--- mm-util.el.orig	2004-12-06 23:20:17.000000000 +0000
+++ mm-util.el	2004-12-06 23:54:27.000000000 +0000
@@ -584,6 +584,73 @@
 		(length (memq (coding-system-base b) priorities)))
 	   t))))
 
+(defun mm-xemacs-find-mime-charset (begin end)
+  "Determine which MIME charset to use to send region as message.
+This uses the XEmacs-specific latin-unity package to better handle the case
+where identical characters from diverse ISO-8859-? character sets can be
+encoded using a single one of the corresponding coding systems. 
+
+It treats `mm-coding-system-priorities' as the list of preferred coding
+systems; a useful example setting for this list in Western Europe would be
+'(iso-8859-1 iso-8859-15 utf-8), which would default to the very standard
+Latin 1 coding system, and only move to coding systems that are less
+supported as is necessary to encode the characters that exist in the buffer.
+
+Latin Unity doesn't know about those non-ASCII Roman characters that are
+available in various East Asian character sets.  As such, its behavior if
+you have a JIS 0212 LATIN SMALL LETTER A WITH ACUTE in a buffer and it can
+otherwise be encoded as Latin 1, won't be ideal.  But this is very much a
+corner case, so don't worry about it.  "
+  (let ((systems mm-coding-system-priorities) csets psets curset chars-region)
+
+    ;; Load the Latin Unity library, if available.
+    (when (and (not (featurep 'latin-unity)) (locate-library "latin-unity"))
+      (require 'latin-unity))
+
+    ;; Now, can we use it?
+    (if (featurep 'latin-unity)
+	(progn 
+	  (assert (featurep 'xemacs) 
+		  (concat "We only expect latin-unity on XEmacs--this code "
+			  "will break on the FSF's Emacs.  "))
+	  (setq csets (latin-unity-representations-feasible-region begin end)
+		psets (latin-unity-representations-present-region begin end)
+		chars-region (delq 'ascii (charsets-in-region begin end)))
+
+	  (catch 'done
+
+	    ;; Pass back the first coding system in the preferred list that
+	    ;; can encode the whole region.
+	    (dolist (curset systems)
+	      (setq curset (latin-unity-massage-name curset 'buffer-default))
+
+	      ;; If the coding system is a universal coding system, then it
+	      ;; can certainly encode all the characters in the region. 
+	      (if (memq curset latin-unity-ucs-list)
+		  (throw 'done (list curset)))
+
+	      ;; If a coding system isn't universal, and isn't in the list
+	      ;; that latin unity knows about, we can't decide whether to
+	      ;; use it here. Leave that until later in mm-find-mime-charset
+	      ;; region function, whence we have been called.
+	      (unless (memq curset latin-unity-coding-systems)
+		(throw 'done nil))
+
+	      ;; Right, we know about this coding system, and it may
+	      ;; conceivably be able to encode all the characters in the
+	      ;; region.
+	      (if (latin-unity-maybe-remap begin end curset csets psets t)
+		  (throw 'done (list curset))))
+
+	    ;; Can't encode using anything from the
+	    ;; mm-coding-system-priorities list. Leave mm-find-mime-charset
+	    ;; to do most of the work.
+	    nil))
+
+      ;; Right, latin unity isn't available; let mm-find-charset-region 
+      ;; take its default action, which equally applies to GNU Emacs. 
+      nil)))
+
 (defun mm-find-mime-charset-region (b e &optional hack-charsets)
   "Return the MIME charsets needed to encode the region between B and E.
 nil means ASCII, a single-element list represents an appropriate MIME
@@ -625,8 +692,12 @@
 			 (setq systems nil
 			       charsets (list cs))))))
 	       charsets))
-	;; Otherwise we're not multibyte, we're XEmacs, or a single
-	;; coding system won't cover it.
+        ;; If we're XEmacs, and some coding system is appropriate,
+        ;; mm-xemacs-find-mime-charset will return an appropriate list.
+        ;; Otherwise, we'll get nil, and the next setq will get invoked.
+        (setq charsets (mm-xemacs-find-mime-charset b e))
+
+        ;; We're not multibyte, or a single coding system won't cover it.
 	(setq charsets
 	      (mm-delete-duplicates
 	       (mapcar 'mm-mime-charset