It's probably the closest we have though, unless you have another source in mind?

If you want statistics that make a minimum of sense, you need a representative sample. "Anyone who was willing to respond" is not representative, which is why any internet survey where anyone can respond is garbage.

To create a representative sample, you need knowledge on your population. For survey on the general population of a given country or state, the census data is used. You know what percentage of your population is in a given age bracket, their occupation, their gender, their family situation, etc. and you build a sample with the same proportions.

Thanks for explaining what a representative sample is, I think we're all a little better off now. I think the question still stands though: does anyone have a more representative survey than the Stack Overflow one?

I wonder how many of them are web devs that picked up OSX thanks to getting into the biz via some mutated media studies course.

Well, I'm a web dev who picked up OSX about 5 years ago, having used Windows in the industry for about 10 years, after I'd used Windows/Linux for 3 years before that during my Computer Science degree. So I guess I don't fit that particular stereotype, and I doubt that many people do (I've never met one).

I think it might come close, though. I mean, yes, there might be some percentage of Linux users that get all their answers from good old mailing lists, but surely the bias should not be that big?

There's a pretty big discrepancy between the popular technologies on StackOverflow and the TIOBE index (http://www.tiobe.com/tiobe_index), so I think there's a significant possibility of bias. I think there are a set of technologies for which StackOverflow is seen as a good resource and other technologies for which StackOverflow has much less useful information and that results in some bias by virtue of self-selection.

I only gave a cursory glance, but I don't think the discrepancy is that vast. The most significant difference I saw was the growth trend and position of Javascript. That said, I think you're right in that StackOverflow's demographic doesn't represent the wider marketplace - web development, particularly of the .NET and Java varieties, as well as DB development is probably over-represented on SO.

We were just talking about this in the office. We all use linux (one uses a Mac - temporary iOS developer...) and we work across the "full stack" from embedded to web.

When working on embedded, C, Postgres, or pretty much anything that is completely UIless we get most of our information from short to the point blogs, with 1/3 or less from SO.

When working on web (php, node, css/js/popular framework) or mobile we pretty much exclusively use SO. The blogs and documentation here are full of stupid terms and too much text.

The other cases I was thinking of were Python and R. Both of which I found were SO heavy, and whose blogs seemed awfully similar to the front end stuff.