> Agreed. But if they don't parse the HTML they don't know what the > content of the document is and so they have no business to mess around > with that content by re-encoding it.
There is no re-encoding! There just might be is all. There might also be a lot of other things going on, and hence a lot of headers sent, and those sending the headers have a responsibility to ensure their accuracy and those receiving them have a responsibility to read and act upon them (though not necessarily with blind trust if it could raise security issues). Anyway, browsers aren't required to examine <meta/> elements at all, though they may, it's only xml declarations that have a strong place in the list of sources of encoding information.
