diff --git a/Doc/library/email.generator.rst b/Doc/library/email.generator.rst
index 930905a..954f175 100644
--- a/Doc/library/email.generator.rst
+++ b/Doc/library/email.generator.rst
@@ -22,6 +22,12 @@
 result in changes to the :class:`~email.message.Message` object as defaults are
 filled in.
 
+:class:`bytes` output can be generated using the :class:`BytesGenerator` class.
+If the message object structure contains non-ASCII bytes, this generator's
+:meth:`~BytesGenerator.flatten` method will emit the original bytes.  Parsing a
+binary message and then flattening it with :class:`BytesGenerator` should be
+idempotent for standards compliant messages.
+
 Here are the public methods of the :class:`Generator` class, imported from the
 :mod:`email.generator` module:
 
@@ -65,6 +71,13 @@
 
       Note that for subparts, no envelope header is ever printed.
 
+      Messages parsed with a Bytes parser that have a
+      :mailheader:`Content-Transfer-Encoding` of 8bit will be converted to a
+      use a 7bit Content-Transfer-Encoding.  Any other non-ASCII bytes in the
+      message structure will be converted to '?' characters.
+
+      .. versionchanged:: 3.2 added support for re-encoding 8bit message bodies.
+
    .. method:: clone(fp)
 
       Return an independent clone of this :class:`Generator` instance with the
@@ -76,11 +89,27 @@
       :class:`Generator`'s constructor.  This provides just enough file-like API
       for :class:`Generator` instances to be used in the :func:`print` function.
 
-As a convenience, see the methods :meth:`Message.as_string` and
-``str(aMessage)``, a.k.a. :meth:`Message.__str__`, which simplify the generation
-of a formatted string representation of a message object.  For more detail, see
+As a convenience, see the :class:`~email.message.Message` methods
+:meth:`~email.message.Message.as_string` and ``str(aMessage)``, a.k.a.
+:meth:`~email.message.Message.__str__`, which simplify the generation of a
+formatted string representation of a message object.  For more detail, see
 :mod:`email.message`.
 
+.. class:: BytesGenerator(outfp, mangle_from_=True, maxheaderlen=78, fmt=None)
+
+   This class has the same API as the :class:`Generator` class, except that
+   *outfp* must be a file like object that will accept :class`bytes` input to
+   its `write` method.  If the message object structure contains non-ASCII
+   bytes, this generator's :meth:`~BytesGenerator.flatten` method will produce
+   them as-is, including preserving parts with a
+   :mailheader:`Content-Transfer-Encoding` of ``8bit``.
+
+   Note that even the :meth:`write` method API is identical:  it expects
+   strings as input, and converts them to bytes by encoding them using
+   the ASCII codec.
+
+   .. versionadded:: 3.2
+
 The :mod:`email.generator` module also provides a derived class, called
 :class:`DecodedGenerator` which is like the :class:`Generator` base class,
 except that non-\ :mimetype:`text` parts are substituted with a format string
diff --git a/Doc/library/email.message.rst b/Doc/library/email.message.rst
index 9dcb2b4..dc305a7 100644
--- a/Doc/library/email.message.rst
+++ b/Doc/library/email.message.rst
@@ -111,9 +111,17 @@
       be decoded if this header's value is ``quoted-printable`` or ``base64``.
       If some other encoding is used, or :mailheader:`Content-Transfer-Encoding`
       header is missing, or if the payload has bogus base64 data, the payload is
-      returned as-is (undecoded).  If the message is a multipart and the
-      *decode* flag is ``True``, then ``None`` is returned.  The default for
-      *decode* is ``False``.
+      returned as-is (undecoded).  In all cases the returned value is binary
+      data.  If the message is a multipart and the *decode* flag is ``True``,
+      then ``None`` is returned.
+
+      When *decode* is ``False`` (the default) the body is returned as a string
+      without decoding the :mailheader:`Content-Transfer-Encoding`.  However,
+      for a :mailheader:`Content-Transfer-Encoding` of 8bit, an attempt is made
+      to decode the original bytes using the `charset` specified by the
+      :mailheader:`Content-Type` header, using the `replace` error handler.  If
+      no `charset` is specified, or if the `charset` given is not recognized by
+      the email package, the body is decoded using the default ASCII charset.
 
 
    .. method:: set_payload(payload, charset=None)
@@ -160,6 +168,10 @@
    Note that in all cases, any envelope header present in the message is not
    included in the mapping interface.
 
+   In a model generated from bytes, any header values that (in contravention
+   of the RFCs) contain non-ASCII bytes will have those bytes transformed
+   into '?' characters when the values are retrieved through this interface.
+
 
    .. method:: __len__()
 
diff --git a/Doc/library/email.parser.rst b/Doc/library/email.parser.rst
index 32f4ff1..77a0b69 100644
--- a/Doc/library/email.parser.rst
+++ b/Doc/library/email.parser.rst
@@ -80,6 +80,14 @@
       if you feed more data to a closed :class:`FeedParser`.
 
 
+.. class:: BytesFeedParser(_factory=email.message.Message)
+
+   Works exactly like :class:`FeedParser` except that the input to the
+   :meth:`~FeedParser.feed` method must be bytes and not string.
+
+   .. versionadded:: 3.2
+
+
 Parser class API
 ^^^^^^^^^^^^^^^^
 
@@ -131,7 +139,7 @@
 
       Similar to the :meth:`parse` method, except it takes a string object
       instead of a file-like object.  Calling this method on a string is exactly
-      equivalent to wrapping *text* in a :class:`StringIO` instance first and
+      equivalent to wrapping *text* in a :class:`~io.StringIO` instance first and
       calling :meth:`parse`.
 
       Optional *headersonly* is a flag specifying whether to stop parsing after
@@ -139,25 +147,78 @@
       the entire contents of the file.
 
