Don't encode emails with base64/qp

[phr]

#3472 introduced a patch, which replaces base64 encoding with quoted-printable.

Anyway, there's even better way to do get rid of unnecessary encoding:

from email import Charset 
Charset.add_charset('utf-8',Charset.SHORTEST,None,'utf-8')

The above will set Content-Transfer-Encoding to 7bit or 8bit as needed - exactly as Mutt or Thunderbird do
and completely solve the problem even for non-latin characters.

[Russell Keith-Magee]

Tests aren't optional. If you're changing code, you need tests - or a good excuse why a test isn't possible.

[phr]

I'm not quite sure what to test here. Only a single parameter needs to be changed
in file django/core/mail.py at line 23:

-Charset.add_charset('utf-8', Charset.SHORTEST, Charset.QP, 'utf-8')
+Charset.add_charset('utf-8', Charset.SHORTEST, None, 'utf-8')

[Luke Plant]

You need to update and add to the tests in tests/regressiontests/mail/tests.py, which fail with the above change. You should add tests that show what happens with 7bit and 8bit data.

[phr]

Yes, sure those tests fail. They test for:

Content-Transfer-Encoding: quoted-printable

at multiple places, which after the proposed change will not pass.
If the input is pure ascii, that header will become:

Content-Transfer-Encoding: 7bit

If the input contains some accentuated character - e.g. 'áéí', there will be 8bit instead.

[gisle]

Patch with adjusted tests

[gisle]

I've included a patch that makes the tests pass, but as demonstrated by the patch this make us start generating email with "Content-Transfer-Encoding: 8bit" when passing in Unicode strings as subject or content. I think that should be avoided, so I don't recommend this patch as is.

[James Bennett]

1.2 is feature-frozen, moving this feature request off the milestone.

[phr]

Replying to gisle:

I've included a patch that makes the tests pass, but as demonstrated by the patch this make us start generating email with "Content-Transfer-Encoding: 8bit" when passing in Unicode strings as subject or content. I think that should be avoided, so I don't recommend this patch as is.

The use of 8bit encoding is exactly the aim of this ticket, so the patch does the right thing IMHO.

Except for a few legacy paths, SMTP is mostly 8bit today - and where not, it's the MTA's role to downconvert to 7bit.

Many email clients are generating 8bit-encoded messages (Thunderbird, Mutt, etc) and people routinely use 8bit email every day. Avoiding it makes no longer any sense, and the current practice of using quoted-printable does not work right for non-latin aplhabets.

Please note that other software already implemented my proposal - see e.g. ​http://trac.edgewall.org/ticket/8252

[Ramiro Morales]

Tests need to be converted to unit tests.

[Ramiro Morales]

Patch updated to trunk status as of now, also added test for #3472/r5143.

[Jannis Leidel]

In [16178]:

Fixed #11212 -- Stopped using quoted-printable encoding for mails with non-ASCII characters but rely on 8bit encoding instead. Thanks, phr, gisle and Ramiro Morales.

[Jacob]

Milestone 1.3 deleted