Discrepancy – The state or quality of being discrepant; disagreement; variance; discordance; dissimilarity; contrariety.
[1913 Webster]
When it comes to web stats on operating systems, we certainly do have discrepancies for July 2011:
Source |
Share |
Value |
NetApplications
|
That Other OS – Desktop
|
93.9% |
Wikipedia
|
That Other OS – Desktop
|
78.14%
|
NetApplications
|
MSIE
|
55.97%
|
Wikipedia
|
MSIE
|
31.96%
|
NetApplications
|
iOS
|
5.09%
|
Wikipedia
|
iOS
|
5.15%
|
NetApplications
|
Android – mobile
|
14.56%
|
Wikipedia
|
Android – mobile
|
17.55%
|
We know Wikipedia’s universe, the sum total of human knowledge described in English but NetApplications hides its sample and indeed has just made the umpteenth wiggle to their data-analysis to try to keep that other OS over 90% share on the desktop…
Clearly, they are not operating in the same universe with such divergent results. It makes no sense that different sources would be so close in one measure but extremely far apart in another. In science, we would call this a “blunder” type error. In IT it appears to be deliberate manipulation to hide the weakness of M$’s cash cow of an operating system.
About Robert Pogson
I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
see Netcraft
Notice that most of the world’s web servers are not big iron and run GNU/Linux far more than that other OS. The blips of popularity for that other OS were when M$ paid people to use their stuff…
Munich migrated from NT. Where was .NET on NT? “Initial release 13 February 2002” so no .NET in Munich unless they rewrote their hundreds of custom applications for it. A lot of work for nothing if they were going to GNU/Linux.
You also miss the boat with the CALs issue, I think. .NET and java, too, use web services, whether WWW over the internet, or just LAN access with TCP/IP. You are too old school, I fear.
“Explain to me how $2.7 billion worth of GNU/Linux serves and $5.9 billion worth of servers running that other OS translates to 71% of server shipments given that servers running that other OS are much more expensive than servers running GNU/Linux. Compare apples to apples: GNU/Linux $0. That other OS $1000+$40 X N where N is the number of CALs.”
You are missing the obvious facts here, #pogson. Linux, like Windows, has been successful at replacing conventional Unix use. Linux has been most successful on large servers, including some mainframe use, because it is so Unix-like in terms of operations and administration. Windows, OTOH, has been seen as a technology replacement for Unix and has attacked the server market from the bottom, growing upwards in terms of scale. Linux servers are likely to be much more expensive than Windows servers because they are the “big iron” sort of machine, costing a lot, but numerically many less than the x86 servers commonly associated with Windows installations.
You build servers from junk box machines and install Linux and think that is typical, but it is not. Your outlook is severly limited by your lack of exposure to “real” commercial situations.
Ch not everything is right on the wikipedia.
IDC units are mostly bogus. Since they are not real Units. Lets take revenue and divide by guess price of server hardware.
http://www.idc.com/getdoc.jsp?containerId=prUS22360110 This is the IDC source document please find the 71.0 or 75.3 percent.
Guess what. The Unit numbers are bogus and IDC deleted them. Also yes both numbers did appear in different versions of that Press Release at idc.com at one point.
Yes the 75.3 or 71.0 to microsoft is what happens when you make units by dividing the revenue numbers by some made up number to get units. Because you don’t have real unit numbers. Note 71.0 was the second figure before they decide they better give up on the idea completely.
IDC edits there Press Releases by god darn stealth when they get caught out with a error ch.
Basically anyone with half a brain could work out something is wrong. In Units Windows drops Linux rises compared to revenue figures if Unit figure is correct. Reason Linux is cheaper per unit than Microsoft OEM installed servers. So more units sell for the same money. Linux Double to Triple compared to revune number when covered to Unit is normal.
Both are quite massive changes. Windows halfs in units compared to revenue normally. So drops to somewhere around the 20 percent. Linux tripples or quads. So you end up with numbers very much looking like netcraft. Of 20 vs 70 with 10 percent in other.
Now that is of course if you have real units not made up crap like IDC did.
IDC wrote this:
”
“
Explain to me how $2.7 billion worth of GNU/Linux serves and $5.9 billion worth of servers running that other OS translates to 71% of server shipments given that servers running that other OS are much more expensive than servers running GNU/Linux. Compare apples to apples: GNU/Linux $0. That other OS $1000+$40 X N where N is the number of CALs.