 
+.. class:: BytesParser(_class=email.message.Message, strict=None)
+
+   This class is exactly parallel to :class:`Parser`, but handles bytes input.
+   The *_class* and *strict* arguments are interpreted in the same way as for
+   the :class:`Parser` constructor.  *strict* is supported only to make porting
+   code easier; it is deprecated.
+
+   .. method:: parse(fp, headeronly=False)
+
+      Read all the data from the binary file-like object *fp*, parse the
+      resulting bytes, and return the message object.  *fp* must support
+      both the :meth:`readline` and the :meth:`read` methods on file-like
+      objects.
+
+      The bytes contained in *fp* must be formatted as a block of :rfc:`2822`
+      style headers and header continuation lines, optionally preceded by a
+      envelope header.  The header block is terminated either by the end of the
+      data or by a blank line.  Following the header block is the body of the
+      message (which may contain MIME-encoded subparts, including subparts
+      with a :mailheader:`Content-Transfer-Encoding` of ``8bit``.
+
+      Optional *headersonly* is a flag specifying whether to stop parsing after
+      reading the headers or not.  The default is ``False``, meaning it parses
+      the entire contents of the file.
+
+   .. method:: parsebytes(bytes, headersonly=False)
+
+      Similar to the :meth:`parse` method, except it takes a byte string object
+      instead of a file-like object.  Calling this method on a byte string is
+      exactly equivalent to wrapping *text* in a :class:`~io.BytesIO` instance
+      first and calling :meth:`parse`.
+
+      Optional *headersonly* is as with the :meth:`parse` method.
+
+   .. versionadded:: 3.2
+
+
 Since creating a message object structure from a string or a file object is such
-a common task, two functions are provided as a convenience.  They are available
+a common task, four functions are provided as a convenience.  They are available
 in the top-level :mod:`email` package namespace.
 
 .. currentmodule:: email
 
-.. function:: message_from_string(s[, _class][, strict])
+.. function:: message_from_string(s, _class=email.message.Message, strict=None)
 
    Return a message object structure from a string.  This is exactly equivalent to
    ``Parser().parsestr(s)``.  Optional *_class* and *strict* are interpreted as
    with the :class:`Parser` class constructor.
 
+.. function:: message_from_bytes(s, _class=email.message.Message, strict=None)
 
-.. function:: message_from_file(fp[, _class][, strict])
+   Return a message object structure from a byte string.  This is exactly
+   equivalent to ``BytesParser().parsebytes(s)``.  Optional *_class* and
+   *strict* are interpreted as with the :class:`Parser` class constructor.
+
+   .. versionadded:: 3.2
+
+.. function:: message_from_file(fp, _class=email.message.Message, strict=None)
 
    Return a message object structure tree from an open :term:`file object`.
    This is exactly equivalent to ``Parser().parse(fp)``.  Optional *_class*
    and *strict* are interpreted as with the :class:`Parser` class constructor.
 
+.. function:: message_from_binary_file(fp, _class=email.message.Message, strict=None)
+
+   Return a message object structure tree from an open binary :term:`file
+   object`.  This is exactly equivalent to ``BytesParser().parse(fp)``.
+   Optional *_class* and *strict* are interpreted as with the :class:`Parser`
+   class constructor.
+
+   .. versionadded:: 3.2
+
 Here's an example of how you might use this at an interactive Python prompt::
 
    >>> import email
diff --git a/Doc/library/email.rst b/Doc/library/email.rst
index d3f1908..8926ae4 100644
--- a/Doc/library/email.rst
+++ b/Doc/library/email.rst
@@ -6,7 +6,7 @@
               email messages, including MIME documents.
 .. moduleauthor:: Barry A. Warsaw <barry@python.org>
 .. sectionauthor:: Barry A. Warsaw <barry@python.org>
-.. Copyright (C) 2001-2007 Python Software Foundation
+.. Copyright (C) 2001-2010 Python Software Foundation
 
 
 The :mod:`email` package is a library for managing email messages, including
@@ -92,6 +92,44 @@
 +---------------+------------------------------+-----------------------+
 | :const:`4.0`  | Python 2.5                   | Python 2.3 to 2.5     |
 +---------------+------------------------------+-----------------------+
+| :const:`5.0`  | Python 3.0 and Python 3.1    | Python 3.0 to 3.2     |
++---------------+------------------------------+-----------------------+
+| :const:`5.1`  | Python 3.2                   | Python 3.0 to 3.2     |
++---------------+------------------------------+-----------------------+
+
+Here are the major differences between :mod:`email` version 5.1 and
+version 5.0:
+
+* It is once again possible to parse messages containing non-ASCII bytes,
+  and to reproduce such messages if the data containing the non-ASCII
+  bytes is not modified.
+
+* New functions :func:`message_from_bytes` and :func:`message_from_binary_file`,
+  and new classes :class:`~email.parser.BytesFeedParser` and
+  :class:`~email.parser.BytesParser` allow binary message data to be parsed
+  into model objects.
+
+* Given bytes input to the model, :meth:`~email.message.Message.get_payload`
+  will by default decode a message body that has a
+  :mailheader:`Content-Transfer-Encoding` of `8bit` using the charset specified
+  in the MIME headers and return the resulting string.
+
+* Given bytes input to the model, :class:`~email.generator.Generator` will
+  convert message bodies that have a :mailheader:`Content-Transfer-Encoding` of
+  8bit to instead have a 7bit Content-Transfer-Encoding.
+
+* New function :class:`~email.generator.BytesGenerator` produces bytes
+  as output, preserving any unchanged non-ASCII data that was
+  present in the input used to build the model, including message bodies
+  with a :mailheader:`Content-Transfer-Encoding` of 8bit.
+
+Here are the major differences between :mod:`email` version 5.0 and version 4:
+
+* All operations are on unicode strings.  Text inputs must be strings,
+  text outputs are strings.  Outputs are limited to the ASCII character
+  set and so can be encoded to ASCII for transmission.  Inputs are also
+  limited to ASCII; this is an acknowledged limitation of email 5.0 and
+  means it can only be used to parse email that is 7bit clean.
 
