MIME overview

A MIME-formatted message contains one or several MIME sections. MIME headers specify how multiple MIME sections are to be interpreted as a whole (whether they are attached together; whether they are alternative representations of the same content; or something even more esoteric). This manual page gives a very brief, terse, overview of basic MIME concepts. The description is biased towards describing the functionality of the makemime utility. See RFC 2045, RFC 2046, RFC 2047, RFC 2048, and RFC 2049 for a formal definition of MIME-formatted messages.

Each file in a MIME message is encoded as a single MIME section. A MIME section consists of at least one header line, "Content-Type:". The "Content-Type:" header gives the type of the data ontained in the file. Other header lines may also be present. Their relative order does not matter. MIME headers are followed by a blank line, then the contents of the file, encoded appropriately. All MIME sections generated by makemime will always contain another header, "Content-Transfer-Encoding:". This header gives the encoding method used for the file; it is an optional header, but makemime always creates it.

The MIME encoding method defaults to "7bit" if this header is absent. 7bit encoding is only suitable for plain text messages in the US-ASCII character set. The "8bit" encoding method is used by plain text messages in other character sets that use octets with the high bit set. An alternative to 8bit encoding is "quoted-printable". The "base64" encoding method is used for files containing binary data (anything other than plain text).

MIME sections that contain text messages have their "Content-Type:" header set to "text/plain"; or "text/html" for HTML messages. There are also several other, rare, content types that can be used. MIME sections that contain other kinds of data will use some other, appropriate "Content-Type:" header, such as "image/gif", or "audio/x-wav".

MIME sections that contain textual content may also use the base64 encoding method, they are not required to use 7bit, 8bit, or quoted-printable. "text/pdf" sections, that contain PDF files, typically contain binary data and must use the base64 encoding. Consequently, MIME sections that typically contain binary data, such as image/gif and audio/x-wav, are free to use encodings other than base64, as long as all the data can be represented by printable characters (but, in practice, that never happens).

MIME sections may also contain other, optional, headers such as "Content-Disposition:", "Content-ID:", and "Content-Name:". Consult the appropriate RFCs for the specific usage of these headers. These headers can be added by makemime by using the -a option, as described below. These headers play no part in creating the overall structure of a MIME-encoded message, and makemime does not care much about these headers. It simply includes them, and their content, upon request.

Multiple files are formatted as a single message MIME message in two steps: first, by creating a MIME section for each file; and then creating a single MIME section that contains other MIME sections. A "multipart/mixed" MIME section contains a collection of MIME sections that represent different objects, attached together. A "multipart/alternative" MIME section contains a collection of MIME sections which are alternative representations of the same object, such as an HTML and a plain text version of the same message. Other "multipart" MIME sections also exist, and their usage is defined by their respective RFCs.

Creating a single MIME section

The -c option reads filename, encodes it appropriately, adds the "Content-Type: type" and "Content-Transfer-Encoding:" MIME headers, then writes the result to standard output. type can be any valid MIME type, except for multipart. Setting filename to "-" reads from standard input. Setting filename to "&n" reads from file descriptor #n.

The -C option sets the MIME charset attribute for text/plain content. The -N option sets the name attribute for Content-Type:.

encoding argument should be specified. It's more efficient to do so. encoding must be one of the following: 7bit, 8bit, quoted-printable, or base64.

If encoding is not specified, makemime reads the filename twice - once to figure out the best encoding method, and the second time to encode filename. If filename is a pipe makemime creates a temporary file, which is not very efficient if filename is large. However letting makemime pick the encoding method is more convenient if filename is relatively small.

Another possibility is to omit encoding and set type to auto. This combination sets "Content-Type:" to either text/plain, or application/octet-stream, based on the selected encoding.

By default the encoded MIME section is written to standard output. The -o option writes the MIME section to outputfile. outputfile may be "&n", which writes the MIME section to a pipe on file descriptor #n.

makemime does not generate any other headers. Particularly, the "Mime-Version:" header is required for MIME-formatted E-mail messages. Additional headers are specified by the -a option, which may be used multiple times to insert multiple headers. makemime doesn't do anything with them except to insert the headers into the generated MIME section.

Note that "Mime-Version:" is only required for the top level MIME section. This header is not required for individual MIME sections that are later combined into a multipart MIME collection.

Creating a multipart MIME collection

The -m option is identical to the -c option, except for three differences.

type must be either "multipart/mixed", "multipart/alternative", or some other MIME multipart content type. Additionally, "encoding" can only be "7bit" or "8bit", and will default to "8bit" if not specified. Finally, filename must be a MIME-formatted section, NOT a regular file. Usually filename is created by a previous invocation of makemime (it can also be a pipe, like the -c option), but it can be created via any other means.

The -m option creates an initial multipart MIME collection, that contains only one MIME section, taken from filename. The collection is written to standard output, or the pipe or to outputfile.

Creating a multipart MIME section

This option adds a MIME section to an existing MIME collection. file1 must be a MIME collection that was previously created by the -m option. file2 must be a MIME section that was previously created by the -c option. The -j options adds the MIME section in file2 to the MIME collection in file1. The result is written to standard output or to outputfile.

file1 and/or file2 may be "@&n" which reads from file descriptor #n. The outputfile may also specify a file descriptor.

file1 and file2 should ideally be created by makemime as well. It's also possible to use MIME-formatted files created by other software, but with some degree of care. makemime is not intended to be a MIME parser, but a MIME generator. However some amount of MIME parsing is necessary to append a MIME section to an existing MIME collection. makemime's parsing is sufficient for appending a new section to a MIME collection, as long as the MIME headers in the MIME collections are straightforward. Very convoluted MIME headers may confuse makemime, and it may not be able to handle them.

Recursive MIME collections

MIME collection may contain other MIME collections as well as MIME sections. The -m and the -j options may use a multipart MIME collection in place of a MIME section automatically because a multipart MIME collection is just a special type of a MIME section. The following example encodes a text message that can be alternatively represented as HTML or plain text, with some additional attachments:

1. Create a MIME collection that has a text/plain and a text/html MIME section.

2. Create a MIME collection consisting of the MIME section generated in step one, plus additional MIME sections containing other attachments.

For example:

output.msg now contains the complete MIME collection. Just add the Subject:, From:, and To: headers (can also be done by additional -a options, of course), and send it on its way.

Building complex MIME encodings

There are several different ways to build complete MIME encodings from multiple MIME sections. One way is to use temporary files to create MIME sections, then combine them together into a single MIME collection. A slightly more complicated approach involves setting up pipes between multiple makemime processes, in order to avoid using temporary files.

This can be done manually, by hand. It is also possible to have makemime do this automatically. makemime will set up these pipes and run multiple instances of itself to create a single MIME collection, with multiple attachments of complexity limited only by your system's limit on the maximum number of open files and pipes.

Any file that's read by the -c, -m, and -j options ( -o specifies a file to create, and doesn't count) may be replaced by a single argument containing a left parenthesis, additional options, then a single argument containing a right parenthesis. A single invocation of makemime can only use one -c, -m, or -j option. However, another -c, -m, or -j option may be specified inside the left and the right parenthesis, and its output is used in place of the file it replaced. In the previous example the third and the fourth invocation of makemime can be replaced with the following command:

Note that the parenthesis must be backslashed, to avoid their special meaning to the shell. An equivalent argument file would have the following contents:

These constructs can be arbitrarily nested, and are limited by the amount of available memory and resources. The entire sequence in the previous section is equivalent to the following command:

An equivalent argument file would be: