation, there are four (sometimes five) machines using the same IP address via NAT.
This same problem would affect your method, even if updates were not being cached.
I believe they tried an Opt In to track kernel updates from unique installs. Basically, if a distro is updating the kernel package; it’s not a random software install and it’s probably in active use.
Few wanted to volunteer for having there updates lowjacked. This was about the time MS was back in the news over updates policy. I believe it was the validation ball-gag they required before you could get XP sp2 updates.
The problem of accurate market statistics will remain and in it’s probably better that it does.
– MS sales statistics will always be the licenses they sold to stockpiles not the number fo activated and regularily updated Windows machines; they keep that stat to themselves, guess it’s not good.
– Only Linux based OS purchased through distribution channels like retails outlets can be tracked as difinitively as the MS/Apple units shipped per year. The very nature of the platform means there will always be untrackable distribution methods including roll your own from tar.gz directly for all the various programs you want.
– It’s probably better that the various random ways to get your prefered nonMS/Apple OS regardless of branding or collection of commodity parts it’s build with. It is a culture and platform meant to impower the end user not the vendor’s shareholders.
My numbers on this are entirely made up, but I think in principle this is sound:
The fact of the matter is that probably 90% of linux users out there are using only 10 different distros. Most of those distros have some manner of package management (be it yast or apt), and are probably hitting less than 30 distinct servers. Given that most distros in my somewhat limited experience have automatically updated themselves, shouldn’t it be fairly easily be possible to ask canonical/suse/redhat/debian how many times a particular update common to almost all installs has been downloaded? That’d be pretty darn anonymous, fairly inclusive (excluding only those who have automatic updates turned off, or who don’t have that common library due to some quirk of their install). That really shouldn’t be terribly hard, should it?
How do the major distros track that across all the mirrors and how do they track a shmuck like me that keeps a local repository mirror to keep traffic load off the ISP’s wires?
Right now I’m mirroring Mandriva 2007.1, 2008.1 repositories for six or so regularily used installs and whatever VM build whim I have on the go. That’s six+ active installs being counted as one, maybe two.
Amazingly, as of today, my dad is a linux user!!
He went and bought an eeepc 900 with linux all by himself! Wonders will never cease! I can see trouble on the horizon for Microsoft..
(Just to clarify; he’s one of those users that have a hard time installing anything, and has steadfast held on to windows, even when my mum bought a mac.)
Edited 2008-08-15 07:01 UTC
And this begs another question: Exactly what counts as a Linux install? Do only the standard distros count, or do the custom-made netbook distros count as well, as many of them have almost no resemblance to desktop Linux? Do we count embedded devices? Servers? How about PDAs and Phones, or hacked devices such as an iPod with linux? Because if we’re just talking about Linux usage, I dare say it’s probably the second most used OS out there in one form or another, even if a system only uses the kernel and no other standard GNU tools. If we’re talking specifically about desktop usage though that estimate goes down considerably.
When getting statistics such as this, remember that Linux is not an os, it’s a kernel. So trying to get figures of how many people use Linux is almost certainly going to be inaccurate. You can monitor the major distros, but that only gives a fraction of how many users there actually may be. For example, I use Archlinux. I wouldn’t be included in a check of the major desktop distros’ usage statistics. But I still use Linux, don’t I?
Second most used? I don’t think so.
Any given person may or may not use a copy of Windows … maybe even two (one at home, one at work). But they are also likely to use a copy of Linux in their TV, in their DVD player or recorder, in their LCD photo frame, possibly in their mobile phone, in their home router, in their NAS device, in their car’s electronics, in their car’s GPS nav unit and in their fish finder!
Edited 2008-08-15 11:22 UTC
Youd have to seporate the various markets and count OS not kernels installed under the OS/Userspace usable combination.
Even now, we don’t cound Windows installs. We count Desktop, Server, cellphones seporately depending on the market.
I do think your pretty close to right though on the most used OS. Just the range of hardware supported from matchbox sized “servers” through industrial and car control computers to cell phones, through R/C radio controls and on, and on.. It blew my mind the first time I read through the arious hardware supported during a humble “make qconfig”.
I have to remain optimistically pesamistic. Optimistic that the OS and other options will continue to bring teh end user choice and benefits outside of the two normal suspects. Pesamistic because getting past the “but there’s no market there” excuse by ISV/IHV will continue to be hard to do without talking the mystic language PointyHair.
