A recent study released by Net
Applications indicates that the market share of Windows among
Internet-connected devices could drop below 90% as soon as the fourth
quarter of this year. This seems to be the same report referred to earlier that showed Linux approaching a 1% use
share.
To read all comments associated with this story, please click here.
Member since:2005-07-08
The questions are: Who paid for the study?
Probably Net Applications. They aren't a survey company.
About 40,000 websites that Net Applications are involved with. I'm not sure how unbiased the sample is either, but their statistics do fit in with the doubled share of Apple computers over the past two years, reflected in sales figures given to stockholders. (Which if you fake them, you're in for a stockholder lawsuit. Steve Jobs already has his hands full with one of those, he doesn't need another.) The increase in Linux share also is reflected in real-world data. The EEE PC, 40% of which are running Linux, is one of the hottest selling laptops now worldwide, while distributions like Ubuntu have a reputation for ease of use that Linux previously did not have.
That's irrelevant, as all browser agent strings also include the OS the browser is being run under. This is easily parsed. I doubt that the amount of non-firefox, non-safari (they have iPhone stats, so they're counting it), and non-IE browsers, are not statistically significant for the purposes of this survey; assuming they're doing any "discarding" at all, because it would be fairly easy to parse what OS is being run in a browser-independant way if you're familiar with browser agent strings. (Most have the "Mozilla" type string that dates back to Netscape Navigator, even IE or Safari which spoof it.)
Probably not many, know of many Windows Opera users? Enough to throw things off double-digit percentage-wise? (Again, Opera is a decent browser, but we're talking about OS statistics and I don't really think Windows Opera users are a significant factor, the only real factor that would affect it since it's the only one that didn't use the Mozilla string.)
That's a good question, but do you know of any 40,000 website conspiracy of Linux and Macintosh computers, set up in a botnet, to throw off a survey of hundreds of thousands of visitors? I didn't think so.
Member since:2006-11-17
It probably doesn't make a big difference for the share or Windows and OS/X (89% is close to 91%).
Only for linux, the 1% share does not seem realistic to me. Maybe in the US (they didn't even tell the language break down of their selected sites). I know many linux browsers fake Windows user agent string. Anyway, above all, I don't trust their selection of web sites. I bet they only selected english web sites or worse US sites. Where I live, linux online share is closer to 15% than 1% in my unbased opinion, but I don't pretend to know anything and certainly don't publish numbers without explaning where they come from in an open manner. The bottom line is that those statistics are worthless to me. A complete specification would add value to it.
Edited 2008-07-08 11:42 UTC
Member since:
2006-11-17
I can pull out numbers like that from my hat (or from somewhere else that is part of my body)
The questions are: Who paid for the study?
Which sites were selected as 'representative for the overall market'? How do they count the OS users? Are they using browser agent-string and if so how many agent-string do they know? How many of them don't they know and get discarded in the result? Do they filter duplicate IP/user? How?
Those numbers mean nothing to me as they are not explained. It's all opaque like if the light was switched off and they told you to go that direction as if they saw something, but they don't want to tell you what.