 Here are the major differences between :mod:`email` version 4 and version 3:
 
diff --git a/Lib/email/__init__.py b/Lib/email/__init__.py
index 8702212..c54a2c7 100644
--- a/Lib/email/__init__.py
+++ b/Lib/email/__init__.py
@@ -4,7 +4,7 @@
 
 """A package for parsing, handling, and generating email messages."""
 
-__version__ = '5.0.0'
+__version__ = '5.1.0'
 
 __all__ = [
     'base64mime',
@@ -16,7 +16,9 @@
     'iterators',
     'message',
     'message_from_file',
+    'message_from_binary_file',
     'message_from_string',
+    'message_from_bytes',
     'mime',
     'parser',
     'quoprimime',
@@ -36,6 +38,13 @@
     from email.parser import Parser
     return Parser(*args, **kws).parsestr(s)
 
+def message_from_bytes(s, *args, **kws):
+    """Parse a bytes string into a Message object model.
+
+    Optional _class and strict are passed to the Parser constructor.
+    """
+    from email.parser import BytesParser
+    return BytesParser(*args, **kws).parsebytes(s)
 
 def message_from_file(fp, *args, **kws):
     """Read a file and parse its contents into a Message object model.
@@ -44,3 +53,11 @@
     """
     from email.parser import Parser
     return Parser(*args, **kws).parse(fp)
+
+def message_from_binary_file(fp, *args, **kws):
+    """Read a binary file and parse its contents into a Message object model.
+
+    Optional _class and strict are passed to the Parser constructor.
+    """
+    from email.parser import Parser
+    return BytesParser(*args, **kws).parse(fp)
diff --git a/Lib/email/feedparser.py b/Lib/email/feedparser.py
index 8db70b3..de8750d 100644
--- a/Lib/email/feedparser.py
+++ b/Lib/email/feedparser.py
@@ -482,3 +482,10 @@
         if lastheader:
             # XXX reconsider the joining of folded lines
             self._cur[lastheader] = EMPTYSTRING.join(lastvalue).rstrip('\r\n')
+
+
+class BytesFeedParser(FeedParser):
+    """Like FeedParser, but feed accepts bytes."""
+
+    def feed(self, data):
+        super().feed(data.decode('ascii', 'surrogateescape'))
diff --git a/Lib/email/generator.py b/Lib/email/generator.py
index e05b67d..40b95c4 100644
--- a/Lib/email/generator.py
+++ b/Lib/email/generator.py
@@ -12,8 +12,9 @@
 import random
 import warnings
 
-from io import StringIO
+from io import StringIO, BytesIO
 from email.header import Header
+from email.message import _has_surrogates
 
 UNDERSCORE = '_'
 NL = '\n'
@@ -72,7 +73,7 @@
             ufrom = msg.get_unixfrom()
             if not ufrom:
                 ufrom = 'From nobody ' + time.ctime(time.time())
-            print(ufrom, file=self._fp)
+            self.write(ufrom + NL)
         self._write(msg)
 
     def clone(self, fp):
@@ -83,6 +84,29 @@
     # Protected interface - undocumented ;/
     #
 
+    # Note that we use 'self.write' when what we are writing is coming from
+    # the source, and self._fp.write when what we are writing is coming from a
+    # buffer (because the Bytes subclass has already had a chance to transform
+    # the data in its write method in that case).  This is an entirely
+    # pragmatic split determined by experiment; we could be more general by
+    # always using write and having the Bytes subclass write method detect when
+    # it has already transformed the input; but, since this whole thing is a
+    # hack anyway this seems good enough.
+
+    # We use these class constants when we need to manipulate data that has
+    # already been written to a buffer (ex: constructing a re to check the
+    # boundary), and the module level NL constant when adding new output to a
+    # buffer via self.write, because 'write' always takes strings.
+    # Having write always take strings makes the code simpler, but there are
+    # a few occasions when we need to write previously created data back
+    # to the buffer or to a new buffer; for those cases we use self._fp.write.
+    _NL = NL
+    _EMPTY = ''
+
+    def _new_buffer(self):
+        # BytesGenerator overrides this to return BytesIO.
+        return StringIO()
+
     def _write(self, msg):
         # We can't write the headers yet because of the following scenario:
         # say a multipart message includes the boundary string somewhere in
@@ -91,13 +115,13 @@
         # parameter.
         #
         # The way we do this, so as to make the _handle_*() methods simpler,
-        # is to cache any subpart writes into a StringIO.  The we write the
-        # headers and the StringIO contents.  That way, subpart handlers can
+        # is to cache any subpart writes into a buffer.  The we write the
+        # headers and the buffer contents.  That way, subpart handlers can
         # Do The Right Thing, and can still modify the Content-Type: header if
         # necessary.
         oldfp = self._fp
         try:
-            self._fp = sfp = StringIO()
+            self._fp = sfp = self._new_buffer()
             self._dispatch(msg)
         finally:
             self._fp = oldfp
@@ -132,16 +156,16 @@
 
     def _write_headers(self, msg):
         for h, v in msg.items():
-            print('%s:' % h, end=' ', file=self._fp)
+            self.write('%s: ' % h)
             if isinstance(v, Header):
-                print(v.encode(maxlinelen=self._maxheaderlen), file=self._fp)
+                self.write(v.encode(maxlinelen=self._maxheaderlen)+NL)
             else:
                 # Header's got lots of smarts, so use it.
                 header = Header(v, maxlinelen=self._maxheaderlen,
                                 header_name=h)
-                print(header.encode(), file=self._fp)
+                self.write(header.encode()+NL)
         # A blank line always separates headers from body
-        print(file=self._fp)
+        self.write(NL)
 
