Mac Musings

The Rest of the WebSideStory

Daniel Knight - 2001.12.26

Last week I ripped into the conclusion by WebSideStory that Linux users were only 0.24% of all desktop computer users. Once the Linux community learned of Linux Lies on Slashdot, we had a very busy day.

I've received a lot of email about the article, some damning me for questioning WebSideStory and their HitBox service, one simply calling me clueless, one raising some good questions about our numbers, and several offering additional reasons for the low Linux number.

What HitBox Does

When a site signs up for HitBox, they get some code to insert on their home page - about 1.7K of JavaScript code. HitBox will be able to tell you how many visitors came to your page, which browsers they use, what OS they're running, their monitor resolution and bit-depth, whether they came via a search engine or other link, and much more.

Of course, browsers that don't support JavaScript won't be counted.

We've asked HitBox/MarketStats/WebSideStory how they determine the visitor's operating system, but have not yet heard back. We suspect they're using the user agent string each browser is supposed to send out. If so, that's a flawed methodology, as we'll explain below.

Negative Feedback

Dissing on Tech Savvy Users

Here's one example:

HotBox [sic], while far from being the end-all be-all of web statistic tracking, offers a good random polling of sites from all over the internet. Including Tech-only sites like your um....websites...ahem.....would only sway the results into something that they didn't want. The survey was tracking DESKTOP that they are obviously not meaning tech-saavy folk.

This is the first time anyone has asserted that "tech-saavy folk" might not be considered real desktop users. Sure, some of us use portables, but I doubt the HitBox (not HotBox) software can distinguish between them.

Dissing on "You Guys"

A really cocky reader writes:

You need to double check your facts. The stats you are complaining about are compiled using data from Hitbox, which is a service and not an application. The web designer places a bit of html and java code (both of which are platform independant) on their page which allows hitbox to compile stat data on their server. The web page owner can then access that data from any computer with a web browser.

I understand that stuff like service, html, and code is really mysterious to you guys, but you could have found out those basics within minutes at the hitbox website.

I explained to this writer that JavaScript is a programming language, so the code used by HitBox does constitute software. No, it's not an application program, but JavaScript creates real running programs.

I also wonder who he means by "you guys" - me, Mac users, Linux users, Web publishers...?

Helpful Feedback

Misidentified Clients

Several readers noted that it's not uncommon for Linux users to set their user agent string to report themselves as Internet Explorer for Windows to prevent them being blocked from some sites. One writer noted that Opera (even on Linux) identifies itself as Internet Explorer by default. Another says Mozilla does the same thing.

If HitBox counts all non-Mac versions of Internet Explorer as Windows versions (since there is no IE for Linux), and if many Linux browsers are set to identify themselves as IE so they can access "IE required" sites, that would really go a long ways toward accounting for HitBox's incredibly Linux undercount.

The Bigger Picture

Another reader noted that our numbers didn't add up to 100% when he added Linux, Windows, and Mac. True, we left out some of the other operating systems. Here's a more complete set of data for November:

 pages: %pages: OS
------: ------: -
374247: 47.79%: Macintosh
350093: 44.70%: Windows
 41888:  5.35%: OS unknown
 15692:  2.00%: Unix
  9799:  1.25%:   Linux
  3031:  0.39%:   SunOS
   785:  0.10%: WebTV
   187:  0.02%: OS/2
   137:  0.02%: BeOS
   109:  0.01%: Amiga

We don't know why "OS unknown" is the third most popular OS. It's possible a fair number of those could be Linux and other *nix users, but also Mac users with obscure browsers and who knows what else. Linux is easily the top *nix platform, with Sun a distant second.

As for WebTV, OS/2, BeOS, and Amiga, it's nice to have them all visit our Mac site.

Positive Feedback

Perhaps my favorite email in response to the article:

Saw your site for the first time (via a link from a slashdot article about linux usage, that I'm reading from a winbox @ work - go figure), and wanted to pass some positive kudos your way. I'm a Mac fan from way back, pressed into service in a Windows-dominated industry. A balanced, comprehensive resource is often too hard to find nowadays, especially when you're preferences don't run with the mainstream herd. Your article on the Linux usage numbers discrepancy was evidence of the kind of balanced reporting much-needed, plus the ready willingness to actually support claims with data (much appreciated!). I'm currently wrestling with getting my DSL installed at home, but when I get it up and running, I hope to be seeing a lot more of your site. I just wanted to send this note saying 'Thanks for doing this' because positive feedback doesn't often get back to the people who need/deserve it, so I figured I'd throw in a contribution on my part against the entropy.

Although we're always swamped with email, we do appreciate the encouraging messages.


In the end, with sites such as Google reporting 1% of their visitors using Linux (we assume that's plus or minus 0.5%, since they round to the nearest whole number) and because a lot of Linux users use browsers that pretend to be Internet Explorer on Windows, we can guess that the real share of Linux users on the Web is at least twice as high as WebSideStory reports (since less than 0.5% would round to 0% in Google's report) and quite possibly in the 1.5% to 2.0% range (since spoofing will underreport the actual Linux numbers).

The sad thing is the typical 4% to 5% figures for Mac users on the Web, since Mac users rarely need to spoof their ID. While claiming to have the better platform, our share of the broader personal computer market is pathetic. (Yes, it's also one of the reasons the Department of Justice was able to demonstrate that Microsoft is a monopoly. That may be the only good thing about the Mac OS having such a small market share).

With Windows at about 90-94% among Web users, they outnumber Mac users about 20 to 1. Linux, which is decidedly geeky (the opposite extreme from the Mac's user friendly), is probably one-fourth to one-fifth as popular as the Mac OS.

Regardless of the exact numbers, which we may never know, it speaks strongly of the herd instinct and desire to conform among the human race. Given several viable alternatives to Windows, only one person in 20 chooses to think different and run the Mac OS, Linux, BeOS, OS/2, or something even more obscure.

It's going to be a tough uphill battle building the user base for any alternative to Windows. Let's hope that OS X will make it easier for Mac and Linux users to share our common Unix heritage in creating the best pair of alternatives to the Wintel juggernaut.