Yeah, there are embedded Linux everywhere, not just stuff like the popcorn hour or Nokia N810. Yesterday I discovered that a fishscanner* I installed at work was running Linux. But I’m not sure I’d count that as an install.
Now that I think about it, it did not come with any notice of GPL…
*An ir-scanner counts and measures fish that swim through the unit. It is used to count fish for example upstream a fishladder.
For example the Fritz-Box DSL modem + router + NATD all in one device that is all around in Germany. The manufacturer installed a kind of embedded Linux there, so is this an install, too? It’s a means that gets all the “Windows” boxes online, by the way, so it should count. 🙂
I’m not surprised that the Mac and Linux have both ~5%. Although there are some specific considerations (perception of security, refinement, ease of use) basically it burns down to price. Wealthy people will most probably buy a Mac to look cool and to step aside from the masses who use Windows, and low-end computers as well as netbooks come with Linux preinstalled for dirt-cheap computers that the least rich can afford.
But to have a true idea of how many people use Linux, the Mozilla Corporation is well positioned. Firefox is a widely-used application, be it by Linux, Mac or Windows users. Mozilla just has to monitor downloads and see the usage share, and to convert it into numbers. If Firefox represents 50% market share (numbers made up), if there are 2 billion computer users and if 10% of its downloads are the X11 version, it can determine the number of Linux users.
Unfortunately, counting Firefox downloads would probably not be as accurate as you think. Most people download the X11 version once and put it on several Linux machines (I know I do).
Is it really most people?
Maybe I’m wrong, but almost all of the people I know that run Linux tend to have multiple computers. They don’t tend to download programs once for each computer, as they tend to transfer files between computers (including seeding apt caches).
The people with only one computer tend to run Windows or MacOS (although there are exceptions, we are talking generalised ‘expectation values’ here).
Therefore I believe that counting Firefox downloads may not be the best way of measuring Linux deployments.
Counting unique MAC addresses might be a better way though.
Yes, MAC addresses is more accurate. Maybe a project like BSDStats.org counting MAC addresses and shipped with the majority of Linux distros would be interesting to have. Distro vendors would have to accept to ship it though. And users would have to click a checkbox to accept to send anonymous stats.
Actually no. What most people do is download it from their distribution’s repositories. So it isn’t counted by Mozilla.
That’s true. Also Firefox comes as a standard install on many distros (Ubuntu being one of them). Also Firefox is available in apt repositories, so you don’t have to download from Mozilla directly.
Firefox counting will be useless!
Lets just say Linux has 90% of the market share and let Microsoft try and prove otherwise! 😉
You get a better-integrated result with Firefox on Linux if you get the distribution-specific version via your distributions repositories rather than downloading the generic version from Mozilla direct.
Getting it via your distribution’s repository also means that the package manager database knows about it (and its dependencies), and it will also be automatically included on your menus.
For this reason, the majority of firefox installations on Linux systems I would suggest are not downloaded direct from Mozilla.
Another problem with using Firefox as a gauge for OS usage is the options available are different between different OS’s. I expect that 75% of Linux users run Firefox, maybe 25% of Windows users run Firefox, and really can’t even guess what percent of Mac users run Firefox.
I keep an installs archive so I can rebuild a system without needing the network connection right away; old habbit from the Modem days. That means I download FF once showing as a Windows FF +1 then reinstall from my own archives updating to the latest version install from time to time. That install can also be distributed to the machines I work on for clients or friends further understating the statistics.
On the nonWindows side, my FF comes in through repositories so again, it’s a Mandriva FF +1 when I update my local repository mirrors further understating that statistic.
Do we really know the actual market share of any OS?
I mean, how do we accurately know how many windows installs there are when a sizable number of windows users run pirated copies of their OS and a sizable number of non-windows users own a windows licence with their prebuilt desktop / laptop despite having wiped their machine the moment they unpacked it.
Beyond generating headlines, I’m not even sure it matters what the actual figures are either. It’s the comunity within that market share that makes the difference. Take OS X for example. Argueably they have small market share yet ask someone in film or music to recommend you a computer and they’d more than likely say Mac.
Exactly.
There was roughly 300million new pcs sold last year, and if osx made up 3% of then, that is 9 million mac users. Odds are, you know one or two of them
If the figures for osx and linux are similar, you also know some of the 4-6 million new linux users last year too
I know this is slightly off topic, but is still related to the topic; If companies are going to use statistics as justification for the support (or there lack of) for a given platform – they might also want to look at the quality of the user base rather than the quantity.
