new file mode 100644
--- /dev/null
+++ b/babel/messages/mofile.py
@@ -0,0 +1,100 @@
+# -*- coding: utf-8 -*-
+#
+# Copyright (C) 2007 Edgewall Software
+# All rights reserved.
+#
+# This software is licensed as described in the file COPYING, which
+# you should have received as part of this distribution. The terms
+# are also available at http://babel.edgewall.org/wiki/License.
+#
+# This software consists of voluntary contributions made by many
+# individuals. For the exact contribution history, see the revision
+# history and logs, available at http://babel.edgewall.org/log/.
+
+"""Writing of files in the ``gettext`` MO (machine object) format.
+
+:see: `The Format of MO Files
+       <http://www.gnu.org/software/gettext/manual/gettext.html#MO-Files>`_
+"""
+
+import array
+import struct
+
+def write_mo(fileobj, catalog, use_fuzzy=False):
+    """Write a catalog to the specified file-like object using the GNU MO file
+    format.
+    
+    >>> from babel.messages import Catalog
+    >>> from gettext import GNUTranslations
+    >>> from StringIO import StringIO
+    
+    >>> catalog = Catalog(locale='en_US')
+    >>> catalog.add('foo', 'Voh')
+    >>> catalog.add((u'bar', u'baz'), (u'Bahr', u'Batz'))
+    >>> catalog.add('fuz', 'Futz', flags=['fuzzy'])
+    >>> buf = StringIO()
+    
+    >>> write_mo(buf, catalog)
+    >>> buf.seek(0)
+    >>> translations = GNUTranslations(fp=buf)
+    >>> translations.ugettext('foo')
+    u'Voh'
+    >>> translations.ungettext('bar', 'baz', 1)
+    u'Bahr'
+    >>> translations.ungettext('bar', 'baz', 2)
+    u'Batz'
+    >>> translations.ugettext('fuz')
+    u'fuz'
+    
+    :param fileobj: the file-like object to write to
+    :param catalog: the `Catalog` instance
+    :param use_fuzzy: whether translations marked as "fuzzy" should be included
+                      in the output
+    """
+    messages = list(catalog)
+    if not use_fuzzy:
+        messages[1:] = [m for m in messages[1:] if not m.fuzzy]
+    messages.sort(lambda x,y: cmp(x.id, y.id))
+
+    ids = strs = ''
+    offsets = []
+
+    for message in messages:
+        # For each string, we need size and file offset.  Each string is NUL
+        # terminated; the NUL does not count into the size.
+        if message.pluralizable:
+            msgid = '\x00'.join([
+                msgid.encode(catalog.charset) for msgid in message.id
+            ])
+            msgstr = '\x00'.join([
+                msgstr.encode(catalog.charset) for msgstr in message.string
+            ])
+        else:
+            msgid = message.id.encode(catalog.charset)
+            msgstr = message.string.encode(catalog.charset)
+        offsets.append((len(ids), len(msgid), len(strs), len(msgstr)))
+        ids += msgid + '\x00'
+        strs += msgstr + '\x00'
+
+    # The header is 7 32-bit unsigned integers.  We don't use hash tables, so
+    # the keys start right after the index tables.
+    keystart = 7 * 4 + 16 * len(messages)
+    valuestart = keystart + len(ids)
+
+    # The string table first has the list of keys, then the list of values.
+    # Each entry has first the size of the string, then the file offset.
+    koffsets = []
+    voffsets = []
+    for o1, l1, o2, l2 in offsets:
+        koffsets += [l1, o1 + keystart]
+        voffsets += [l2, o2 + valuestart]
+    offsets = koffsets + voffsets
+
+    fileobj.write(struct.pack('Iiiiiii',
+        0x950412deL,                # magic
+        0,                          # version
+        len(messages),              # number of entries
+        7 * 4,                      # start of key index
+        7 * 4 + len(messages) * 8,  # start of value index
+        0, 0                        # size and offset of hash table
+    ) + array.array("i", offsets).tostring() + ids + strs)