Multipurpose Internet Mail Extensions (MIME) Part One: Format of Internet Message Bodies

The term CRLF, in this set of documents, refers to the sequence of octets corresponding to the two US-ASCII characters CR (decimal value 13) and LF (decimal value 10) which, taken together, in this order, denote a line break in RFC 822 mail.¶

The term "character set" is used in MIME to refer to a method of converting a sequence of octets into a sequence of characters. Note that unconditional and unambiguous conversion in the other direction is not required, in that not all characters may be representable by a given character set and a character set may provide more than one sequence of octets to represent a particular sequence of characters.¶

This definition is intended to allow various kinds of character encodings, from simple single-table mappings such as US-ASCII to complex table switching methods such as those that use ISO 2022's techniques, to be used as character sets. However, the definition associated with a MIME character set name must fully specify the mapping to be performed. In particular, use of external profiling information to determine the exact mapping is not permitted.¶

NOTE: The term "character set" was originally to describe such straightforward schemes as US-ASCII and ISO-8859-1 which have a simple one-to-one mapping from single octets to single characters. Multi-octet coded character sets and switching techniques make the situation more complex. For example, some communities use the term "character encoding" for what MIME calls a "character set", while using the phrase "coded character set" to denote an abstract mapping from integers (not octets) to characters.¶

The term "message", when not further qualified, means either a (complete or "top-level") RFC 822 message being transferred on a network, or a message encapsulated in a body of type "message/rfc822" or "message/partial".¶

The term "entity", refers specifically to the MIME-defined header fields and contents of either a message or one of the parts in the body of a multipart entity. The specification of such entities is the essence of MIME. Since the contents of an entity are often called the "body", it makes sense to speak about the body of an entity. Any sort of field may be present in the header of an entity, but only those fields whose names begin with "content-" actually have any MIME-related meaning. Note that this does NOT imply thay they have no meaning at all -- an entity that is also a message has non- MIME header fields whose meanings are defined by RFC 822.¶

The term "body part" refers to an entity inside of a multipart entity.¶

The term "body", when not further qualified, means the body of an entity, that is, the body of either a message or of a body part.¶

NOTE: The previous four definitions are clearly circular. This is unavoidable, since the overall structure of a MIME message is indeed recursive.¶

"7bit data" refers to data that is all represented as relatively short lines with 998 octets or less between CRLF line separation sequences [RFC-821]. No octets with decimal values greater than 127 are allowed and neither are NULs (octets with decimal value 0). CR (decimal value 13) and LF (decimal value 10) octets only occur as part of CRLF line separation sequences.¶

"8bit data" refers to data that is all represented as relatively short lines with 998 octets or less between CRLF line separation sequences [RFC-821]), but octets with decimal values greater than 127 may be used. As with "7bit data" CR and LF octets only occur as part of CRLF line separation sequences and no NULs are allowed.¶

"Binary data" refers to data where any sequence of octets whatsoever is allowed.¶

"Lines" are defined as sequences of octets separated by a CRLF sequences. This is consistent with both RFC 821 and RFC 822. "Lines" only refers to a unit of data in a message, which may or may not correspond to something that is actually displayed by a user agent.¶

Note that the definition of "tspecials" is the same as the RFC 822 definition of "specials" with the addition of the three characters "/", "?", and "=", and the removal of ".".¶

Note also that a subtype specification is MANDATORY -- it may not be omitted from a Content-Type header field. As such, there are no default subtypes.¶

The type, subtype, and parameter names are not case sensitive. For example, TEXT, Text, and TeXt are all equivalent top-level media types. Parameter values are normally case sensitive, but sometimes are interpreted in a case-insensitive fashion, depending on the intended use. (For example, multipart boundaries are case-sensitive, but the "access-type" parameter for message/External-body is not case-sensitive.)¶