     #
     # Handlers for writing types and subtypes
@@ -153,9 +177,15 @@
             return
         if not isinstance(payload, str):
             raise TypeError('string payload expected: %s' % type(payload))
+        if _has_surrogates(msg._payload):
+            charset = msg.get_param('charset')
+            if charset is not None:
+                del msg['content-transfer-encoding']
+                msg.set_payload(payload, charset)
+                payload = msg.get_payload()
         if self._mangle_from_:
             payload = fcre.sub('>From ', payload)
-        self._fp.write(payload)
+        self.write(payload)
 
     # Default body handler
     _writeBody = _handle_text
@@ -170,21 +200,21 @@
             subparts = []
         elif isinstance(subparts, str):
             # e.g. a non-strict parse of a message with no starting boundary.
-            self._fp.write(subparts)
+            self.write(subparts)
             return
         elif not isinstance(subparts, list):
             # Scalar payload
             subparts = [subparts]
         for part in subparts:
-            s = StringIO()
+            s = self._new_buffer()
             g = self.clone(s)
             g.flatten(part, unixfrom=False)
             msgtexts.append(s.getvalue())
         # Now make sure the boundary we've selected doesn't appear in any of
         # the message texts.
-        alltext = NL.join(msgtexts)
+        alltext = self._NL.join(msgtexts)
         # BAW: What about boundaries that are wrapped in double-quotes?
-        boundary = msg.get_boundary(failobj=_make_boundary(alltext))
+        boundary = msg.get_boundary(failobj=self._make_boundary(alltext))
         # If we had to calculate a new boundary because the body text
         # contained that string, set the new boundary.  We don't do it
         # unconditionally because, while set_boundary() preserves order, it
@@ -195,9 +225,9 @@
             msg.set_boundary(boundary)
         # If there's a preamble, write it out, with a trailing CRLF
         if msg.preamble is not None:
-            print(msg.preamble, file=self._fp)
+            self.write(msg.preamble + NL)
         # dash-boundary transport-padding CRLF
-        print('--' + boundary, file=self._fp)
+        self.write('--' + boundary + NL)
         # body-part
         if msgtexts:
             self._fp.write(msgtexts.pop(0))
@@ -206,14 +236,14 @@
         # --> CRLF body-part
         for body_part in msgtexts:
             # delimiter transport-padding CRLF
-            print('\n--' + boundary, file=self._fp)
+            self.write('\n--' + boundary + NL)
             # body-part
             self._fp.write(body_part)
         # close-delimiter transport-padding
-        self._fp.write('\n--' + boundary + '--')
+        self.write('\n--' + boundary + '--')
         if msg.epilogue is not None:
-            print(file=self._fp)
-            self._fp.write(msg.epilogue)
+            self.write(NL)
+            self.write(msg.epilogue)
 
     def _handle_multipart_signed(self, msg):
         # The contents of signed parts has to stay unmodified in order to keep
@@ -232,23 +262,23 @@
         # block and the boundary.  Sigh.
         blocks = []
         for part in msg.get_payload():
-            s = StringIO()
+            s = self._new_buffer()
             g = self.clone(s)
             g.flatten(part, unixfrom=False)
             text = s.getvalue()
-            lines = text.split('\n')
+            lines = text.split(self._NL)
             # Strip off the unnecessary trailing empty line
-            if lines and lines[-1] == '':
-                blocks.append(NL.join(lines[:-1]))
+            if lines and lines[-1] == self._EMPTY:
+                blocks.append(self._NL.join(lines[:-1]))
             else:
                 blocks.append(text)
         # Now join all the blocks with an empty line.  This has the lovely
         # effect of separating each block with an empty line, but not adding
         # an extra one after the last one.
-        self._fp.write(NL.join(blocks))
+        self._fp.write(self._NL.join(blocks))
 
     def _handle_message(self, msg):
-        s = StringIO()
+        s = self._new_buffer()
         g = self.clone(s)
         # The payload of a message/rfc822 part should be a multipart sequence
         # of length 1.  The zeroth element of the list should be the Message
@@ -265,6 +295,90 @@
             payload = s.getvalue()
         self._fp.write(payload)
 
+    # This used to be a module level function; we use a classmethod for this
+    # and _compile_re so we can continue to provide the module level function
+    # for backward compatibility by doing
+    #   _make_boudary = Generator._make_boundary
+    # at the end of the module.  It *is* internal, so we could drop that...
+    @classmethod
+    def _make_boundary(cls, text=None):
+        # Craft a random boundary.  If text is given, ensure that the chosen
+        # boundary doesn't appear in the text.
+        token = random.randrange(sys.maxsize)
+        boundary = ('=' * 15) + (_fmt % token) + '=='
+        if text is None:
+            return boundary
+        b = boundary
+        counter = 0
+        while True:
+            cre = cls._compile_re('^--' + re.escape(b) + '(--)?$', re.MULTILINE)
+            if not cre.search(text):
+                break
+            b = boundary + '.' + str(counter)
+            counter += 1
+        return b
+
+    @classmethod
+    def _compile_re(cls, s, flags):
+        return re.compile(s, flags)
+
+
+class BytesGenerator(Generator):
+    """Generates a bytes version of a Message object tree.
+
+    Functionally identical to the base Generator except that the output is
+    bytes and not string.  When surrogates were used in the input to encode
+    bytes, these are decoded back to bytes for output.
+
+    The outfp object must accept bytes in its write method.
+    """
+
+    # Bytes versions of these constants for use in manipulating data from
+    # the BytesIO buffer.
+    _NL = NL.encode('ascii')
+    _EMPTY = b''
+
+    def write(self, s):
+        self._fp.write(s.encode('ascii', 'surrogateescape'))
+
+    def _new_buffer(self):
+        return BytesIO()
+
+    def _write_headers(self, msg):
+        # This is almost the same as the string version, except for handling
+        # strings with 8bit bytes.
+        for h, v in msg._headers:
+            self.write('%s: ' % h)
+            if isinstance(v, Header):
+                self.write(v.encode(maxlinelen=self._maxheaderlen)+NL)
+            elif _has_surrogates(v):
+                # If we have raw 8bit data in a byte string, we have no idea
+                # what the encoding is.  There is no safe way to split this
+                # string.  If it's ascii-subset, then we could do a normal
+                # ascii split, but if it's multibyte then we could break the
+                # string.  There's no way to know so the least harm seems to
+                # be to not split the string and risk it being too long.
+                self.write(v+NL)
+            else:
+                # Header's got lots of smarts and this string is safe...
+                header = Header(v, maxlinelen=self._maxheaderlen,
+                                header_name=h)
+                self.write(header.encode()+NL)
+        # A blank line always separates headers from body
+        self.write(NL)
+
+    def _handle_text(self, msg):
+        # If the string has surrogates the original source was bytes, so
+        # just write it back out.
+        if _has_surrogates(msg._payload):
+            self.write(msg._payload)
+        else:
+            super(BytesGenerator,self)._handle_text(msg)
+
+    @classmethod
+    def _compile_re(cls, s, flags):
+        return re.compile(s.encode('ascii'), flags)
+
 
