Table of Contents
Quite a bit of effort has been made to make Mutt the premier text-mode
MIME MUA. Every effort has been made to provide the functionality that
the discerning MIME user requires, and the conformance to the standards
wherever possible. When configuring Mutt for MIME, there are two extra
types of configuration files which Mutt uses. One is the
mime.types file, which contains the mapping of file
extensions to IANA MIME types. The other is the
mailcap file, which specifies the external commands
to use for handling specific MIME types.
MIME is short for “Multipurpose Internet Mail Extension” and describes mechanisms to internationalize and structure mail messages. Before the introduction of MIME, messages had a single text part and were limited to us-ascii header and content. With MIME, messages can have attachments (and even attachments which itself have attachments and thus form a tree structure), nearly arbitrary characters can be used for sender names, recipients and subjects.
Besides the handling of non-ascii characters in message headers, to Mutt
the most important aspect of MIME are so-called MIME types. These are
constructed using a major and
minor type separated by a forward slash. These
specify details about the content that follows. Based upon these, Mutt
decides how to handle this part. The most popular major type is
“text” with minor types for plain text,
HTML and various other formats. Major types also exist for images,
audio, video and of course general application data (e.g. to separate
cryptographically signed data with a signature, send office documents,
and in general arbitrary binary data). There's also the
multipart major type which represents the root of a
subtree of MIME parts. A list of supported MIME types can be found in
Table 5.1, “Supported MIME types”.
MIME also defines a set of encoding schemes for transporting MIME
content over the network: 7bit,
8bit, quoted-printable,
base64 and binary. There're some
rules when to choose what for encoding headers and/or body (if needed),
and Mutt will in general make a good choice.
Mutt does most of MIME encoding/decoding behind the scenes to form messages conforming to MIME on the sending side. On reception, it can be flexibly configured as to how what MIME structure is displayed (and if it's displayed): these decisions are based on the content's MIME type. There are three areas/menus in dealing with MIME: the pager (while viewing a message), the attachment menu and the compose menu.
When you select a message from the index and view it in the pager, Mutt
decodes as much of a message as possible to a text representation. Mutt
internally supports a number of MIME types, including the
text major type (with all minor types), the
message/rfc822 (mail messages) type and some
multipart types. In addition, it recognizes a variety
of PGP MIME types, including PGP/MIME and
application/pgp.
Mutt will denote attachments with a couple lines describing them. These lines are of the form:
[-- Attachment #1: Description --] [-- Type: text/plain, Encoding: 7bit, Size: 10000 --]
Where the Description is the description or filename given for the attachment, and the Encoding is one of the already mentioned content encodings.
If Mutt cannot deal with a MIME type, it will display a message like:
[-- image/gif is unsupported (use 'v' to view this part) --]
The default binding for <view-attachments> is
“v”, which displays the attachment menu for a message. The
attachment menu displays a list of the attachments in a message. From
the attachment menu, you can save, print, pipe, delete, and view
attachments. You can apply these operations to a group of attachments
at once, by tagging the attachments and by using the
<tag-prefix> operator. You can also reply to
the current message from this menu, and only the current attachment (or
the attachments tagged) will be quoted in your reply. You can view
attachments as text, or view them using the mailcap viewer definition
(the mailcap mechanism is explained later in detail).
Finally, you can apply the usual message-related functions (like <resend-message>,
and the <reply> and
<forward> functions) to attachments of type
message/rfc822.
See table Table 9.7, “Default Attachment Menu Bindings” for all available functions.
The compose menu is the menu you see before you send a message. It allows you to edit the recipient list, the subject, and other aspects of your message. It also contains a list of the attachments of your message, including the main body. From this menu, you can print, copy, filter, pipe, edit, compose, review, and rename an attachment or a list of tagged attachments. You can also modifying the attachment information, notably the type, encoding and description.
Attachments appear as follows by default:
- 1 [text/plain, 7bit, 1K] /tmp/mutt-euler-8082-0 <no description> 2 [applica/x-gunzip, base64, 422K] ~/src/mutt-0.85.tar.gz <no description>
The “-” denotes that Mutt will delete the file after
sending (or postponing, or canceling) the message. It can be toggled
with the <toggle-unlink> command (default: u).
The next field is the MIME content-type, and can be changed with the
<edit-type> command (default: ^T). The next
field is the encoding for the attachment, which allows a binary message
to be encoded for transmission on 7bit links. It can be changed with
the <edit-encoding> command (default: ^E). The
next field is the size of the attachment, rounded to kilobytes or
megabytes. The next field is the filename, which can be changed with
the <rename-file> command (default: R). The
final field is the description of the attachment, and can be changed
with the <edit-description> command (default:
d). See $attach_format for a full
list of available expandos to format this display to your needs.
To get most out of MIME, it's important that a MIME part's content type
matches the content as closely as possible so that the recipient's
client can automatically select the right viewer for the
content. However, there's no reliable for Mutt to know how to detect
every possible file type. Instead, it uses a simple plain text mapping
file that specifies what file extension corresponds to what MIME
type. This file is called mime.types.
When you add an attachment to your mail message, Mutt searches your
personal mime.types file at
$HOME/.mime.types, and then the system
mime.types file at
/usr/local/share/mutt/mime.types or
/etc/mime.types
Each line starts with the full MIME type, followed by a space and space-separated list of file extensions. For example you could use:
Example 5.1. mime.types
application/postscript ps eps application/pgp pgp audio/x-aiff aif aifc aiff
A sample mime.types file comes with the Mutt
distribution, and should contain most of the MIME types you are likely
to use.
If Mutt can not determine the MIME type by the extension of the file you
attach, it will run the command specified in
$mime_type_query_command.
If that command is not specified, Mutt will look at the file. If the file
is free of binary information, Mutt will assume that the file is plain text,
and mark it as text/plain. If the file contains binary
information, then Mutt will mark it as
application/octet-stream. You can change the MIME
type that Mutt assigns to an attachment by using the
<edit-type> command from the compose menu
(default: ^T), see Table 5.1, “Supported MIME types” for supported
major types. Mutt recognizes all of these if the appropriate entry is
found in the mime.types file. Non-recognized mime
types should only be used if the recipient of the message is likely to
be expecting such attachments.
Table 5.1. Supported MIME types
| MIME major type | Standard | Description |
|---|---|---|
application | yes | General application data |
audio | yes | Audio data |
image | yes | Image data |
message | yes | Mail messages, message status information |
model | yes | VRML and other modeling data |
multipart | yes | Container for other MIME parts |
text | yes | Text data |
video | yes | Video data |
chemical | no | Mostly molecular data |
MIME types are not arbitrary, they need to be assigned by IANA.
Mutt supports RFC 1524 MIME Configuration, in particular the Unix specific format specified in Appendix A of RFC 1524. This file format is commonly referred to as the “mailcap” format. Many MIME compliant programs utilize the mailcap format, allowing you to specify handling for all MIME types in one place for all programs. Programs known to use this format include Firefox, lynx and metamail.
In order to handle various MIME types that Mutt doesn't have built-in support for, it parses a series of external configuration files to find an external handler. The default search string for these files is a colon delimited list containing the following files:
$HOME/.mailcap
$PKGDATADIR/mailcap
$SYSCONFDIR/mailcap
/etc/mailcap
/usr/etc/mailcap
/usr/local/etc/mailcap
where $HOME is your home directory. The
$PKGDATADIR and the $SYSCONFDIR
directories depend on where Mutt is installed: the former is the default
for shared data, the latter for system configuration files.
The default search path can be obtained by running the following command:
mutt -nF /dev/null -Q mailcap_path
In particular, the metamail distribution will install a mailcap file,
usually as /usr/local/etc/mailcap, which contains
some baseline entries.
A mailcap file consists of a series of lines which are comments, blank, or definitions.
A comment line consists of a # character followed by anything you want.
A blank line is blank.
A definition line consists of a content type, a view command, and any number of optional fields. Each field of a definition line is divided by a semicolon “;” character.
The content type is specified in the MIME standard
“type/subtype” notation. For example,
text/plain, text/html,
image/gif, etc. In addition, the mailcap format
includes two formats for wildcards, one using the special
“*” subtype, the other is the implicit wild, where you only
include the major type. For example, image/*, or
video will match all image types and video types,
respectively.
The view command is a Unix command for viewing the type specified. There
are two different types of commands supported. The default is to send
the body of the MIME message to the command on stdin. You can change
this behavior by using %s as a parameter to your view
command. This will cause Mutt to save the body of the MIME message to a
temporary file, and then call the view command with the
%s replaced by the name of the temporary file. In
both cases, Mutt will turn over the terminal to the view program until
the program quits, at which time Mutt will remove the temporary file if
it exists. This means that mailcap does not work
out of the box with programs which detach themselves from the terminal
right after starting, like open on Mac OS X. In order
to nevertheless use these programs with mailcap, you probably need
custom shell scripts.
So, in the simplest form, you can send a text/plain
message to the external pager more on standard input:
text/plain; more
Or, you could send the message as a file:
text/plain; more %s
Perhaps you would like to use lynx to interactively view a
text/html message:
text/html; lynx %s
In this case, lynx does not support viewing a file from standard input,
so you must use the %s syntax.
Some older versions of lynx contain a bug where they will
check the mailcap file for a viewer for text/html.
They will find the line which calls lynx, and run it. This causes lynx
to continuously spawn itself to view the object.
On the other hand, maybe you don't want to use lynx interactively, you
just want to have it convert the text/html to
text/plain, then you can use:
text/html; lynx -dump %s | more
Perhaps you wish to use lynx to view text/html files,
and a pager on all other text formats, then you would use the following:
text/html; lynx %s text/*; more
The interpretation of shell meta-characters embedded in MIME parameters
can lead to security problems in general. Mutt tries to quote
parameters in expansion of %s syntaxes properly, and
avoids risky characters by substituting them, see the $mailcap_sanitize variable.
Although Mutt's procedures to invoke programs with mailcap seem to be safe, there are other applications parsing mailcap, maybe taking less care of it. Therefore you should pay attention to the following rules:
Keep the %-expandos away from shell quoting. Don't quote them with single or double quotes. Mutt does this for you, the right way, as should any other program which interprets mailcap. Don't put them into backtick expansions. Be highly careful with evil statements, and avoid them if possible at all. Trying to fix broken behavior with quotes introduces new leaks — there is no alternative to correct quoting in the first place.
If you have to use the %-expandos' values in context where you need
quoting or backtick expansions, put that value into a shell variable and
reference the shell variable where necessary, as in the following
example (using $charset inside the backtick expansion
is safe, since it is not itself subject to any further expansion):
text/test-mailcap-bug; cat %s; copiousoutput; test=charset=%{charset} \
&& test "`echo $charset | tr '[A-Z]' '[a-z]'`" != iso-8859-1
In addition to the required content-type and view command fields, you can add semi-colon “;” separated fields to set flags and other options. Mutt recognizes the following optional fields:
This flag tells Mutt that the command passes possibly large amounts of
text on standard output. This causes Mutt to invoke a pager (either
the internal pager or the external pager defined by the pager variable)
on the output of the view command. Without this flag, Mutt assumes that
the command is interactive. One could use this to replace the pipe to
more in the lynx -dump example in
the Basic section:
text/html; lynx -dump %s ; copiousoutput
This will cause lynx to format the text/html output
as text/plain and Mutt will use your standard pager
to display the results.
Mutt will set the COLUMNS environment variable to
the width of the pager. Some programs make use of this environment
variable automatically. Others provide a command line argument that
can use this to set the output width:
text/html; lynx -dump -width ${COLUMNS:-80} %s; copiousoutput
Note that when using the built-in pager, only entries with this flag will be considered a handler for a MIME type — all other entries will be ignored.
Mutt uses this flag when viewing attachments with auto_view, in order to decide whether it should honor the setting of the $wait_key variable or not. When an attachment is viewed using an interactive program, and the corresponding mailcap entry has a needsterminal flag, Mutt will use $wait_key and the exit status of the program to decide if it will ask you to press a key after the external program has exited. In all other situations it will not prompt you for a key.
This flag specifies the command to use to create a new attachment of a specific MIME type. Mutt supports this from the compose menu.
This flag specifies the command to use to create a new attachment of a specific MIME type. This command differs from the compose command in that Mutt will expect standard MIME headers on the data. This can be used to specify parameters, filename, description, etc. for a new attachment. Mutt supports this from the compose menu.
This flag specifies the command to use to print a specific MIME type. Mutt supports this from the attachment and compose menus.
This flag specifies the command to use to edit a specific MIME type. Mutt supports this from the compose menu, and also uses it to compose new attachments. Mutt will default to the defined $editor for text attachments.
