> >>Suppose you made a document and sent it to me via conventional post. > >> > >>The last agent handling the document would be the mail carrier. > >>Does the mail carrier have the right to open the mailing and > >>replace your document with garbage? > >> > >> > >> > >>No, however if I receive a letter in the post written in German I'm going > to ask someone to translate it rather than try to cope with a language (c.f. > encoding) I don't understand. > >> > >> > Yes, if that's what you ask for. But as I know some German I may prefer > to do my own translation. And if the recipient is a German who knows no > English, they certainly aren't going to be amused if their letters get > translated whether they want them to be or not. So the mail carrier > should do this only if specifically asked to do so. >
Indeed. Remember the problem here isn't a server performing translation, transliteration or re-encoding - but rather a server misidentifying an encoding (hence my analogy of the translator having a nervous break-down, that and the fact that the image struck me as funny). However to enable a correctly functioning server to perform such re-encoding *when asked to do so* we have to have the rule that HTTP-headers over-ride embedded self-description for text-based formats. This causes problems in cases like those described, but not when the webserver has a rough idea of what the hell it is doing. One could argue against the rule of headers having precedence on the basis that it is brittle, but it is no more brittle than trusting copy-and-paste <meta/> elements which are also likely to be wrong (trust me I've seen enough that my anecdotal experience is approaching statistical validity). But one day it will all be Unicode... one day...