 
 _FMT = '[Non-text (%(type)s) part of message omitted, filename %(filename)s]'
@@ -325,23 +439,9 @@
 
 
 
-# Helper
+# Helper used by Generator._make_boundary
 _width = len(repr(sys.maxsize-1))
 _fmt = '%%0%dd' % _width
 
-def _make_boundary(text=None):
-    # Craft a random boundary.  If text is given, ensure that the chosen
-    # boundary doesn't appear in the text.
-    token = random.randrange(sys.maxsize)
-    boundary = ('=' * 15) + (_fmt % token) + '=='
-    if text is None:
-        return boundary
-    b = boundary
-    counter = 0
-    while True:
-        cre = re.compile('^--' + re.escape(b) + '(--)?$', re.MULTILINE)
-        if not cre.search(text):
-            break
-        b = boundary + '.' + str(counter)
-        counter += 1
-    return b
+# Backward compatibility
+_make_boundary = Generator._make_boundary
diff --git a/Lib/email/message.py b/Lib/email/message.py
index 923b26c..a835ce2 100644
--- a/Lib/email/message.py
+++ b/Lib/email/message.py
@@ -24,8 +24,26 @@
 # existence of which force quoting of the parameter value.
 tspecials = re.compile(r'[ \(\)<>@,;:\\"/\[\]\?=]')
 
+# How to figure out if we are processing strings that come from a byte
+# source with undecodable characters.
+_has_surrogates = re.compile(
+    '([^\ud800-\udbff]|\A)[\udc00-\udfff]([^\udc00-\udfff]|\Z)').search
+
 
 # Helper functions
+def _sanitize_surrogates(value):
+    # If the value contains surrogates, re-decode and replace the original
+    # non-ascii bytes with '?'s.  Used to sanitize header values before letting
+    # them escape as strings.
+    if not isinstance(value, str):
+        # Header object
+        return value
+    if _has_surrogates(value):
+        original_bytes = value.encode('ascii', 'surrogateescape')
+        return original_bytes.decode('ascii', 'replace').replace('\ufffd', '?')
+    else:
+        return value
+
 def _splitparam(param):
     # Split header parameters.  BAW: this may be too simple.  It isn't
     # strictly RFC 2045 (section 5.1) compliant, but it catches most headers
@@ -184,44 +202,72 @@
         If the message is a multipart and the decode flag is True, then None
         is returned.
         """
-        if i is None:
-            payload = self._payload
-        elif not isinstance(self._payload, list):
+        # Here is the logic table for this code, based on the email5.0.0 code:
+        #   i     decode  is_multipart  result
+        # ------  ------  ------------  ------------------------------
+        #  None   True    True          None
+        #   i     True    True          None
+        #  None   False   True          _payload (a list)
+        #   i     False   True          _payload element i (a Message)
+        #   i     False   False         error (not a list)
+        #   i     True    False         error (not a list)
+        #  None   False   False         _payload
+        #  None   True    False         _payload decoded (bytes)
+        # Note that Barry planned to factor out the 'decode' case, but that
+        # isn't so easy now that we handle the 8 bit data, which needs to be
+        # converted in both the decode and non-decode path.
+        if self.is_multipart():
+            if decode:
+                return None
+            if i is None:
+                return self._payload
+            else:
+                return self._payload[i]
+        # For backward compatibility, Use isinstance and this error message
+        # instead of the more logical is_multipart test.
+        if i is not None and not isinstance(self._payload, list):
             raise TypeError('Expected list, got %s' % type(self._payload))
-        else:
-            payload = self._payload[i]
+        payload = self._payload
+        cte = self.get('content-transfer-encoding', '').lower()
+        # payload can be bytes here, (I wonder if that is actually a bug?)
+        if isinstance(payload, str):
+            if _has_surrogates(payload):
+                bpayload = payload.encode('ascii', 'surrogateescape')
+                if not decode:
+                    try:
+                        payload = bpayload.decode(self.get_param('charset', 'ascii'), 'replace')
+                    except LookupError:
+                        payload = bpayload.decode('ascii', 'replace')
+            elif decode:
+                try:
+                    bpayload = payload.encode('ascii')
+                except UnicodeError:
+                    # This won't happen for RFC compliant messages (messages
+                    # containing only ASCII codepoints in the unicode input).
+                    # If it does happen, turn the string into bytes in a way
+                    # guaranteed not to fail.
+                    bpayload = payload.encode('raw-unicode-escape')
         if not decode:
             return payload
-        # Decoded payloads always return bytes.  XXX split this part out into
-        # a new method called .get_decoded_payload().
-        if self.is_multipart():
-            return None
-        cte = self.get('content-transfer-encoding', '').lower()
         if cte == 'quoted-printable':
-            if isinstance(payload, str):
-                payload = payload.encode('ascii')
-            return utils._qdecode(payload)
+            return utils._qdecode(bpayload)
         elif cte == 'base64':
             try:
-                if isinstance(payload, str):
-                    payload = payload.encode('ascii')
-                return base64.b64decode(payload)
+                return base64.b64decode(bpayload)
             except binascii.Error:
                 # Incorrect padding
-                pass
+                return bpayload
         elif cte in ('x-uuencode', 'uuencode', 'uue', 'x-uue'):
-            in_file = BytesIO(payload.encode('ascii'))
+            in_file = BytesIO(bpayload)
             out_file = BytesIO()
             try:
                 uu.decode(in_file, out_file, quiet=True)
                 return out_file.getvalue()
             except uu.Error:
                 # Some decoding problem
-                pass
-        # Is there a better way to do this?  We can't use the bytes
-        # constructor.
+                return bpayload
         if isinstance(payload, str):
-            return payload.encode('raw-unicode-escape')
+            return bpayload
         return payload
 
