i read an article today proclaiming that linux only has 0.36% of the desktop market based on a sample set of two million web users. in one online discussion about it some people posted links to stats on other websites that showed similarly dismal number hovering around the 1% mark. then i looked at some general interest (e.g. non-technical, non-geeky, non-specialist content) sites that i have the stats to and they are reporting 3-4% for linux. then there was the private study i saw recently that put it over the 5% mark and of course my own personal experiences of constantly running into linux desktop users ... except in airports where i never see them. what's going on here?
i recently read freakonomics at the recommendation of adriaan de groot, and reflecting on that book i decided to consider: well, what if all those numbers are accurate? how could that be and what would that say about linux desktops?
interestingly, those websites with ~1% linux visitors were web site developer sites. perhaps people who run linux aren't the same people who make web sites. or perhaps those that do make web sites on linux don't learn how to do so by reading those kinds of tutorial sites. perhaps they read books or they learned somewhere else, such as from friends, colleagues, school, etc...
but what about the two million users? i think the same thing may be at play here. the websites profiled are general interest sites. not the sort the average geek might look through perhaps. or perhaps ....
... perhaps what everyone says about linux deployment is true. no, not that it isn't happening ;) but that it isn't happening amongst home users, road warriors, etc. but that it is happening in schools, companies (and particularly single purpose systems) and government offices. that would explain the absence in the airports and the recreational web. it would also match what we're good at and what we're not.
so i wonder if what's happening is that measuring usage by web traffic tends to measure the home market where we are dismal. this would make all the numbers jive quite well.
as a side note .. i was over at t.'s tonight and we installed a web app on her suse 10.0 OSS system. she likes the suse, but it turns out that version doesn't package php pear (suse9 did) nor the gettext support in php. wtf? where is the suse of yore, the suse i loved, the suse i adored with every package? i guess the answer is that suse is not for me, or i'm not for it. one or the other, we have grown apart.
*a moment of silence*
update: so suse10oss does have php-pear packages, but t.'s yast (despite having the correct sources listed) isn't showing them. arg! will have to look into it and see why yast is breaking. note: perhaps instead of writing a gtk+ front end to yast (gotta love that NIH) perhaps those developer hours could be spent improving it? no, that'd make sense.
also, a comment on the blog brought up another good point regarding linux usage stat differential on website: it varies wildly by geography, so depending on the country/region that the website caters to the usage stats can also vary wildly. good point.
the same comment noted that in their region it's primarily due to cost being seen as the primary benefit to open source and using software against the authors will (aka "piracy") isn't seen as a bad thing.