text/* parts of email messages must use \r\n as the
newline separator. For unencoded messages. smtplib and
friends take care of the translation from \n to \r\n in
the SMTP processing.

Parts which are unencoded (i.e. 7bit character sets)
MUST use \n line endings, or smtplib with translate to
\r\r\n.

Parts that get encoded using quoted-printable can use
either, because the qp-encoder assumes input data is
text and reencodes with \n.

However, parts which get encoded using base64 are NOT
translated, and so must use \r\n line endings.

This means you have to guess whether your text is going
to get encoded or not (admittedly, usually not that
hard), and translate the line endings appropriately
before generating a Message instance.

I think the fix would be for Charset.encode_body() to
alway force the encoder to text mode
(i.e.binary=False), since it seems unlikely to have a
Charset for something which is not text.

Email sprint candidate.

Well, it's two years later, but I did look at this during the sprints at PyCon, though I didn't get as far as posting it then (I only just now rediscovered the patch on my laptop).

Python3 no longer has a "binary" flag on base64mime.encode, so here is a proposed patch for Python3.  I'm not sure if this should be backported or not, but I'm leaning that way.  Theoretically it should be only an improvement, but I can easily imagine unix-only programs unknowingly depending on the previous non-translation of newlines.  Still, since email is about intermachine communication and this clearly makes it more RFC compliant, the change is a legitimate bug fix and the chance of breakage is relatively small.

Tests are still needed.

Added some tests against the patch provided by R. David Murray. See attached patch.

Tests pass against default.

I almost applied this patch, but my gut is having second thoughts about it.  I don't think this is the correct solution.  The correct solution would be to delay the encoding of the body part until the message generation phase, and use the requested linesep at that point.  That is, in 3.2 and 3.3 I've changed the paradigm from "always use \n and convert the final string at need" to "specify the linesep when flattening the message".

I'm uploading the completed patch here so I don't lose it, but I don't think I'm going to use it in this form.