     def set_payload(self, payload, charset=None):
@@ -340,7 +386,7 @@
         Any fields deleted and re-inserted are always appended to the header
         list.
         """
-        return [v for k, v in self._headers]
+        return [_sanitize_surrogates(v) for k, v in self._headers]
 
     def items(self):
         """Get all the message's header fields and values.
@@ -350,7 +396,7 @@
         Any fields deleted and re-inserted are always appended to the header
         list.
         """
-        return self._headers[:]
+        return [(k, _sanitize_surrogates(v)) for k, v in self._headers]
 
     def get(self, name, failobj=None):
         """Get a header value.
@@ -361,7 +407,7 @@
         name = name.lower()
         for k, v in self._headers:
             if k.lower() == name:
-                return v
+                return _sanitize_surrogates(v)
         return failobj
 
     #
@@ -381,7 +427,7 @@
         name = name.lower()
         for k, v in self._headers:
             if k.lower() == name:
-                values.append(v)
+                values.append(_sanitize_surrogates(v))
         if not values:
             return failobj
         return values
diff --git a/Lib/email/parser.py b/Lib/email/parser.py
index 06014e2..b83e0f7 100644
--- a/Lib/email/parser.py
+++ b/Lib/email/parser.py
@@ -7,7 +7,7 @@
 __all__ = ['Parser', 'HeaderParser']
 
 import warnings
-from io import StringIO
+from io import StringIO, TextIOWrapper
 
 from email.feedparser import FeedParser
 from email.message import Message
@@ -89,3 +89,47 @@
 
     def parsestr(self, text, headersonly=True):
         return Parser.parsestr(self, text, True)
+
+
+class BytesParser:
+
+    def __init__(self, *args, **kw):
+        """Parser of binary RFC 2822 and MIME email messages.
+
+        Creates an in-memory object tree representing the email message, which
+        can then be manipulated and turned over to a Generator to return the
+        textual representation of the message.
+
+        The input must be formatted as a block of RFC 2822 headers and header
+        continuation lines, optionally preceeded by a `Unix-from' header.  The
+        header block is terminated either by the end of the input or by a
+        blank line.
+
+        _class is the class to instantiate for new message objects when they
+        must be created.  This class must have a constructor that can take
+        zero arguments.  Default is Message.Message.
+        """
+        self.parser = Parser(*args, **kw)
+
+    def parse(self, fp, headersonly=False):
+        """Create a message structure from the data in a binary file.
+
+        Reads all the data from the file and returns the root of the message
+        structure.  Optional headersonly is a flag specifying whether to stop
+        parsing after reading the headers or not.  The default is False,
+        meaning it parses the entire contents of the file.
+        """
+        fp = TextIOWrapper(fp, encoding='ascii', errors='surrogateescape')
+        return self.parser.parse(fp, headersonly)
+
+
+    def parsebytes(self, text, headersonly=False):
+        """Create a message structure from a byte string.
+
+        Returns the root of the message structure.  Optional headersonly is a
+        flag specifying whether to stop parsing after reading the headers or
+        not.  The default is False, meaning it parses the entire contents of
+        the file.
+        """
+        text = text.decode('ASCII', errors='surrogateescape')
+        return self.parser.parsestr(text, headersonly)
diff --git a/Lib/email/test/test_email.py b/Lib/email/test/test_email.py
index 8f95125..e5e51c6 100644
--- a/Lib/email/test/test_email.py
+++ b/Lib/email/test/test_email.py
@@ -9,8 +9,9 @@
 import difflib
 import unittest
 import warnings
+import textwrap
 
-from io import StringIO
+from io import StringIO, BytesIO
 from itertools import chain
 
 import email
@@ -34,7 +35,7 @@
 from email import base64mime
 from email import quoprimime
 
-from test.support import findfile, run_unittest
+from test.support import findfile, run_unittest, unlink
 from email.test import __file__ as landmark
 
 
@@ -2070,6 +2071,10 @@
         msg, text = self._msgobj('msg_36.txt')
         self._idempotent(msg, text)
 
+    def test_message_signed_idempotent(self):
+        msg, text = self._msgobj('msg_45.txt')
+        self._idempotent(msg, text)
+
     def test_content_type(self):
         eq = self.assertEquals
         unless = self.assertTrue
@@ -2186,7 +2191,8 @@
         all.sort()
         self.assertEqual(all, [
             'base64mime', 'charset', 'encoders', 'errors', 'generator',
-            'header', 'iterators', 'message', 'message_from_file',
+            'header', 'iterators', 'message', 'message_from_binary_file',
+            'message_from_bytes', 'message_from_file',
             'message_from_string', 'mime', 'parser',
             'quoprimime', 'utils',
             ])
@@ -2687,6 +2693,266 @@
         self.assertTrue(msg.get_payload(0).get_payload().endswith('\r\n'))
 
