UTF-8 quoted-printable, multiline subject for Thunderbird?

Question

Let's say I want to compose an email header with UTF-8, quoted-printable encoded subject, which is "test — UNIX-утилита для проверки типа файла и сравнения значений". I can confirm the bytes of the characters using:

$ echo "UNIX-утилита ..." | perl utfinfo.pl
Got 16 uchars
Char: 'U' u: 85 [0x0055] b: 85 [0x55] n: LATIN CAPITAL LETTER U [Basic Latin]
Char: 'N' u: 78 [0x004E] b: 78 [0x4E] n: LATIN CAPITAL LETTER N [Basic Latin]
Char: 'I' u: 73 [0x0049] b: 73 [0x49] n: LATIN CAPITAL LETTER I [Basic Latin]
Char: 'X' u: 88 [0x0058] b: 88 [0x58] n: LATIN CAPITAL LETTER X [Basic Latin]
Char: '-' u: 45 [0x002D] b: 45 [0x2D] n: HYPHEN-MINUS [Basic Latin]
Char: 'у' u: 1091 [0x0443] b: 209,131 [0xD1,0x83] n: CYRILLIC SMALL LETTER U [Cyrillic]
Char: 'т' u: 1090 [0x0442] b: 209,130 [0xD1,0x82] n: CYRILLIC SMALL LETTER TE [Cyrillic]
Char: 'и' u: 1080 [0x0438] b: 208,184 [0xD0,0xB8] n: CYRILLIC SMALL LETTER I [Cyrillic]
...

So, I'm trying to get the UTF-8, quoted printable representation of this. For instance, using Python's quopri:

$ python -c 'import quopri; a="test — UNIX-утилита для проверки типа файла и сравнения значений"; print(quopri.encodestring(a));'
test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=
=D1=8F =D0=BF=D1=80=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=
=D0=B0 =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0=D0=B2=D0=BD=
=D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9

... or PHP's quoted_printable_encode, which gives the exact same output:

$ php -r '$a="test — UNIX-утилита для проверки типа файла и сравнения значений"; echo quoted_printable_encode($a)."\n";'
test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=
=D1=8F =D0=BF=D1=80=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=
=D0=B0 =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0=D0=B2=D0=BD=
=D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9

So, to test, I make a text file called test.eml, and try to simply wrap this output in the =?UTF-8?Q? ... ?= tags for the Subject: line, making sure that line endings are CRLF \r\n:

Message-Id: <4c428d27a41043e2b2b07e@example.com>
Subject: =?UTF-8?Q?test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=
=D1=8F =D0=BF=D1=80=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=
=D0=B0 =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0=D0=B2=D0=BD=
=D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9?=
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Hello world

... but if I open this in Thunderbird, I get a corrupt output:

I've read somewhere that multiline in long header fields is covered by RFC0822 "LONG HEADER FIELDS", and basically, the line ending should be followed by a space. So I indent the continuation lines by one space:

Message-Id: <4c428d27a41043e2b2b07e@example.com>
Subject: =?UTF-8?Q?test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=
 =D1=8F =D0=BF=D1=80=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=
 =D0=B0 =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0=D0=B2=D0=BD=
 =D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9?=
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Hello world

... and I get a slighly different subject in Thunderbird, but still corrupt:

Now, if I delete =\r\n from the first three continuation lines, so the subject is all in one line:

Message-Id: <4c428d27a41043e2b2b07e@example.com>
Subject: =?UTF-8?Q?test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=D1=8F =D0=BF=D1=80=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=D0=B0 =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0=D0=B2=D0=BD=D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9?=
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Hello world

... then actually Thunderbird shows the subject line well:

... but then my header is in conflict with the recommendation from RFC 2822 - 2.1.1. Line Length Limits which says "Each line of characters MUST be no more than 998 characters, and SHOULD be no more than 78 characters, excluding the CRLF."; specifically the line limit of 78 characters.

