There's a nice piece of work by Charles Arthur in The Guardian today that puts a fascinating post from one of Microsoft's top PR people under the microscope. It's all well worth reading, but naturally the following numbers from the memo and Arthur's analysis were of particular interest:
24% Linux Server market share in 2005.
33% Predicted Linux Server market share for 2007 (made in 2005).
21.2% Actual Linux Server market share, Q4 2009.
Analysis: This is a really interesting one, because it is a distortion of reality that would have Steve Jobs applauding at its subtlety. You look at those numbers and think: wow, Linux servers really aren't popular. How odd, because you'll notice that you come across Linux servers all over the place: Google, Facebook (which runs F5's Big IP, which is Linux), Yahoo, Amazon, Wordpress.com (which hosts millions of blogs), Twitter... so why such a small number? (The only major site I could quickly find that runs Windows Server is eBay.)
Answer: because those "market share" figures are for Linux server licences sold. Microsoft doesn't count them - and because the market research companies can't count them - if money doesn't change hands. True, this indicates that companies selling Linux servers (principally hardware) aren't making headway against Windows Server. But what it doesn't tell you is what progress Linux is making overall on the web. For that, you need Netcraft. And that suggests that Linux has a really big market share.
There are two things that are worth noting about the figures that the Microsoft post quoted. First, that it bothered to include them alongside all the other sectors where Microsoft is under pressure. This is a clear indication that the company is feeling the heat from GNU/Linux in the server market: you don't bother defending yourself unless you're under attack.
But what's more telling is that Microsoft couldn't find any better figures to make its case. As Arthur rightly points out, the way they were gathered seriously undercounts the true usage of GNU/Linux. The fact that despite that distortion GNU/Linux still has pretty decent numbers – 21% is hardly shabby – means that the real market shares in terms of real-world deployments will be much worse for Microsoft. Hence the attack.