The attached message was sent to the Gnus list, but the source of the
problem is Emacs' coding system priorities.  Emacs prefers mac-roman
over utf-8 when decoding the text given below.

I noticed that PROBLEMS in Emacs 21.3 prerelease no longer says that
Unicode support is not complete (which it did in 21.2), so I assume
the Unicode support has been completed, or at least the bugs has been
fixed.

So, shouldn't UTF-8 be preferred over mac-roman in emacs 21.3?
Doesn't even Macintosh use UTF-8 now?

Perhaps it is too late for 21.3, and if so, shouldn't we at least make
this modification to HEAD?  I verified that the same problem still
applies.  (You can reproduce it easily by simply cut'n'pasting the
characters below to a new buffer in a emacs -q and try to save it, if
your mail reader decodes this mail properly.)

(Note that I run emacs in an UTF-8 locale, so this message will be
encoded as UTF-8, even though it contains the same characters as
originally reported.)