Note that the value of a quoted string parameter does not include the quotes. That is, the quotation marks in a quoted-string are not a part of the value of the parameter, but are merely used to delimit that parameter value. In addition, comments are allowed in accordance with RFC 822 rules for structured header fields. Thus the following two forms¶

are completely equivalent.¶

Beyond this syntax, the only syntactic constraint on the definition of subtype names is the desire that their uses must not conflict. That is, it would be undesirable to have two different communities using "Content-Type: application/foobar" to mean two different things. The process of defining new media subtypes, then, is not intended to be a mechanism for imposing restrictions, but simply a mechanism for publicizing their definition and usage. There are, therefore, two acceptable mechanisms for defining new media subtypes: ¶

1. Private values (starting with "X-") may be defined bilaterally between two cooperating agents without outside registration or standardization. Such values cannot be registered or standardized.
2. New standard values should be registered with IANA as described in RFC 2048.

The second document in this set, RFC 2046, defines the initial set of media types for MIME.¶

This default is assumed if no Content-Type header field is specified. It is also recommend that this default be assumed when a syntactically invalid Content-Type header field is encountered. In the presence of a MIME-Version header field and the absence of any Content-Type header field, a receiving User Agent can also assume that plain US-ASCII text was the sender's intent. Plain US-ASCII text may still be assumed in the absence of a MIME-Version or the presence of an syntactically invalid Content-Type header field, but the sender's intent might have been otherwise.¶

These values are not case sensitive -- Base64 and BASE64 and bAsE64 are all equivalent. An encoding type of 7BIT requires that the body is already in a 7bit mail-ready representation. This is the default value -- that is, "Content-Transfer-Encoding: 7BIT" is assumed if the Content-Transfer-Encoding header field is not present.¶

This single Content-Transfer-Encoding token actually provides two pieces of information. It specifies what sort of encoding transformation the body was subjected to and hence what decoding operation must be used to restore it to its original form, and it specifies what the domain of the result is.¶

The transformation part of any Content-Transfer-Encodings specifies, either explicitly or implicitly, a single, well-defined decoding algorithm, which for any sequence of encoded octets either transforms it to the original sequence of octets which was encoded, or shows that it is illegal as an encoded sequence. Content-Transfer- Encodings transformations never depend on any additional external profile information for proper operation. Note that while decoders must produce a single, well-defined output for a valid encoding no such restrictions exist for encoders: Encoding a given sequence of octets to different, equivalent encoded sequences is perfectly legal.¶

Three transformations are currently defined: identity, the "quoted- printable" encoding, and the "base64" encoding. The domains are "binary", "8bit" and "7bit".¶

The Content-Transfer-Encoding values "7bit", "8bit", and "binary" all mean that the identity (i.e. NO) encoding transformation has been performed. As such, they serve simply as indicators of the domain of the body data, and provide useful information about the sort of encoding that might be needed for transmission in a given transport system. The terms "7bit data", "8bit data", and "binary data" are all defined in Section 2.¶

The quoted-printable and base64 encodings transform their input from an arbitrary domain into material in the "7bit" range, thus making it safe to carry over restricted transports. The specific definition of the transformations are given below.¶

The proper Content-Transfer-Encoding label must always be used. Labelling unencoded data containing 8bit characters as "7bit" is not allowed, nor is labelling unencoded non-line-oriented data as anything other than "binary" allowed.¶

Unlike media subtypes, a proliferation of Content-Transfer-Encoding values is both undesirable and unnecessary. However, establishing only a single transformation into the "7bit" domain does not seem possible. There is a tradeoff between the desire for a compact and efficient encoding of largely- binary data and the desire for a somewhat readable encoding of data that is mostly, but not entirely, 7bit. For this reason, at least two encoding mechanisms are necessary: a more or less readable encoding (quoted-printable) and a "dense" or "uniform" encoding (base64).¶

Mail transport for unencoded 8bit data is defined in RFC 1652. As of the initial publication of this document, there are no standardized Internet mail transports for which it is legitimate to include unencoded binary data in mail bodies. Thus there are no circumstances in which the "binary" Content-Transfer-Encoding is actually valid in Internet mail. However, in the event that binary mail transport becomes a reality in Internet mail, or when MIME is used in conjunction with any other binary-capable mail transport mechanism, binary bodies must be labelled as such using this mechanism.¶

