Statistics, Market Share, Logic, and Why There Probably Can't Be Only One
My apologies to "The Highlander" franchise (except for "The Quickening," which I'll never mention again -- ever) for not accepting the "there can be only one" approach for every aspect of my life. And to bMighty's Matthew McKenzie, I owe a giant round of applause and thanks for stating quite eloquently something I've been thinking about for some time but have been unable to tame into a post.
McKenzie discusses the importance -- or relative lack thereof -- of Net Application's revelation last month that Linux surpassed 1% of the operating system market share. While statistics has the unique distinction of being an area of mathematics that doesn't flat-out discourage creativity in calculation, and the gathered data depends as much on where it is collected as how it is interpreted -- I feel he's right in his conclusion it's not a win, or a loss, or even terribly indicative of a trend.
Framed in the "overall marketshare" terminology, the information (or how it was gathered and calculated) isn't necessarily questionable, it's more that it's meaningless. It's nebulous, even when one looks at several months worth of data. I agree with McKenzie that Freeform Dynamics' look at how Linux is used in various business settings answers an actual question -- and the answer can be used to ask further questions, form opinions -- and maybe one day even explain to some degree what 1% of the market share really means.
Perhaps, finally, I can tame my thoughts into this sentence: A single number, or a string of numbers answering the same question over any length of time can not give as clear a picture as multiple, carefully targeted points of data collected at a single point in time.
Never mind that this single number has unleashed arguments from all angles about how this proves Linux will never achieve world domination, and how proprietary software platforms are doomed -- with one claim just as easily supported by the numbers as all the others -- it is perpetuating a myth.
Operating systems aren't immortal beings, and by rights, there can't be (there shouldn't be) only one. McKenzie reiterates Brian Proffitt's take on the Freeform Dynamics report, highlighting the idea that visual artists and graphic designers generally find using Linux less appealing (or workable) than those in IT departments, or even generalized office workers.
There can't be only one operating system, I think, for much the same reason there can't be one, or two, or three Linux distributions. Are sixty thousand forks of the same project a problem? Yes. Should there just be a single unified Linux distribution? Of course not. No one system can be everything to everyone, and no one system (however powerful, or stable) can do everything perfectly that just one person might require of it in the course of a day.
While observing trends and measuring market share are important, the results (good or bad) shouldn't be any platform's measure of self-worth or validation. It's a data point to build on (we're weak in this area, strong in this area, our platform is being used a lot more this quarter, where did all of our uses go?) in order to improve and stay relevant.
It's good to make sure the numbers used to stay relevant actually say something.