The main idea of this support is to behave in standard way if standard means are available, and make "best guess" if not. That is, if your browser sends the Accept-Charset: header in the request, it is honored, and if not, the server tries to guess which charset you most probably prefer from the User-Agent: header. So, if your browser does not send any valid Cyrillic charset in the Accept-Charset:, be prepared to get Russian documents in windows-1251 (AKA cp1251) encoding if you are running M$ Windows, Macintosh Cyrillic if you are running OS7 on a Mac, oem866 (AKA cp866) if you are running M$ DOS, and koi8-r in all other cases.
The original package also allows support for broken clients that use fonts non-native for the platform they are running on, e.g. when koi8 fonts are installed on a Windows machine. This is done by making several virtual servers running on the same document tree, and sending documents in encoding specified via the name of the server. On www.fido7.org this is NOT done. If you are running Windows with koi8 fonts, you are out of luck. You just have these options: