I recently read two posts throwing mud at the 1% statistic. The first didn’t offer any positive reasoning and just said the statistics had to be based on sales and that wasn’t fair. Probably link/flamebait… It worked. The second tries to stomp both sales and web statistics but falls short again. Neither give evidence to show Linux at a higher level.
To be clear, all three of us are talking about desktop and mobile Linux users. Not servers. Linux clearly (and quite rightly) dominates there.
I’m going to lead off with some points (for and against) the two major numbers used coming up with the 1% share number. I’ll finish up with some statistics of my own.
Sales data: Junk
- Only shows what operating systems went to customers, not what they use.
- Only new sales (refurbs and recycles don’t count)
- Obviously completely useless for what we’re trying to count here. True “apples and oranges”.
Web statistics: Pretty accurate
- Aggregated web-collected statistics put Linux at 1.10% (as of August 2010)
- They work by collating statistics from multiple websites. Your browser announces its user agent string every request it makes.As an example, mine is:
Mozilla/5.0 (X11; U; Linux x86_64; en-GB; rv:188.8.131.52) Gecko/20100723 Ubuntu/10.04 (lucid) Firefox/3.6.8. The operating system is extracted and counted.
- User agent strings can be altered (required for some websites) but can be detrimental on other websites so it’s not usually left in an altered state.
- Non-browsers can pretend to be browsers, by using a real browser’s user agent string. Apparently AVG does this. But if you use a client-side script to count your users, usually only browsers trigger as real visits.
- IE sometimes needs additional files. Firefox and Chrome are more lean in their requests. But this only has an effect if you count the “hits” and not the visits. I don’t.
- Only counts computers that are used to browse the web but this applies to all operating systems equally. They can all be used without the internet.
But there is a catch: demographics
As people commenting have been quick to point out, some sites (like W3Schools.com) report a relatively high Linux market share. To see why you only have to look at where the data comes from.
Its content is purely technical. It’s a learning resource and reference for web developers and programmers. A technical task demands technical people. It’s also English only which again, limits statistics drawn from there.
Compare that to something like the Wikimedia foundation which is multi-lingual, widely used by most demographics and you get a much better global statistic.
I’m a web developer. I maintain sites that cover technical, corporate and extremely non-technical users. This is worth mentioning as there are biases that have an effect on who visits the site.
Compared to something like W3Schools or Wikimedia these site barely register on the scale but they give me a direct handle on the numbers and I know who the users are.
Notes about this data: These 7 websites account for the top 92% of visitors. I use Google Analytics (a client-side solution) and I block the script here so I don’t count towards the stats. Stats are rounded. Between the sites, there are a good 200,000 unique visitors each month. They’re all English language sites.
I think it’s also important to point out that I am a Linux user and have been for years. It’s in most elements of my personal and professional life and that’s great. My point is I want the market share to be as high as possible, I’m not beating on Linux but I won’t stand by while people claim its market share is something its not based on their own feelings or poorly thought out arguments. It doesn’t do the community any service.
I’m trying to use all data with as many grains of salt as necessary to point out a realistic number. In the UK and US this appears to be around 1%. Globally it might be higher but none of the statistics (mine especially) cater specifically for foreign markets. I would be interested to see stats from a popular eastern website to compare.
There might be flaws in the data collection and collation but 1% isn’t a myth, it’s the most scientifically-based figure we have at the moment.
I’m sure in time both the figure itself and the way we can collect the data will improve but if you don’t like it, instead of telling everybody it must be 10% because it “must be”, come up with another way to poll people.
About Oli: I’m a Django and Python programmer, occasional designer, Ubuntu member, Ask Ubuntu moderator and technical blogger. I occasionally like to rant about subjects I should probably learn more about but I usually mean well.