NOTE: The five values defined for the Content-Transfer-Encoding field imply nothing about the media type other than the algorithm by which it was encoded or the transport system requirements if unencoded.¶

Implementors may, if necessary, define private Content-Transfer- Encoding values, but must use an x-token, which is a name prefixed by "X-", to indicate its non-standard status, e.g., "Content-Transfer- Encoding: x-my-new-encoding". Additional standardized Content- Transfer-Encoding values must be specified by a standards-track RFC. The requirements such specifications must meet are given in RFC 2048. As such, all content-transfer-encoding namespace except that beginning with "X-" is explicitly reserved to the IETF for future use.¶

Unlike media types and subtypes, the creation of new Content- Transfer-Encoding values is STRONGLY discouraged, as it seems likely to hinder interoperability with little potential benefit¶

If a Content-Transfer-Encoding header field appears as part of a message header, it applies to the entire body of that message. If a Content-Transfer-Encoding header field appears as part of an entity's headers, it applies only to the body of that entity. If an entity is of type "multipart" the Content-Transfer-Encoding is not permitted to have any value other than "7bit", "8bit" or "binary". Even more severe restrictions apply to some subtypes of the "message" type.¶

It should be noted that most media types are defined in terms of octets rather than bits, so that the mechanisms described here are mechanisms for encoding arbitrary octet streams, not bit streams. If a bit stream is to be encoded via one of these mechanisms, it must first be converted to an 8bit byte stream using the network standard bit order ("big-endian"), in which the earlier bits in a stream become the higher-order bits in a 8bit byte. A bit stream not ending at an 8bit boundary must be padded with zeroes. RFC 2046 provides a mechanism for noting the addition of such padding in the case of the application/octet-stream media type, which has a "padding" parameter.¶

This must be interpreted to mean that the body is a base64 US-ASCII encoding of data that was originally in ISO-8859-1, and will be in that character set again after decoding.¶

Certain Content-Transfer-Encoding values may only be used on certain media types. In particular, it is EXPRESSLY FORBIDDEN to use any encodings other than "7bit", "8bit", or "binary" with any composite media type, i.e. one that recursively includes other Content-Type fields. Currently the only composite media types are "multipart" and "message". All encodings that are desired for bodies of type multipart or message must be done at the innermost level, by encoding the actual body that needs to be encoded.¶

It should also be noted that, by definition, if a composite entity has a transfer-encoding value such as "7bit", but one of the enclosed entities has a less restrictive value such as "8bit", then either the outer "7bit" labelling is in error, because 8bit data are included, or the inner "8bit" labelling placed an unnecessarily high demand on the transport system because the actual included data were actually 7bit-safe.¶

NOTE ON ENCODING RESTRICTIONS: Though the prohibition against using content-transfer-encodings on composite body data may seem overly restrictive, it is necessary to prevent nested encodings, in which data are passed through an encoding algorithm multiple times, and must be decoded multiple times in order to be properly viewed. Nested encodings add considerable complexity to user agents: Aside from the obvious efficiency problems with such multiple encodings, they can obscure the basic structure of a message. In particular, they can imply that several decoding operations are necessary simply to find out what types of bodies a message contains. Banning nested encodings may complicate the job of certain mail gateways, but this seems less of a problem than the effect of nested encodings on user agents.¶

Any entity with an unrecognized Content-Transfer-Encoding must be treated as if it has a Content-Type of "application/octet-stream", regardless of what the Content-Type header field actually says.¶

NOTE ON THE RELATIONSHIP BETWEEN CONTENT-TYPE AND CONTENT-TRANSFER- ENCODING: It may seem that the Content-Transfer-Encoding could be inferred from the characteristics of the media that is to be encoded, or, at the very least, that certain Content-Transfer-Encodings could be mandated for use with specific media types. There are several reasons why this is not the case. First, given the varying types of transports used for mail, some encodings may be appropriate for some combinations of media types and transports but not for others. (For example, in an 8bit transport, no encoding would be required for text in certain character sets, while such encodings are clearly required for 7bit SMTP.)¶

