Recently there has been an article floating around touting that Windows is selling better than Linux. Now, there are a multitude of ways to discredit an article these days. Do as I did with Walt Mossberg and go after what he doesn't write about. Note that crucial facts or history are missed or distorted, and credibility of the writing is removed. Another way is to go after the funding of the article or report, such as the numerous reports funded by Microsoft with pre-determined results that Microsoft Windows is more secure, faster, or whatever, than a competing Operating System.
With Windows and Linux sales though... any story reporting sales between the two starts out on a false assumption. That assumption is that Linux and Windows are distributed through selling of a physical product or a license. The problem with the assumption is that many Linux distributions are explicitly licensed so they cannot be purchased on a hardware medium. The result is that most Linux's are distributed through free to access FTP servers, or simply copied and handed out to other users. In the cases where a consumer Linux has an attached serial number or purchased license, most users will avoid the product. This is something Mepis learned first hand during it's experience with TechAlign and Frontier Linux.
The fact is, the majority of Linux's growth as an operating system has come at the hands of grass roots activists. The majority of Linux's growth continues as a grass roots activity.
This is not to say that the work of IBM, Novell, RedHat, Dell, HP, Lenovo, AMD, and other hardware and software vendors is not having an impact on the growth of Linux. Due to the work of many Independent Hardware and Software vendors Linux is found everywhere from the consumer desktop to dedicated devices like Tivo and Playstation.
The problem is, comparing sales of Windows to Linux completely ignores the grassroots factor.
Now, some arguments state that the grassroots factor isn't very large, and that most of the work of Linux is found in the server market.
I can pretty much destroy that idea with one word. Debian.
Fact is Debian is one of the largest server operating systems in use with the latest version of Debian, Etch, easily qualifying as one of the top three Linux based operating systems in use today. Yet, nobody sells Debian. Where do all of the installs come from them if Debian can't be sold? Okay, that is a rhetorical question as users obviously download and install Debian themselves, but that is an activity accountants cannot account for.
Going beyond Debian(pure) there are the huge grassroots factors of Mepis Linux and Ubuntu Linux, both which fall under the Debian tree. Again, accountants cannot explain the success of either of these two versions of Linux in terms of sales, since there is no direct sales data to be obtained. Mepis and Ubuntu are freely available to download, freely available to copy, and freely available to give to other users.
Debian isn't the end of the story once one starts to include community based editions of commercial distributions such as Fedora and OpenSuse. Neither of those OS's are sold in any retail channels, and I'm fairly confident that the Fedora following alone is probably larger than Ubuntu and Mepis put together.
The end result is that any report which purports that Windows is selling better than Linux ignores a crucial factor of Linux distributions.
The tracking situation isn't likely to change within the near or far future as it is. Many Linux developers are heavily concerned with user freedoms, and simply don't want to be tracked, or be in a position to enable tracking. Many Linux users are also in a similar position to avoid tracking measures, again referencing the deluge of hate posts generated when TechAlign issued serial numbers with the Mepis based Frontier Linux.
Does this mean that all Linux tracking data is effectively flawed? In short, yes.
Does this also mean that any sales reports comparing Linux and Windows server sales are also effectively flawed? The gut reaction is again, yes. The reasoned out answer is probably not. Most Linux and Windows servers are classed to deliver a particular power per watt or throughput amount. Comparing functionally equivalent Linux and Windows servers from one independent hardware vendor is of use for tracking sales data. Attempting to track the server sales overall? Isn't possible because of Debian.
In many ways, tracking Linux sales versus Windows sales is a lot like a synthetic benchmark. If you narrow the field enough, then comparisons are valid. But, when you try to look at the entire field? The effect falls apart.