Quoted-printable, MIME and email encoding

Why email needs its own encoding layer, how encoded words in headers work, when a body is base64 rather than quoted-printable, and the line length limits that cause silent breakage.

Email was designed for 7-bit text

SMTP originally carried only 7-bit ASCII lines of at most 1000 characters. MIME adds headers that describe how the real content is encoded, which is why a modern message has several layers of encoding before the text you read.

Subject: =?UTF-8?Q?Caf=C3=A9_meeting?=
From: =?UTF-8?B?QW5hIFBlem5pY2s=?= <[email protected]>
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0

The caf=C3=A9 is open until 18:00 =
on weekdays.

Q means quoted-printable, B means base64, and the encoded word may not exceed 75 characters.
Adjacent encoded words are joined; the underscore in a Q-word represents a space.
A soft line break is a trailing = with no content after it — the continuation is joined with no space inserted.
Non-ASCII in a header must be encoded; a raw accented byte is rejected or mangled in transit.

Quoted-printable or base64

Encoding	Expansion	Readable as text	Best for
7bit / 8bit	None	Yes	Pure ASCII content
quoted-printable	Modest, most bytes literal	Mostly	Mostly-ASCII text with accents
base64	About 33 percent	No	Binary attachments, non-Latin text
binary	None	No	Only with an extension such as CHUNKING

import quopri, base64

raw = "The café is open until 18:00 on weekdays.".encode("utf-8")

encoded = quopri.encodestring(raw)
print(encoded.decode())
# The caf=C3=A9 is open until 18:00 on weekdays.

assert quopri.decodestring(encoded) == raw

# base64 for the same payload
b64 = base64.encodebytes(raw).decode()
assert base64.decodebytes(b64.encode()) == raw

Quoted-printable keeps a mostly-ASCII message readable in a raw dump, which is why it is preferred for text bodies. Base64 is more compact for binary data because it has no per-byte worst case of tripling.

Line length and the traps

Encoded lines must stay under 76 characters; longer lines are broken, and base64 padding can be split across lines in ways some decoders reject.
A soft break in quoted-printable adds no space, so re-wrapping text changes the meaning.
Mixing transfer encodings inside one message requires a multipart structure with a boundary.
Attachment filenames are themselves encoded words, and older clients only understand a subset of charsets.
A message that declares UTF-8 but encodes the body as 8bit will be corrupted by any relay that normalises to 7-bit.

from email.message import EmailMessage

msg = EmailMessage()
msg["Subject"] = "Café meeting"              # the library encodes the header for you
msg["From"] = "[email protected]"
msg["To"] = "[email protected]"
msg.set_content("The café is open until 18:00.")

with open("report.pdf", "rb") as f:
    msg.add_attachment(f.read(), maintype="application",
                       subtype="pdf", filename="report.pdf")

# let the library choose the transfer encoding
assert msg["Content-Transfer-Encoding"] in ("quoted-printable", "base64")

💡

Never hand-write MIME. The header encoding, line folding, boundary selection and charset declaration interact, and a small mistake produces a message that looks fine to one client and arrives as a wall of equals signs in another.

FAQ

Why does my email body show equals signs?

A client failed to decode quoted-printable, or the decoder was given the wrong charset. The =C3=A9 sequences are the UTF-8 bytes of an accented letter.

Should I use base64 for a plain text body?

Only when the text is mostly non-Latin. Base64 hides the content from inspection and adds a third more bytes, which is why quoted-printable is the default for text.

Escape sequences across languages Compression vs encoding: gzip, deflate and Brotli

Last refreshed 2026-09-18.

Quoted-printable, MIME and email encoding

Email was designed for 7-bit text

Quoted-printable or base64

Line length and the traps

FAQ

Related