Second, certain media types may require different types of transfer encoding under different circumstances. For example, many PostScript bodies might consist entirely of short lines of 7bit data and hence require no encoding at all. Other PostScript bodies (especially those using Level 2 PostScript's binary encoding mechanism) may only be reasonably represented using a binary transport encoding. Finally, since the Content-Type field is intended to be an open-ended specification mechanism, strict specification of an association between media types and encodings effectively couples the specification of an application protocol with a specific lower-level transport. This is not desirable since the developers of a media type should not have to be aware of all the transports in use and what their limitations are.¶

The quoted-printable and base64 encodings are designed so that conversion between them is possible. The only issue that arises in such a conversion is the handling of hard line breaks in quoted- printable encoding output. When converting from quoted-printable to base64 a hard line break in the quoted-printable form represents a CRLF sequence in the canonical form of the data. It must therefore be converted to a corresponding encoded CRLF in the base64 form of the data. Similarly, a CRLF sequence in the canonical form of the data obtained after base64 decoding must be converted to a quoted- printable hard line break, but ONLY when converting text data.¶

There was some confusion, in the previous versions of this RFC, regarding the model for when email data was to be converted to canonical form and encoded, and in particular how this process would affect the treatment of CRLFs, given that the representation of newlines varies greatly from system to system, and the relationship between content-transfer-encodings and character sets. A canonical model for encoding is presented in RFC 2049 for this reason.¶

The Quoted-Printable encoding is intended to represent data that largely consists of octets that correspond to printable characters in the US-ASCII character set. It encodes the data in such a way that the resulting octets are unlikely to be modified by mail transport. If the data being encoded are mostly US-ASCII text, the encoded form of the data remains largely recognizable by humans. A body which is entirely US-ASCII may also be encoded in Quoted-Printable to ensure the integrity of the data should the message pass through a character-translating, and/or line-wrapping gateway.¶

In this encoding, octets are to be represented as determined by the following rules: ¶

1. (General 8bit representation) Any octet, except a CR or LF that is part of a CRLF line break of the canonical (standard) form of the data being encoded, may be represented by an "=" followed by a two digit hexadecimal representation of the octet's value. The digits of the hexadecimal alphabet, for this purpose, are "0123456789ABCDEF". Uppercase letters must be used; lowercase letters are not allowed. Thus, for example, the decimal value 12 (US-ASCII form feed) can be represented by "=0C", and the decimal value 61 (US- ASCII EQUAL SIGN) can be represented by "=3D". This rule must be followed except when the following rules allow an alternative encoding.

2. (Literal representation) Octets with decimal values of 33 through 60 inclusive, and 62 through 126, inclusive, MAY be represented as the US-ASCII characters which correspond to those octets (EXCLAMATION POINT through LESS THAN, and GREATER THAN through TILDE, respectively).

3. (White Space) Octets with values of 9 and 32 MAY be represented as US-ASCII TAB (HT) and SPACE characters, respectively, but MUST NOT be so represented at the end of an encoded line. Any TAB (HT) or SPACE characters on an encoded line MUST thus be followed on that line by a printable character. In particular, an "=" at the end of an encoded line, indicating a soft line break (see rule #5) may follow one or more TAB (HT) or SPACE characters. It follows that an octet with decimal value 9 or 32 appearing at the end of an encoded line must be represented according to Rule #1. This rule is necessary because some MTAs (Message Transport Agents, programs which transport messages from one user to another, or perform a portion of such transfers) are known to pad lines of text with SPACEs, and others are known to remove "white space" characters from the end of a line. Therefore, when decoding a Quoted-Printable body, any trailing white space on a line must be deleted, as it will necessarily have been added by intermediate transport agents.

