The Goal

I'd be astonished if there hadn't been some sort of migration away from IE over the last few months, but inertia can be a powerful force: a lot of people are just plain used to IE by now and they may figure that the hassle of changing outweighs the risk of security breaches. We'll see.

Hasn't this already been done? Well, sort of. Tech news sites occasionally host polls asking which browsers their readers prefer, but I'm not sure how trustworthy these are. Are people allowed to vote more than once? Do they have philosophical reasons for voting for browsers they don't actually use? Do they publicly support Browser A while privately thinking Browser B is better in certain ways? I want to know what browsers techies actually use, not what they want other people to think they use.

My goal is not to analyze which browsers are used by computer users in general; plenty of surveys and analyses have been done (here's a pretty good one), and we all know that IE is currently (and has been for many years) the juggernaut of the browser world. Instead, I want to find out what browsers are used by serious techies who are comfortable installing any browser they want, and who understand what makes some browsers better than others.

While we're at it, let's also take a look at which operating systems they use. This takes almost no additional effort and might reveal some surprises. This issue has been controversial lately, with Google removing OS statistics from their zeitgeist page (apparently because they weren't happy about people using the statistics to derive market share figures). So that's all the more reason to do our own research.

Methodology

In January 2004 I published an article on OSNews that described a suite of small benchmarks I had designed to test file and I/O performance on nine major languages or variants. This article contained two links to different pages on my personal website: the author's biography had a link to my homepage, and a link within the text took readers to my benchmark source code. Then in March of 2004 I published a book review on Slashdot (the book was a history of the concept of infinity), which contained a link to my homepage at the end of the article.

I'm arbitrarily but conveniently defining "techie" as someone who would read an article on language performance benchmarks on OSNews or a technical/mathematical book review on Slashdot, and I'm taking the subset of readers who clicked through to my homepage (whether directly or via the source code page) as a proxy for all techies. My website logs the user-agent string, referral page, IP address, and date of each visit. To calculate the percentage of techies who use each browser type, I removed multiple log entries from the same IP address, so each user was recorded only once. Then I eliminated all log entries that were not referred from OSNews, Slashdot, or my own site (the last was intended to capture users who were coming from my source code page). Unfortunately, the MySQL installation on my host's servers was corrupted in early March, which meant that I lost two months of data. But I was left with a large enough sample size to prove useful: I ended up with 785 log entries with unique IP addresses and appropriate referral strings between 6 March 2004 and 17 August 2004.

Table of contents