So, how can I obtain the proper multi-line quoted-printable representation of an UTF-8 Subject header string, so I can use it in an .eml file split at 78 characters - and have Thunderbird correctly read it?

Answer 1

When I ask python to create an email with that subject, here's what it does:

$ python
Python 2.7.9 (default, Mar  1 2015, 18:22:53)
[GCC 4.9.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from email.message import Message
>>> from email.header import Header
>>> msg = Message()
>>> import quopri
>>> h = Header(quopri.decodestring('test =E2=80=94 UNIX-'
     '=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=D1=8F'
     '=D0=BF=D1=80=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8'
     '=D0=BF=D0=B0 =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8'
     '=D1=81=D1=80=D0=B0=D0=B2=D0=BD=D0=B5=D0=BD=D0=B8=D1=8F '
     '=D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9?='), 'UTF-8')
>>> msg['Subject'] = h
>>> print msg.as_string()
Subject: =?utf-8?b?dGVzdCDigJQgVU5JWC3Rg9GC0LjQu9C40YLQsCDQtNC70Y8g0L/RgNC+0LI=?=
 =?utf-8?b?0LXRgNC60Lgg0YLQuNC/0LAg0YTQsNC50LvQsCDQuCDRgdGA0LDQstC90LU=?=
 =?utf-8?b?0L3QuNGPINC30L3QsNGH0LXQvdC40Lk/?=


>>>

So it uses base64 encoding instead of quoted-printable, but my strong suspicion, based on this, is that the answer is that each line must begin and end the escape.

Indeed:

>>> import email
>>> s = '''Subject: =?UTF-8?Q?test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8=D0?=
...  =?UTF-8?Q?=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=D1=8F =D0=BF=D1=80=D0?=
...  =?UTF-8?Q?=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=D0=B0?=
...  =?UTF-8?Q? =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0=D0?=
...  =?UTF-8?Q?=B2=D0=BD=D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0=D1?=
...  =?UTF-8?Q?=87=D0=B5=D0=BD=D0=B8=D0=B9?=
...
... Hello.
... '''
>>> e = email.message_from_string(s.replace('\n', '\r\n'))
>>> email.header.decode_header(e['Subject'])
[('test \xe2\x80\x94 UNIX-\xd1\x83\xd1\x82\xd0\xb8\xd0\xbb\xd0\xb8\xd1\x82\xd0\xb0 \xd0\xb4\xd0\xbb\xd1\x8f \xd0\xbf\xd1\x80\xd0\xbe\xd0\xb2\xd0\xb5\xd1\x80\xd0\xba\xd0\xb8 \xd1\x82\xd0\xb8\xd0\xbf\xd0\xb0 \xd1\x84\xd0\xb0\xd0\xb9\xd0\xbb\xd0\xb0 \xd0\xb8 \xd1\x81\xd1\x80\xd0\xb0\xd0\xb2\xd0\xbd\xd0\xb5\xd0\xbd\xd0\xb8\xd1\x8f \xd0\xb7\xd0\xbd\xd0\xb0\xd1\x87\xd0\xb5\xd0\xbd\xd0\xb8\xd0\xb9', 'utf-8')]
>>> decoded = email.header.decode_header(e['Subject'])
>>> print decoded[0][0].decode(decoded[0][1])
test — UNIX-утилита для проверки типа файла и сравнения значений

---

EDIT: However, even with the above added in .eml file, Thunderbird fails again:

... but this time it indicates it got some of the chars correct. And indeed, breakage occurs where lines are broken "in the middle of a character"; say if for the sequence 0xD1, 0x83 for the character у, the =D1?= ends one line, and the Q?=83 starts the other, then Thunderbird cannot parse that. So after manual rearrangement, this snippet can be obtained:

