What type of site are the statistics from? A site like "Linux Today" may contain a significantly higher percentage of non-IE browsers while MSN will likely show that 98 percent of its visitors use IE. A broader survey of disparate sites would be necessary to really determine actual browser use across the Web.
Mozilla and Opera are both capable of "lying" about what type of browser they are in order to circumvent poor Web site coding that restricts access to IE and/or Netscape. Opera 6.0, by default identifies itself in the agent-header string as "IE". Mozilla can be modified to do the same. Even if you do determine that 90 percent of your visitors are using, IE you can't be certain this is true. A portion of those may be browsers masquerading as IE in order to access your site, because your Web site authors are coding specifically to a single browser or failing to comply with standards.
Market share for browsers cannot be determine by "downloads" or "sales" or "installations by default. Many Linux window managers such as Gnome and KDE install Netscape by default, but it is not necessarily used. Anecdotal though it may be, our lab in Green Bay, Wisconsin has 12 client machines set up to dual boot Windows 2000 and RedHat 7.3 with KDE. Every Windows 2000 machine installs IE and every RH 7.3 installs Netscape and Mozilla. We don't actually use these browsers. Or if we do, it is rarely. Should these count toward market share? Of course not.