4. (Line Breaks) A line break in a text body, represented as a CRLF sequence in the text canonical form, must be represented by a (RFC 822) line break, which is also a CRLF sequence, in the Quoted-Printable encoding. Since the canonical representation of media types other than text do not generally include the representation of line breaks as CRLF sequences, no hard line breaks (i.e. line breaks that are intended to be meaningful and to be displayed to the user) can occur in the quoted-printable encoding of such types. Sequences like "=0D", "=0A", "=0A=0D" and "=0D=0A" will routinely appear in non-text data represented in quoted- printable, of course.

   Note that many implementations may elect to encode the local representation of various content types directly rather than converting to canonical form first, encoding, and then converting back to local representation. In particular, this may apply to plain text material on systems that use newline conventions other than a CRLF terminator sequence. Such an implementation optimization is permissible, but only when the combined canonicalization-encoding step is equivalent to performing the three steps separately.

5. (Soft Line Breaks) The Quoted-Printable encoding REQUIRES that encoded lines be no more than 76 characters long. If longer lines are to be encoded with the Quoted-Printable encoding, "soft" line breaks must be used. An equal sign as the last character on a encoded line indicates such a non-significant ("soft") line break in the encoded text.

This provides a mechanism with which long lines are encoded in such a way as to be restored by the user agent. The 76 character limit does not count the trailing CRLF, but counts all other characters, including any equal signs.¶

Since the hyphen character ("-") may be represented as itself in the Quoted-Printable encoding, care must be taken, when encapsulating a quoted-printable encoded body inside one or more multipart entities, to ensure that the boundary delimiter does not appear anywhere in the encoded body. (A good strategy is to choose a boundary that includes a character sequence such as "=_" which can never appear in a quoted-printable body. See the definition of multipart messages in RFC 2046.)¶

NOTE: The quoted-printable encoding represents something of a compromise between readability and reliability in transport. Bodies encoded with the quoted-printable encoding will work reliably over most mail gateways, but may not work perfectly over a few gateways, notably those involving translation into EBCDIC. A higher level of confidence is offered by the base64 Content-Transfer-Encoding. A way to get reasonably reliable transport through EBCDIC gateways is to also quote the US-ASCII characters¶

according to rule #1.¶

Because quoted-printable data is generally assumed to be line- oriented, it is to be expected that the representation of the breaks between the lines of quoted-printable data may be altered in transport, in the same manner that plain text mail has always been altered in Internet mail when passing between systems with differing newline conventions. If such alterations are likely to constitute a corruption of the data, it is probably more sensible to use the base64 encoding rather than the quoted-printable encoding.¶

NOTE: Several kinds of substrings cannot be generated according to the encoding rules for the quoted-printable content-transfer- encoding, and hence are formally illegal if they appear in the output of a quoted-printable encoder. This note enumerates these cases and suggests ways to handle such illegal substrings if any are encountered in quoted-printable data that is to be decoded. ¶

1. An "=" followed by two hexadecimal digits, one or both of which are lowercase letters in "abcdef", is formally illegal. A robust implementation might choose to recognize them as the corresponding uppercase letters.
2. An "=" followed by a character that is neither a hexadecimal digit (including "abcdef") nor the CR character of a CRLF pair is illegal. This case can be the result of US-ASCII text having been included in a quoted-printable part of a message without itself having been subjected to quoted-printable encoding. A reasonable approach by a robust implementation might be to include the "=" character and the following character in the decoded data without any transformation and, if possible, indicate to the user that proper decoding was not possible at this point in the data.
3. An "=" cannot be the ultimate or penultimate character in an encoded object. This could be handled as in case (2) above.
4. Control characters other than TAB, or CR and LF as parts of CRLF pairs, must not appear. The same is true for octets with decimal values greater than 126. If found in incoming quoted-printable data by a decoder, a robust implementation might exclude them from the decoded data and warn the user that illegal characters were discovered.
5. Encoded lines must not be longer than 76 characters, not counting the trailing CRLF. If longer lines are found in incoming, encoded data, a robust implementation might nevertheless decode the lines, and might report the erroneous encoding to the user.