 
+class Test8BitBytesHandling(unittest.TestCase):
+    # In Python3 all input is string, but that doesn't work if the actual input
+    # uses an 8bit transfer encoding.  To hack around that, in email 5.1 we
+    # decode byte streams using the surrogateescape error handler, and
+    # reconvert to binary at appropriate places if we detect surrogates.  This
+    # doesn't allow us to transform headers with 8bit bytes (they get munged),
+    # but it does allow us to parse and preserve them, and to decode body
+    # parts that use an 8bit CTE.
+
+    bodytest_msg = textwrap.dedent("""\
+        From: foo@bar.com
+        To: baz
+        Mime-Version: 1.0
+        Content-Type: text/plain; charset={charset}
+        Content-Transfer-Encoding: {cte}
+
+        {bodyline}
+        """)
+
+    def test_known_8bit_CTE(self):
+        m = self.bodytest_msg.format(charset='utf-8',
+                                     cte='8bit',
+                                     bodyline='pöstal').encode('utf-8')
+        msg = email.message_from_bytes(m)
+        self.assertEqual(msg.get_payload(), "pöstal\n")
+        self.assertEqual(msg.get_payload(decode=True),
+                         "pöstal\n".encode('utf-8'))
+
+    def test_unknown_8bit_CTE(self):
+        m = self.bodytest_msg.format(charset='notavalidcharset',
+                                     cte='8bit',
+                                     bodyline='pöstal').encode('utf-8')
+        msg = email.message_from_bytes(m)
+        self.assertEqual(msg.get_payload(), "p��stal\n")
+        self.assertEqual(msg.get_payload(decode=True),
+                         "pöstal\n".encode('utf-8'))
+
+    def test_8bit_in_quopri_body(self):
+        # This is non-RFC compliant data...without 'decode' the library code
+        # decodes the body using the charset from the headers, and because the
+        # source byte really is utf-8 this works.  This is likely to fail
+        # against real dirty data (ie: produce mojibake), but the data is
+        # invalid anyway so it is as good a guess as any.  But this means that
+        # this test just confirms the current behavior; that behavior is not
+        # necessarily the best possible behavior.  With 'decode' it is
+        # returning the raw bytes, so that test should be of correct behavior,
+        # or at least produce the same result that email4 did.
+        m = self.bodytest_msg.format(charset='utf-8',
+                                     cte='quoted-printable',
+                                     bodyline='p=C3=B6stál').encode('utf-8')
+        msg = email.message_from_bytes(m)
+        self.assertEqual(msg.get_payload(), 'p=C3=B6stál\n')
+        self.assertEqual(msg.get_payload(decode=True),
+                         'pöstál\n'.encode('utf-8'))
+
+    def test_invalid_8bit_in_non_8bit_cte_uses_replace(self):
+        # This is similar to the previous test, but proves that if the 8bit
+        # byte is undecodeable in the specified charset, it gets replaced
+        # by the unicode 'unknown' character.  Again, this may or may not
+        # be the ideal behavior.  Note that if decode=False none of the
+        # decoders will get involved, so this is the only test we need
+        # for this behavior.
+        m = self.bodytest_msg.format(charset='ascii',
+                                     cte='quoted-printable',
+                                     bodyline='p=C3=B6stál').encode('utf-8')
+        msg = email.message_from_bytes(m)
+        self.assertEqual(msg.get_payload(), 'p=C3=B6st��l\n')
+        self.assertEqual(msg.get_payload(decode=True),
+                        'pöstál\n'.encode('utf-8'))
+
+    def test_8bit_in_base64_body(self):
+        # Sticking an 8bit byte in a base64 block makes it undecodable by
+        # normal means, so the block is returned undecoded, but as bytes.
+        m = self.bodytest_msg.format(charset='utf-8',
+                                     cte='base64',
+                                     bodyline='cMO2c3RhbAá=').encode('utf-8')
+        msg = email.message_from_bytes(m)
+        self.assertEqual(msg.get_payload(decode=True),
+                         'cMO2c3RhbAá=\n'.encode('utf-8'))
+
+    def test_8bit_in_uuencode_body(self):
+        # Sticking an 8bit byte in a uuencode block makes it undecodable by
+        # normal means, so the block is returned undecoded, but as bytes.
+        m = self.bodytest_msg.format(charset='utf-8',
+                                     cte='uuencode',
+                                     bodyline='<,.V<W1A; á ').encode('utf-8')
+        msg = email.message_from_bytes(m)
+        self.assertEqual(msg.get_payload(decode=True),
+                         '<,.V<W1A; á \n'.encode('utf-8'))
+
+
+    headertest_msg = textwrap.dedent("""\
+        From: foo@bar.com
+        To: báz
+        Subject: Maintenant je vous présente mon collègue, le pouf célèbre
+        \tJean de Baddie
+        From: göst
+
+        Yes, they are flying.
+        """).encode('utf-8')
+
+    def test_get_8bit_header(self):
+        msg = email.message_from_bytes(self.headertest_msg)
+        self.assertEqual(msg.get('to'), 'b??z')
+        self.assertEqual(msg['to'], 'b??z')
+
+    def test_print_8bit_headers(self):
+        msg = email.message_from_bytes(self.headertest_msg)
+        self.assertEqual(str(msg),
+                         self.headertest_msg.decode(
+                            'ascii', 'replace').replace('�', '?'))
+
+    def test_values_with_8bit_headers(self):
+        msg = email.message_from_bytes(self.headertest_msg)
+        self.assertListEqual(msg.values(),
+                              ['foo@bar.com',
+                               'b??z',
+                               'Maintenant je vous pr??sente mon '
+                                   'coll??gue, le pouf c??l??bre\n'
+                                   '\tJean de Baddie',