IDC also reports, “Demand for x86 servers continued to improve in 2Q11, with revenues growing 15.1% in the quarter to $8.4 billion worldwide as unit shipments increased 5.4% to 1.9 million servers.”
If 71% of servers shipped with that other OS, you are talking about 1.9millionX71% = 1.35 million servers and the price of a server with that other OS would have to be much less than the price of a server with GNU/Linux. Then there’s the matter of server consolidation. Who knows what loads are running in the virtual machines…
“71.0% of all quarterly server shipments.”
And that’s the number of boxes, what else.
“Is by value including OS. Also this is fully assembled servers not custom orders.”
Where do you get this strange idea ?
Sorry for repeating myself:
Last year, the WP page “Usage share of Operating Systems†showed this interesting statistics:
http://en.wikipedia.org/w/index.php?title=Usage_share_of_operating_systems&oldid=379301547#Servers
(Unfortunately, it has since been replaced by one focussing on web servers and leaving out most of the details of this version.)
So what can we learn ? Unix has a far higher share of revenues than units – think “Big Ironâ€. Linux is big on web servers in general, but on those with SSL certificates (indicating something worth protecting is running on them, eg online shops) it virtually ties with Windows. On servers in general, Windows runs on 75% of units – given that on web servers it is much less popular, it must have way more than 75% on non-web servers. If you want to know why, just work in the IT department of some big company – you will find lots of Windows servers running AD, file, print, Exchange, Sharepoint, even a lot of DB servers …
Ch IDC numbers are all revenue.
71.0% of all quarterly server shipments. Is by value including OS. Also this is fully assembled servers not custom orders.
Big issue here if I have redhat and I provide Dell with a image of redhat to install that is setup for my environment. 50 dollar charge only is added to the hardware and its marked as shipping with Linux.
So basically there is a huge miss count CH. Not even part of the Linux OS cost is being counted.
W3Techs numbers are down right rough. I know this.
Issue its really hard to get “market share by units sold”.
“BTW: On big database servers you will see quite some Solaris and other proprietary Unix, often on SPARC etc.” These are getting rare.
ch we need to be demarding that companies doing these stats start serving up some useful unit numbers.
A server running Linux costs much less than a server running that other OS usually so revenue share skew the data. Look at actual units in operation, not dollars.
e.g. A server on a LAN with 100 CALs costs $5000 more than a server running GNU/Linux. If the server hardware costs $2000, the cost ratio is $7000:$2000 = 3.5
Many servers ship with no OS because the shop will have its pet load ready to go in some disk image. In schools where I have worked, schools with servers will usually have two servers, one for firewall/cache running GNU/Linux and one for AD. Larger organizations will have databases, storage and such which are just fine on GNU/Linux, FreeBSD etc.
“I think GNU/Linux does own servers.”
Actually no – you believe that to be the case, although we have been over this. To quote your source:
” Linux servers represent 20.5% of all server revenue in the quarter”
“Quarterly revenue of $5.9 billion for Windows servers represented 45.5% of overall quarterly factory revenue and 71.0% of all quarterly server shipments.”
So somebody else is shipping 70% of boxes and makes almost half the revenue – how does it follow that Linux owns the server ? (Yes, I’m referring to all servers, not just web servers.)
BTW: On big database servers you will see quite some Solaris and other proprietary Unix, often on SPARC etc.
I think GNU/Linux does own servers. The only service that I see dominated by that other OS is holding the hands of their desktop OS. Web, HPC, multimedia rendering, database are heavy users of GNU/Linux. In smart phones, consider the rate of growth as well as share. Android/Linux has passed iPhone long ago, Symbian is dying, Phoney 7 never existed, and Android/Linux is growing rapidly on tablets. I read today that ARMed notebooks are in the wings. Folks are planning to put “8” on them but Android/Linux will be there as well. I suspect the world won’t wait on M$ much longer.
There are 100million+ active web servers out there and 60% run GNU/Linux. That other OS? 12.33%.
Ch by units shipped in Servers Linux holds at least 70 percent share. Note at least. Suspected higher. Windows is 20 percent or less. 70 percent is pretty much owned. Shops not running something Linux are becoming insanely rare.
Revenue based numbers have it the other way over.
Embedded is around 40 to 50 percent Linux depending on who numbers you are reading. Yes 40 percent is still double its nearest competitor. In embedded the only one with growing market share the last 4 years has been Linux. So that is well on the way to being owned.