WARNING TO IMPLEMENTORS: If binary data is encoded in quoted- printable, care must be taken to encode CR and LF characters as "=0D" and "=0A", respectively. In particular, a CRLF sequence in binary data should be encoded as "=0D=0A". Otherwise, if CRLF were represented as a hard line break, it might be incorrectly decoded on platforms with different line break conventions.¶

IMPORTANT: The addition of LWSP between the elements shown in this BNF is NOT allowed since this BNF does not specify a structured header field.¶

The Base64 Content-Transfer-Encoding is designed to represent arbitrary sequences of octets in a form that need not be humanly readable. The encoding and decoding algorithms are simple, but the encoded data are consistently only about 33 percent larger than the unencoded data. This encoding is virtually identical to the one used in Privacy Enhanced Mail (PEM) applications, as defined in RFC 1421.¶

A 65-character subset of US-ASCII is used, enabling 6 bits to be represented per printable character. (The extra 65th character, "=", is used to signify a special processing function.)¶

NOTE: This subset has the important property that it is represented identically in all versions of ISO 646, including US-ASCII, and all characters in the subset are also represented identically in all versions of EBCDIC. Other popular encodings, such as the encoding used by the uuencode utility, Macintosh binhex 4.0 [RFC-1741], and the base85 encoding specified as part of Level 2 PostScript, do not share these properties, and thus do not fulfill the portability requirements a binary transport encoding for mail must meet.¶

The encoding process represents 24-bit groups of input bits as output strings of 4 encoded characters. Proceeding from left to right, a 24-bit input group is formed by concatenating 3 8bit input groups. These 24 bits are then treated as 4 concatenated 6-bit groups, each of which is translated into a single digit in the base64 alphabet. When encoding a bit stream via the base64 encoding, the bit stream must be presumed to be ordered with the most-significant-bit first. That is, the first bit in the stream will be the high-order bit in the first 8bit byte, and the eighth bit will be the low-order bit in the first 8bit byte, and so on.¶

Each 6-bit group is used as an index into an array of 64 printable characters. The character referenced by the index is placed in the output string. These characters, identified in Table 1, below, are selected so as to be universally representable, and the set excludes characters with particular significance to SMTP (e.g., ".", CR, LF) and to the multipart boundary delimiters defined in RFC 2046 (e.g., "-").¶

The encoded output stream must be represented in lines of no more than 76 characters each. All line breaks or other characters not found in Table 1 must be ignored by decoding software. In base64 data, characters other than those in Table 1, line breaks, and other white space probably indicate a transmission error, about which a warning message or even a message rejection might be appropriate under some circumstances.¶

Special processing is performed if fewer than 24 bits are available at the end of the data being encoded. A full encoding quantum is always completed at the end of a body. When fewer than 24 input bits are available in an input group, zero bits are added (on the right) to form an integral number of 6-bit groups. Padding at the end of the data is performed using the "=" character. Since all base64 input is an integral number of octets, only the following cases can arise: (1) the final quantum of encoding input is an integral multiple of 24 bits; here, the final unit of encoded output will be an integral multiple of 4 characters with no "=" padding, (2) the final quantum of encoding input is exactly 8 bits; here, the final unit of encoded output will be two characters followed by two "=" padding characters, or (3) the final quantum of encoding input is exactly 16 bits; here, the final unit of encoded output will be three characters followed by one "=" padding character.¶

Because it is used only for padding at the end of the data, the occurrence of any "=" characters may be taken as evidence that the end of the data has been reached (without truncation in transit). No such assurance is possible, however, when the number of octets transmitted was a multiple of three and no "=" characters are present.¶

Any characters outside of the base64 alphabet are to be ignored in base64-encoded data.¶

Care must be taken to use the proper octets for line breaks if base64 encoding is applied directly to text material that has not been converted to canonical form. In particular, text line breaks must be converted into CRLF sequences prior to base64 encoding. The important thing to note is that this may be done directly by the encoder rather than in a prior canonicalization step in some implementations.¶

NOTE: There is no need to worry about quoting potential boundary delimiters within base64-encoded bodies within multipart entities because no hyphen characters are used in the base64 encoding.¶