When is MSIE not MSIE?

I have referenced my Norwegian weblog a couple of times already.
I am using this site as a testing ground. One of the things I wanted to test was: How many browsers that claim to be Microsoft Internet Explorer is actually MSIE?


I used “Conditional Comments”:http://msdn.microsoft.com/workshop/author/dhtml/overview/ccomment_ovw.asp to include a specific CSS file if the browser is IE. Partly to create fallback styles because MSIE cannot deal with display: table. And, as I said, partly to find MSIEs real market share on my site.
Since I am also including a common CSS file for all browsers that is also picked up by MSIE, finding the ratio of browsers that claim to be IE, and browsers that are IE is easy.
I created an excerpt of my access log, containing accesses from 2003-02-23 00:00 to 2003-02-25 16:09. The log is in all 8719 lines.
All CSS-capable browsers would have requested the file named /styles/screen.css, but only true MSIE browsers would have requested _/styles/msie-screen.css_.
All in all, 675 lines of the log contained requests for _/styles/screen.css_.
Of these 675 lines where a request for a stylesheet was made, 302 also contained MSIE as part of the User-Agent string. Which _apparently_ gives MSIE a 44,7% market share.
I mentioned the conditional comments. They are only parsed and read by a true Microsoft Internet Explorer. So, if I grep the log for GET /styles/msie-screen.css, I have the true number of MSIE requests among CSS-capable browsers.
By doing this check, the number was reduced to 245, leaving Internet Explorer at a market share of 36,3%. 18,8% of all browsers that claim to be MSIE, aren’t.
I also got some very weak indication on MSIE versions used: MSIE 6.0 in a firm lead at 92,9%, MSIE 5.5. at 4,9% and MSIE 5.0 accounted for the rest.
*Disclaimer:* This test is made from a sample from a period of only 64 hours and 9 minutes on a relatively low-traffic site. The market share should not be confused with MSIE’s general market share, and the test is here only to give an indication on how common it is for other browsers to misrepresent themselves as Internet Explorer.

Next Post

4 Comments

  1. That’s a clever way of detecting the differences. I wonder what e.g. Google uses for it’s browser statistics in their Zeitgeist ( http://www.google.com/press/zeitgeist.html ) … if your results are correct, there may be an over-representation of MSIE in the stats…

  2. I think most sites use only some form of parsing of the User-Agent string. I can’t yet seem to actually have found a site that uses CC’s in any way.
    The only thing I can tell for sure is that to some extent, MSIE will always be over-represented in stats. If the true number is 18% or if it is 1,8% I cannot tell. I will return with results when I have data spanning a longer period, and post the results again to see if my preliminary findings are right.
    I’m also kind of hoping someone else can do the same for a week or two to either confirm or dismiss my findings.

  3. Market share of Internet Explorer inflated?

    It certainly looks that way from the small tests that Arve Bersvendsen has done: When is MSIE not MSIE? It seems a lot of browsers…

  4. Thanks for this article. Although it’s older than two years I read it only today. Until now (as you certainly know; due to my other comment a few days ago 😉 ), I wasn’t aware of the problems that are created by changing the stylesheets on the basis of the user-agent string.