+                               "g??st"])
+
+    def test_items_with_8bit_headers(self):
+        msg = email.message_from_bytes(self.headertest_msg)
+        self.assertListEqual(msg.items(),
+                              [('From', 'foo@bar.com'),
+                               ('To', 'b??z'),
+                               ('Subject', 'Maintenant je vous pr??sente mon '
+                                              'coll??gue, le pouf c??l??bre\n'
+                                              '\tJean de Baddie'),
+                               ('From', 'g??st')])
+
+    def test_get_all_with_8bit_headers(self):
+        msg = email.message_from_bytes(self.headertest_msg)
+        self.assertListEqual(msg.get_all('from'),
+                              ['foo@bar.com',
+                               'g??st'])
+
+    non_latin_bin_msg = textwrap.dedent("""\
+        From: foo@bar.com
+        To: báz
+        Subject: Maintenant je vous présente mon collègue, le pouf célèbre
+        \tJean de Baddie
+        Mime-Version: 1.0
+        Content-Type: text/plain; charset="utf-8"
+        Content-Transfer-Encoding: 8bit
+
+        Да, они летят.
+        """).encode('utf-8')
+
+    def test_bytes_generator(self):
+        msg = email.message_from_bytes(self.non_latin_bin_msg)
+        out = BytesIO()
+        email.generator.BytesGenerator(out).flatten(msg)
+        self.assertEqual(out.getvalue(), self.non_latin_bin_msg)
+
+    # XXX: ultimately the '?' should turn into CTE encoded bytes
+    # using 'unknown-8bit' charset.
+    non_latin_bin_msg_as7bit = textwrap.dedent("""\
+        From: foo@bar.com
+        To: b??z
+        Subject: Maintenant je vous pr??sente mon coll??gue, le pouf c??l??bre
+        \tJean de Baddie
+        Mime-Version: 1.0
+        Content-Type: text/plain; charset="utf-8"
+        Content-Transfer-Encoding: base64
+
+        0JTQsCwg0L7QvdC4INC70LXRgtGP0YIuCg==
+        """)
+
+    def test_generator_handles_8bit(self):
+        msg = email.message_from_bytes(self.non_latin_bin_msg)
+        out = StringIO()
+        email.generator.Generator(out).flatten(msg)
+        self.assertEqual(out.getvalue(), self.non_latin_bin_msg_as7bit)
+
+    def test_bytes_generator_with_unix_from(self):
+        # The unixfrom contains a current date, so we can't check it
+        # literally.  Just make sure the first word is 'From' and the
+        # rest of the message matches the input.
+        msg = email.message_from_bytes(self.non_latin_bin_msg)
+        out = BytesIO()
+        email.generator.BytesGenerator(out).flatten(msg, unixfrom=True)
+        lines = out.getvalue().split(b'\n')
+        self.assertEqual(lines[0].split()[0], b'From')
+        self.assertEqual(b'\n'.join(lines[1:]), self.non_latin_bin_msg)
+
+    def test_message_from_binary_file(self):
+        fn = 'test.msg'
+        self.addCleanup(unlink, fn)
+        with open(fn, 'wb') as testfile:
+            testfile.write(self.non_latin_bin_msg)
+        m = email.parser.BytesParser().parse(open(fn, 'rb'))
+        self.assertEqual(str(m), self.non_latin_bin_msg_as7bit)
+
+    latin_bin_msg = textwrap.dedent("""\
+        From: foo@bar.com
+        To: Dinsdale
+        Subject: Nudge nudge, wink, wink
+        Mime-Version: 1.0
+        Content-Type: text/plain; charset="latin-1"
+        Content-Transfer-Encoding: 8bit
+
+        oh là là, know what I mean, know what I mean?
+        """).encode('latin-1')
+
+    latin_bin_msg_as7bit = textwrap.dedent("""\
+        From: foo@bar.com
+        To: Dinsdale
+        Subject: Nudge nudge, wink, wink
+        Mime-Version: 1.0
+        Content-Type: text/plain; charset="iso-8859-1"
+        Content-Transfer-Encoding: quoted-printable
+
+        oh l=E0 l=E0, know what I mean, know what I mean?
+        """)
+
+    def test_string_generator_reencodes_to_quopri_when_appropriate(self):
+        m = email.message_from_bytes(self.latin_bin_msg)
+        self.assertEqual(str(m), self.latin_bin_msg_as7bit)
+
+    def test_decoded_generator_emits_unicode_body(self):
+        m = email.message_from_bytes(self.latin_bin_msg)
+        out = StringIO()
+        email.generator.DecodedGenerator(out).flatten(m)
+        #DecodedHeader output contains an extra blank line compared
+        #to the input message.  RDM: not sure if this is a bug or not,
+        #but it is not specific to the 8bit->7bit conversion.
+        self.assertEqual(out.getvalue(),
+            self.latin_bin_msg.decode('latin-1')+'\n')
+
+    def test_bytes_feedparser(self):
+        bfp = email.feedparser.BytesFeedParser()
+        for i in range(0, len(self.latin_bin_msg), 10):
+            bfp.feed(self.latin_bin_msg[i:i+10])
+        m = bfp.close()
+        self.assertEqual(str(m), self.latin_bin_msg_as7bit)
+
+
+class TestBytesGeneratorIdempotent(TestIdempotent):
+
+    def _msgobj(self, filename):
+        with openfile(filename, 'rb') as fp:
+            data = fp.read()
+        msg = email.message_from_bytes(data)
+        return msg, data
+
+    def _idempotent(self, msg, data):
+        b = BytesIO()
+        g = email.generator.BytesGenerator(b, maxheaderlen=0)
+        g.flatten(msg)
+        self.assertEqual(data, b.getvalue())
+
+    maxDiff = None
+
+    def assertEqual(self, str1, str2):
+        self.assertListEqual(str1.split(b'\n'), str2.split(b'\n'))
+
+
+
 class TestBase64(unittest.TestCase):
     def test_len(self):
         eq = self.assertEqual
diff --git a/Misc/NEWS b/Misc/NEWS
index 7694e76..24b64f3 100644
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -92,6 +92,9 @@
 Library
 -------
 
+- Issue #4661: email can now parse bytes input and generate either converted
+  7bit output or bytes output.  Email version bumped to 5.1.0.
+
 - Issue #1589: Add ssl.match_hostname(), to help implement server identity
   verification for higher-level protocols.