In Lib/email/encoders.py:

def encode_7or8bit(msg):
    """Set the Content-Transfer-Encoding header to 7bit or 8bit."""
    orig = msg.get_payload(decode=True)
    if orig is None:
        # There's no payload.  For backwards compatibility we use 7bit
        msg['Content-Transfer-Encoding'] = '7bit'
        return
    # We play a trick to make this go fast.  If encoding/decode to ASCII
    # succeeds, we know the data must be 7bit, otherwise treat it as 8bit.
    try:
        if isinstance(orig, str):
            orig.encode('ascii')
        else:
            orig.decode('ascii')
    except UnicodeError:
        charset = msg.get_charset()


msg.get_payload(decode=True) always return bytes so there is no point of these lines:

        if isinstance(orig, str):
            orig.encode('ascii')
        else:
            orig.decode('ascii')

Attached the patch to refactor this function.

Good point, thanks for the patch.

Then special case for iso-2022-* is not needed too.

Yes, in theory that should be true at this point.

Okay, here is the patch on which we removed checking of special case for iso-2022-*.

New changeset 29e5dd1f608a by R David Murray in branch 'default':
#19957: Simplify encode_7or8bit now that _payload is always str.
http://hg.python.org/cpython/rev/29e5dd1f608a

Thanks, Vajrasky and Serhiy.