Having 300million sales per year may sound alot, but when over 50% of them can’t afford to purchase a legal copy of your piece of software – its completely pointless; these same people struggled to purchase a $400 machine – you *really* think that they have another $300 to spend on the legal version of your software? this is the delusion I see the Windows/PC embrace, thinking that some how PC’s are cheaper, and because they sell more, it means there are significantly more customers willing and able to purchase their (if they are a programmer) software.
Going back to Linux/*NIX – how many end users that actually use *NIX is interesting – percentage wise, for example – how many use it, and know they are using it? how many devices are being used right now which have linux but the end user doesn’t know – in the case of the eeepc or a smart phone of some description?
At the end of the day, does ‘raw numbers’ really matter? like with the quality versus quantity argument I made above – does huge numbers prove anything?
An fully functional OEM version of Windows Vista Home Basic can be as low as $60 when purchased in a store (checked on shopping.com). Included with the price of a computer I’m sure you can even remove a few more bucks. Spread out over 5 years for instance (estimate time of use of a computer before you give it away), then $60 represents only a dollar per month. It’s nothing. $60 is what my wife spends when she goes to the supermarket every Friday
Except that Home Basic is dog crap.
Vista is cheaper than it was when first released. An OEM version of Vista Ultimate can be purchased for as little as $160 at Tech Bargains.
I have Windows Vista Business SP1 on my main desktop and use it as an occasional terminal server for my Linux machines. I hacked it to provide support for unlimited concurrent RDP sessions.
It works fairly well and I can easily disable seamless operation to remote a full desktop rather than just application windows.
Edited 2008-08-15 15:51 UTC
[quote]Except that Home Basic is dog crap.[/quote]
Maybe. I chose the cheapest version of Vista to match the low-end $400 computer mentioned above.
Yes, but somehow I’m guessing you don’t live in India, or Thailand, or Malaysia, or Iraq, or ….
For many of these people the cost does make a difference. (Remembering of course that many Indians etc are unbelievably wealthy).
Who said anything about the cost of Windows; I said purchasing software AFTER purchasing the hardware. The hardware already comes with Windows, but what about Microsoft Office – what about all the software over $100 which are on offer.
The same people who talk about the huge numbers of computers out there, ignore the fact that unless these people purchase middleware like office suites, photo tweakers etc. the huge size does not help their own bottom line. Ok, I’ll give a callus example of where it doesn’t make sense. Indonesia has a population of 230million – do you think that someone would be stupid enough to think, “wow! that is a huge market to sell beer to” when 86% of the population are Muslim? Look at who makes up the user based rather than just looking at the numbers. Numbers by themselves are meaningless – they give absolutely no context.
People in the third world don’t have $100 just laying around, sorry to break the bad news, but for everyone else around the world, they don’t have money growing off tree’s. Even in New Zealand – do you *REALLY* think that a person sitting in the US$21,400 can afford to spend over a hundred dollars on a piece of software?
* Btw, $21,400 is the NZ GDP per capita (in US dollars); which IIRC is about the same mark as Malaysia the last time I checked.
Edited 2008-08-16 07:02 UTC
Indeed, this is something a lot of people does not seem to grasp.
$100 is fine if you make $2000/month. It is not so ok when you make $300-$400/month.
Even at $2000 cash in hand, it isn’t much; in New Zealand the cost for paying off the mortgage is going to be around $300 per week, then you have food for the family, maybe around $120-$130 if you’re really focused, $30-$40 for petrol, $20 per week for electricity and $10 per week for telephone.
Within that space of time, you’re looking at $2000 per month how it can be easily used up. $500 per week is not alot of money when you look at the cost of living these days.
Edited 2008-08-16 10:03 UTC
Do people really need to pay for additional applications? I mean, there are very good free applications (not necessary opensource). Some that com to mind:
OpenOffice.org; 7-Zip; Zip; jZip; ClamWin, AVG, Comodo AntiVirus, BitDefender Free Edition, Avira AntiVir PersonalEdition Classic; Spyware Terminator, Comodo BOClean, Ad-Aware, Spyware Blaster; Comodo Backup; GIMP; Pain.NET; Pidgin; Comodo Firewall Pro; WinSCP (SSH), SmartFTP; Filezilla; Amaya, Seamonkey’s Composer, nVu, Eclipse, KompoZer, Maguma Open Studio; Aptana, Trellian WebPage, HAPedit, tsWebEditor; Opera, Seamonkey, Firefox; Safari; ConTEXT, Crimson Editor; Eclipse PDT; Notepad++, Dev-C++, Dev-PHP IDE, PHP Coder; burnatonce, CDBurnerXP; Inkscape; mplayer; Audacity; Thunderbird.
All free of charge
Forgetting the anti-virus and anti-spyware applications for the moment, if you were to put the rest on a system and use it as your desktop … what would you need Windows for (other than as a vulnerable target for the anti-virus and anti-spyware to protect)?
You can get the equivalent applications on a Linux OS, also for free … but if you get a decent distribution they will all be pre-installed for you. That will save you the trouble of finding and downloading the applications.
Advice … forget the .NET applications. Lock-in. Instead of Paint.NET, use Krita.
On a low-powered limited memory system, installing Linux will also work better because you won’t need to have the anti-virus and anti-spyware running.
Win-win.
Kia Ora Kaiwai,
You are right, and I’ll try and elaborate.
For commercial purposes the marketers need the statistics to tell them who the “decision makers” and “influencers” in software purchasing decisions. Aunt Dolly will pay for what comes on her computer, but maybe not much else.
Its the smart IT folk who choose what gets installed on corporate desktop (for example). The relatively small number have disproportionate power when it comes deciding what gets purchased (outside of the consumer space). These people are usually university educated, although there are exceptions of course. They probably also have a disproportionalty high usage of alternative operating systems. Their numbers will always be relatively small, but their effect on the marketplace large. Therefore, there is no real need for folks to worry that only 10% of the smartest users are using alternative operating systems. That will grow naturally (in the commercial world) as they exercise their influence over time.
One more point. Your title mentioned statistics so I thought I’d add something about that. Techniques for estimating population properties from samples are very well established. The trick is to get a truly independent organisation to perform the sampling. The other thing is that the results have to presented ‘in context’, such as “x% of home user desktops have both windows and mac osx” or “y% percent of corporate desktops use windows 2000” or “zx% of engineers use solaris”. Without these qualifiers the “statistics” presented should be rejected by you all as completely meaningless (unfortunately, many journalists are very poor at mathematical reasoning so merely parrot figures out of context since they don’t really understand how statistics works).
There is no need for a scheme to count every single Linux download when random sampling of a large enough population will result in similar information.
Edited 2008-08-15 21:43 UTC
You don’t get it. Microsoft doesn’t give a damn about sales of Windows. They only like sales because it gives them marketshare, not for the money. (Okay, not true: the bean counters *always* want the money.)
Where Microsoft makes its money is the *platform*. Microsoft is and always has been a platform company. Windows is such a crap OS because Microsoft doesn’t make OSes, they make platforms.
What does the Platform get them? It gets sales of their other software. They sell office, mssql, and on and on. Each add-on costs money and almost all of them broaden the platform. Other companies then buy developer tools, write software and sell it for money. Owning the platform gives MS a competitive edge.
Microsoft didn’t like Netscape because Netscape was promoting a platform that did not rely on their own. Microsoft doesn’t like PHP+Apache for the same reasons. Linux is not really something Microsoft worries about because there isn’t one big Linux platform. If there ever were then MS would react very badly.
Really? Wow…I wonder what they worry about if not about making money
Both.
They make money from all the software licenses they sell. Using a tied platform allows them to sell more and to secure their business.
Could you elaborate? I don’t see the relationship between having a poor-quality OS and developing a platform.
Really? Why does Microsoft airs “Get the facts” advertising promoting their Windows OS and undermining Linux? Didn’t Steve Ballmer state opensource software was MS’biggest threat, referring to Linux, MySQL and OpenOffice.org?
The big Linux platform in question is Linux itself, which competes with Windows. The competitor to Windows Server, the .NET framework and MSSQL is the LAMP stack that has had dramatic success. So yes, there is a Linux platform and they do worry.
There is no Linux platform, there are a variety of ‘platforms’ that happen to use the Linux kernel (but don’t require it). Even the “lamp stack” isn’t a platform in the Microsoft style. Is it Mysql? Postgres? Perl? PHP? Python? Ruby? It’s all of these things, really, which is bad from a platform perspective.
MS is not much bothered by Linux-the-kernel. The extent that they are not campaigning against it is proof of this. The get the facts campaign is a pittance.