More shockly 30 percent of the embedded market owns to one Linux Distribution. Wind River Linux. With the 20 percent that Wind River holds in VxWorks. They have 50 percent of the market under their thumb.
So embedded is not as divided as it first appears.
Guess who Wind River is. Yes Intel. Also guess what the biggest competitor to Linux is in the embedded market VxWorks. Also surprising is that VxWorks runs Linux binaries. So standard binary format in embedded these days are Linux Elf files.
Note for satisfaction rating in embedded that the Linux worked basically Linux is around 80 percent. 40 percent higher than its closest competition.
Supercomputers are owned by Linux without question.
Now here is the shocker. Remember more embedded devices ship every year than PCs. What is the total value of 50 percent of the embed market in revenue to a OS maker and support provider.
“Revenue US$359.7 million”
Yep less than 1 billion dollar a year. Redhat makes more money.
Now lets rationalize this. If MS gets paid at the same pay rates embedded OS makers take total income for the year would be at best 1 to 2 billion for the complete PC market and server market share MS holds. MS would fold at this level of income.
This tells you something about Linux that less than half a billion dollars is a profitable income.
Embedded is not a huge profit making area for a OS maker.
“Which cult were you talking about,”
The cult where you can say/write “Really, how long has GNU/Linux *not* found its way to the desktop when it practically owns everything else” without anyone questioning it.
(Hint: Linux doesn’t “own” servers. Android is strong on smartphones, but at ~33% doesn’t “own” that sector. And I doubt there’s any “owner” in embedded.)
Gropey the Clown. ARM shares there tech with everyone equally. Yes ARM information is some of the reason why some Windows CE devices could be converted to Linux.
Oldman external usb harddrive is not exactly hit or miss. There is a hardware bug. Annoying as hell bug.
Devices with a true USB device port external usb harddrives work no problems. Devices using the “usb on the go” interface it is a little dicey. Sometimes you have to turn the device on and off a few times for it to get out of the idea that the harddrive is a PC and treats the harddrive as a device.
Particular models external hard drives are worse. The ones that support pretending to be a PC to allow downloading from cameras and the like.
The best way is connect a usb key to the usb to go port fire phone up then connect the external hard drive after the usb key has appeared.
Yes this effect all phones with “usb on the go” including Windows Phone 7 devices. The issue is the usb port starting up in the wrong mode due to hardware making mistake and worse some of this hardware can only change mode on power on so once OS is loaded it too late todo anything.
Please don’t blame a hardware error that everyone is fighting with. On android.
“Insert an SDâ„¢ Card into the SDâ„¢ Card slot, or an external USB device via the USB port on your tabletâ€
Thats one vendor Pog. The actual USB support for USB hard drives is still hit or miss from what I can determine, even in the Android 3.2 beta, at least as it stands now.
oldman wrote, “Even though it has a native file system and USB support, android appears to still have no solid support for external hard disks, and even a 64Gb tablet becomes tight repidly these days.”
e.g. Toshiba Thrive tablet User’s Guide – page 101:
“Using the TOSHIBA File Manager, you can back up your data to an SDâ„¢ Card, or an external USB storage device.
Insert an SDâ„¢ Card into the SDâ„¢ Card slot, or an external USB device via the USB port on your tablet”
“Android/Linux is not GNU/Linux but a major component of it is the Linux kernel. Android can be used as a desktop OS on x86 and on PCs running ARM.”
And I am running a system 370 emulator running MVS 3.8 on my desktop (http://mvs380.sourceforge.net/) trying to resurrect an ancient piece of code that actually may be of use, but I would hardly call it something that is ever going to be mainstream.
BTW – Even though it has a native file system and USB support, android appears to still have no solid support for external hard disks, and even a 64Gb tablet becomes tight repidly these days.
Android/Linux is not GNU/Linux but a major component of it is the Linux kernel. Android can be used as a desktop OS on x86 and on PCs running ARM.
It does not matter what deals ARM and M$ make as long as they are not exclusive deals. If M$ bought out ARM, for instance, there would be a problem but alternative technologies would spring up. It’s not a secret how to make a cheap, fast, cool processor. The Chinese, for instance plan to be self-sufficient in IT within a few years. That means no Intel and no M$.
China is in the process of making the Dawning 6000 supercomputer with no foreign CPUs. M$ is not going to buy out China.
