2 Basic Ideas

For to be usable in different languages and topics, the internet contains some thousand various characters; only a few of them can be reached by typing. These often used characters are saved as 'key x was pressed '; the characters are decoded while reading the document.
This method is good for home-computing but doesn't work in the international Internet. It is possible to ask the browser to show a document in a distinct way - e.g.

<meta http-equiv="content-type" content="text/html; charset=ISO-8859-5">

for cyrilic. Except for some misunderstandings and the question, if the browser should ignore the user's commands when detecting a charset, this should work if characters from different topics aren't mixed:

Günther Schäfer a gagné 3½ ¥.

It isn't possible to attach this sentence to one single family of characters; the solution is to declare the suspected characters as symbols:

G&uuml;nther Sch&auml;fer a gagn&eacute; 3&frac12; &yen;.

It is not too funny to replace the letters by hand - the basic idea of Character exChange was to give this task to a machine with a single click.