When selling a platform where about every element ties you to the OS, the OS itself only needs to be “good enough”.
Actually I’d say once you get enought inertia none of your products needs to be above average. Just improve the minimum to stay “good enough” and keep pushing the lock-in.
How many TOMTOM satnavs were sold in the last five years… ??
Not sure why that has not been turned into the best hackable device known to NIX
Then again I bet there are many more Linux VMs then any other platform… I must have a hundred or so…
I am sure there are not many people with a 100 Mac VMs, just think of the cost..
Edited 2008-08-15 11:13 UTC
osX doesn’t run so well without the hardware it’s designed to be the embedded OS for. I’d love to get an osX VM running but that means legally quiestionable work arounds or a development in VMware that hasn’t happened yet.
Actually, I’m unsure, can parallels split osX into mulitple VM segments on the same brick of hardware?
in denmark i can tell you one thing, there are more linux than osx users..
unless ofcourse theres some super dense blob of osx users somewhere hidden in a secret jobs church the normal society knows nothing about.
I say this because it simply is true, just by walking around, its not uncommon to stumble into a linux user, but people using osx.. i’ve seen that randomly around, like.. maybe 2-3 times in my life. Granted, there are now alot of crapple intel boxes around, all of which undoubtedly counts as crapple marketshare atleast one time, but is osx being used? no way in hell, those people bought it to install winblows on.
I think pretty much what we’re all saying is “Yawn”. So, what if MS claims x number of users. We all know it means little. Personally, for about 9 years now all I’ve used windows for is email because management forces it on us. But then I’m a developer/programmer. In my department, pretty much everyone that uses windows gets all their content from Linux/Unix servers so are they windows users or *NIX users? It really doesn’t matter. The stranglehold at the moment is Office and someday that will drop away also. But even then, I bet MS will claim that more than 90% of all the pc’s in the world run windows. Didn’t they call that the “Big Lie” theory. Repeat something often enough and everyone will start to think it’s a fact? Works in politics.
There are probably between 10 and 20 million desktop Linux users and many more server and embedded users.
Linux has about 50% market share in our house. My father and I use Fedora 9. My mother has an XP laptop with a restricted user account. I was contemplating the idea of migrating her over to Fedora but all she knows how to do with a computer is click the Start menu and open Internet Explorer :-P.
Linuxappfinder and osalt have been great sites for locating equivalent FOSS and commercial apps on Linux.
We also access our Windows software from within Linux with Wine, Mono, VNC (rarely) and my enhanced Seamless Terminal Server solution.
The latter solution refers to this:
http://img99.imageshack.us/img99/6199/olympicsfm1.png
Edited 2008-08-15 15:11 UTC
Is someone with a Nokia Web Tablet (which runs Linux and Maemo) a Linux user?
How about someone running a Linux-based Linksys router as a front-end for their Windows boxes and/or Macs?
How about someone with a Buffalo LinkStation (has Linux in firmware)?
How would you count a household with one Linux server and two Windows desktop users who access the server all the time?
Edited 2008-08-15 18:51 UTC
I am a Zenwalk/Slackware user and I am proud of it.
-2501
meaning, the only thing that the market pays attention to … are packaged HW/OS/Support purchases through an OEM such as Dell and HP, or OS/Support purchases through a vendor like Red Hat or Oracle. Those are hard numbers. Projections based on browser agent strings, addresses, and other kinds of data are extremely weak, even if you’re optimistic, because there’s no way to tell whether double- or triple- or quadruple-counting is occurring — which overlaps sales statistics. Ultimately, market interest is driven by SALES, not usage, so it almost doesn’t matter how many people are using free downloaded copies of Linux. OEMs and ISVs generally don’t care about those people, because they aren’t feeding $$$ back into the system. It’s all about the Benjamins…
Edited 2008-08-16 22:27 UTC
while the main purpose of debian’s popularity contest
http://popcon.debian.org
is to note which programs are popular, it also tracks how many submissions it gets. currently, that’s around 74 thousand.
*slurp* A one
*slurp* A two
*crunch* A three…
There are three linux users.
[sorry, could not resist, I know I should have]
If you want to “count” Linux users, a server with “irresistible” multimedia software for Linux would be a very good option.
For example, please look at these statistics of bandwidth usage from debian-multimedia.org:
http://www.debian-multimedia.org/statistics.php
This gives us a good indication how many installations of debian-based Linux installations are currently in use at the least.
Forty-two.