Message-Id: <4c428d27a41043e2b2b07e@example.com>
Subject: =?UTF-8?Q?test =E2=80=94 UNIX-=D1=83=D1=82=D0=B8?=
 =?UTF-8?Q?=D0=BB=D0=B8=D1=82=D0=B0 =D0=B4=D0=BB=D1=8F =D0=BF=D1=80?=
 =?UTF-8?Q?=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8 =D1=82=D0=B8=D0=BF=D0=B0?=
 =?UTF-8?Q? =D1=84=D0=B0=D0=B9=D0=BB=D0=B0 =D0=B8 =D1=81=D1=80=D0=B0?=
 =?UTF-8?Q?=D0=B2=D0=BD=D0=B5=D0=BD=D0=B8=D1=8F =D0=B7=D0=BD=D0=B0?=
 =?UTF-8?Q?=D1=87=D0=B5=D0=BD=D0=B8=D0=B9?=
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Hello world

... which opens fine as an .eml message in Thunderbird (same as this image from OP).

EDIT2: Also PHP seems to do it right, with this invocation of mb_encode_mimeheader (directly pasteable in .eml file):

$ php -r '$a="test — UNIX-утилита для проверки типа файла и сравнения значений"; mb_internal_encoding("UTF-8"); echo mb_encode_mimeheader($a, "UTF-8", "Q")."\n";'
test =?UTF-8?Q?=E2=80=94=20UNIX-=D1=83=D1=82=D0=B8=D0=BB=D0=B8=D1=82?=
 =?UTF-8?Q?=D0=B0=20=D0=B4=D0=BB=D1=8F=20=D0=BF=D1=80=D0=BE=D0=B2=D0=B5?=
 =?UTF-8?Q?=D1=80=D0=BA=D0=B8=20=D1=82=D0=B8=D0=BF=D0=B0=20=D1=84=D0=B0?=
 =?UTF-8?Q?=D0=B9=D0=BB=D0=B0=20=D0=B8=20=D1=81=D1=80=D0=B0=D0=B2=D0=BD?=
 =?UTF-8?Q?=D0=B5=D0=BD=D0=B8=D1=8F=20=D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD?=
 =?UTF-8?Q?=D0=B8=D0=B9?=

Answer 2

The problem with your test.eml is that your RFC2047 encoding is broken. The Q encoding is based on quoted-printable, but is not entirely the same. In particular, each space needs to be encoded as either =20 or _, and you cannot escape line breaks with a final =.

Fundamentally, each =?...?= sequence needs to be a single, unambiguous token per RFC 822. You can either break up your input into multiple such tokens and leave the spaces unencoded, or encode the spaces. Note that spaces between two such tokens are not significant, so encoding the spaces into the sequences makes more sense.

Message-Id: <4c428d27a41043e2b2b07e@example.com>
Subject: =?UTF-8?Q?test_=E2=80=94_UNIX-=D1=83=D1=82=D0=B8=D0=BB?=
 =?UTF-8?Q?=D0=B8=D1=82=D0=B0_=D0=B4=D0=BB_=D1=8F_=D0=BF=D1=80?=
 =?UTF-8?Q?=D0=BE=D0=B2=D0=B5=D1=80=D0=BA=D0=B8_=D1=82=D0=B8=D0=BF?=
 =?UTF-8?Q?=D0=B0_=D1=84=D0=B0=D0=B9=D0=BB=D0=B0_=D0=B8_=D1=81?=
 =?UTF-8?Q?=D1=80=D0=B0=D0=B2=D0=BD_=D0=B5=D0=BD=D0=B8=D1=8F_?=
 =?UTF-8?Q?=D0=B7=D0=BD=D0=B0=D1=87=D0=B5=D0=BD=D0=B8=D0=B9?=
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Hello world

Of course, with this exposition, quoted-printable isn't really legible at all, and probably takes much more space than base64, so you might prefer to go with the B encoding in the end after all.

Unless you are writing a MIME library yourself, the simple solution is to not care, and let the library piece this together for you. PHP is more problematic (the standard library lacks this functionality, and the third-party libraries are somewhat uneven--find one you trust, and stick to it), but in Python, simply pass in a Unicode string, and the email library will encode it if necessary.