The problem as I see it is that in order to display different character sets on the same HTML page we have no choice but to convert them all to UTF-8. This is obvious when considering the mailbox view. Imagine subjects in different encodings and languages. Once converted to UTF-8 they can all be properly displayed on the same page under a single character set. HTML only allows a web page to have one character set. So while it is feasible to enable outgoing messages to use a different character set, we cannot correctly display that on the compose page, because the interface and the contents of the textarea would be in different encodings.  The only thing I can think of would be to convert to UTF-8 as we do now while the reply is being composed, then after hitting send we convert back to the original character set.

I already assigned the bug you created for this to myself in the tracker. I appreciate the feedback. I will think about it some more and let you know what I come up with.

Take care,
Jason