Probably unrelated to this post but since you guys always like to demonize Microsoft and glorify ARM, I just thought I’d like to show you this little tidbit of information:
http://www.arm.com/community/software-enablement/microsoft/index.php
“ARM and Microsoft have been collaborating for more than 14 years across a range of technologies and product categories.”
Your savior is in bed with the your devil.
:O
“Ivan lets be blunt the population of android devices have over doubled.”
Ok, I’ll be blunt. Android is not Linux nor is it a desktop operating system and it is highly suspect of FOSS fans to act like it is when it suits their purposes.
“People like me really would just love to get our hands on the real numbers.”
The ‘real’ numbers are there, you just don’t want to accept that your ‘choice’ isn’t more popular.
“Slightly circular, but that’s okay for a cult.”
Which cult were you talking about, the Microsoft cult? Or, the cult of Microsoft?
Not quite. The strength of M$’s lock-in is less when M$ has multiple OS on the go.
In the past, M$ usually had only one old OS rattling around. It was fairly easy to sell the new one based on feature-bloat and compatibility with new hardware. With two old OS rattling around, clearly, people see that the new OS is not that important and good hardware lasts a lot longer than 3 years. Too bad for monopoly. With these changes in attitude to M$ comes increasing openness to FLOSS which is much more flexible for the users/customers. People can see advantages to GNU/Linux, for instance, because the software works for them an not M$. People with 5 year old hardware can get new life out of the machines without paying for a new licence…
Twitter, this only tells that there is less time to release a version of Windows.
And now a little lesson:
You quote the part of the NetApplications statistic that deals explicitely with “Desktop” and note “discrepancies” when comparing it with the Wikimedia “overall” statistics. However, if you compare desktop to desktop, then the Wikimedia figures for OSes are:
Windows 89,4%
MacOS 8,7%
Linux 1,8%
Slightly different from NetApplications’ numbers, but nowhere near as different as you suggested.
“There is virtually no one free and clear of any credibility problem who backs up that 1% metric.”
Might that be because any source coming up with a number in the 1%-region is automatically “unfree and has a credibility problem” to you ?
But I like Twitter’s comment even better: “It is no wonder that their numbers don’t reflect reality. The real question is why sites like Wikipedia don’t.”
So you know what reality should be like, and therefore anyone who says different must be wrong. Nice logic. Slightly circular, but that’s okay for a cult.
What I find most intriguing is the fragmentation of the Windows world. According to Wikipedia, no version of Windows has more than 33% of the market. It has been this way for a while and that’s why web developers write to real standards instead of IE version whatever now. The situation in Microsoft browsers is even worse, with the largest Microsoft share at 19% for any one version. That’s lower than browsers that follow real standards, Of course, the situation is worse for Microsoft than the numbers show.
Netcraft is a Microsoft owned advertising firm more than a research firm. It is no wonder that their numbers don’t reflect reality.
The real question is why sites like Wikipedia don’t. My guess is that botnet traffic is aimed at them. This is not beneath the morals of a company like Microsoft that regularly astroturfs.
http://en.wikipedia.org/wiki/Usage_share_of_operating_systems This is a good page.
Ivan lets be blunt the population of android devices have over doubled. More like 15 times when it was being recorded at 0.5 of a percent in webstats compared to now when its reporting in webstates as somewhere between 1 to 2 percent.
So we know webstats are not worth crap basically. There is direct evidence that a lot are slipping through the cracks not counted.
There are very few good metrics. Like we know when 1/2 Million Linux systems were deployed in Brazil. But due to where it was deployed it did not register at all in the web status.
Yes there have been some quite big deskop Linux deployments that should have registered in stats if the stats were working.
Question is how far broken are the stats. Yes the very thing MS people pull out is known to be bust.
People like me really would just love to get our hands on the real numbers.
Rather than “statistical noise” it’s more likely leakage from all the rest of the sectors finding their way into the desktop. Really, how long has GNU/Linux *not* found its way to the desktop when it practically owns everything else.
No doubt there is a huge volume of noise protesting that desktop Linux is not above 1%. All of it seems to be directed from Redmond WA. There is far too much going on in the world of GNU/Linux for there to be less than 1% of the desktop. Microsoft are just taking advantage of the difficulty in measuring something that’s given away for free. There is virtually no one free and clear of any credibility problem who backs up that 1% metric.
What’s the discrepancy? Wikipedia is advertising total page hits Marketshare.com is separating mobile from desktop.
Two different stats (both leave desktop Linux as statistical noise, however).