I just found a link to here on el Reg.
Some really interesting info, gathered from about 10,000 volunteer machines. Of course there are plenty of arguments that the sample is not representative. What with them having to volunteer on-line, install some (never in a million years gonna get corporate approval) software, no non Windows OSes, etc.
So I’m thinking the best way to treat the data is to believe anything that backs up my personal views and assume a flawed sample for anything that might suggest I’m wrong. (don’t tell me you don’t do that?? ;-)). That and wonder why some of the charts add up to way more than 100%.
In reality I think the sample will be heavily biased to fairly (but not highly) tech savvy individuals.
The big surprise for me is how high the Office 2007 figures are, balanced in a way by how high the OOo proportion is. And Vista seems to have done better than I would expect. Both of these could be explained by the (assumed) sample bias to individuals.
What do you think? Any surprises in there for you?