This field specifies the format for the file denoted by
%s in the command fields. Certain programs will
require a certain file extension, for instance, to correctly view a
file. For instance, lynx will only interpret a file as
text/html if the file ends in
.html. So, you would specify lynx as a
text/html viewer with a line in the mailcap file
like:
text/html; lynx %s; nametemplate=%s.html
This field specifies a command to run to test whether this mailcap entry should be used. The command is defined with the command expansion rules defined in the next section. If the command returns 0, then the test passed, and Mutt uses this entry. If the command returns non-zero, then the test failed, and Mutt continues searching for the right entry. Note that the content-type must match before Mutt performs the test. For example:
text/html; firefox -remote 'openURL(%s)' ; test=RunningX text/html; lynx %s
In this example, Mutt will run the program RunningX
which will return 0 if the X Window manager is running, and non-zero if
it isn't. If RunningX returns 0, then Mutt will run
firefox to display the text/html object. If RunningX
doesn't return 0, then Mutt will go on to the next entry and use lynx to
display the text/html object.
When searching for an entry in the mailcap file, Mutt will search for
the most useful entry for its purpose. For instance, if you are
attempting to print an image/gif, and you have the
following entries in your mailcap file, Mutt will search for an entry
with the print command:
image/*; xv %s
image/gif; ; print= anytopnm %s | pnmtops | lpr; \
nametemplate=%s.gif
Mutt will skip the image/* entry and use the
image/gif entry with the print command.
In addition, you can use this with auto_view to denote two
commands for viewing an attachment, one to be viewed automatically, the
other to be viewed interactively from the attachment menu using the
<view-mailcap> function (bound to
“m” by default). In addition, you can then use the test
feature to determine which viewer to use interactively depending on your
environment.
text/html; firefox -remote 'openURL(%s)' ; test=RunningX text/html; lynx %s; nametemplate=%s.html text/html; lynx -dump %s; nametemplate=%s.html; copiousoutput
For auto_view, Mutt
will choose the third entry because of the
copiousoutput tag. For interactive viewing, Mutt
will run the program RunningX to determine if it
should use the first entry. If the program returns non-zero, Mutt will
use the second entry for interactive viewing. The last entry is for
inline display in the pager and the
<view-attach> function in the attachment menu.
Entries with the copiousoutput tag should always be
specified as the last one per type. For non-interactive use, the last
entry will then actually be the first matching one with the tag set.
For non-interactive use, only copiousoutput-tagged
entries are considered. For interactive use, Mutt ignores this tag and
treats all entries equally. Therefore, if not specified last, all
following entries without this tag would never be considered for
<view-attach> because the
copiousoutput before them matched already.
The various commands defined in the mailcap files are passed to the
/bin/sh shell using the system(3)
function. Before the command is passed to /bin/sh
-c, it is parsed to expand various special parameters with
information from Mutt. The keywords Mutt expands are:
As seen in the basic mailcap section, this variable is expanded to a filename specified by the calling program. This file contains the body of the message to view/print/edit or where the composing program should place the results of composition. In addition, the use of this keyword causes Mutt to not pass the body of the message to the view/print/edit program on stdin.
Mutt will expand %t to the text representation of the
content type of the message in the same form as the first parameter of
the mailcap definition line, i.e. text/html or
image/gif.
Mutt will expand this to the value of the specified parameter from the Content-Type: line of the mail message. For instance, if your mail message contains:
Content-Type: text/plain; charset=iso-8859-1
then Mutt will expand %{charset} to
“iso-8859-1”. The default metamail mailcap file uses this
feature to test the charset to spawn an xterm using the right charset to
view the message.
This will be replaced by a literal %.
Mutt does not currently support the %F and
%n keywords specified in RFC 1524. The main purpose
of these parameters is for multipart messages, which is handled
internally by Mutt.
This mailcap file is fairly simple and standard:
# I'm always running X :) video/*; xanim %s > /dev/null image/*; xv %s > /dev/null # I'm always running firefox (if my computer had more memory, maybe) text/html; firefox -remote 'openURL(%s)'
This mailcap file shows quite a number of examples:
# Use xanim to view all videos Xanim produces a header on startup, # send that to /dev/null so I don't see it video/*; xanim %s > /dev/null # Send html to a running firefox by remote text/html; firefox -remote 'openURL(%s)'; test=RunningFirefox # If I'm not running firefox but I am running X, start firefox on the # object text/html; firefox %s; test=RunningX # Else use lynx to view it as text text/html; lynx %s # This version would convert the text/html to text/plain text/html; lynx -dump %s; copiousoutput # I use enscript to print text in two columns to a page text/*; more %s; print=enscript -2Gr %s # Firefox adds a flag to tell itself to view jpegs internally image/jpeg;xv %s; x-mozilla-flags=internal # Use xv to view images if I'm running X # In addition, this uses the \ to extend the line and set my editor # for images image/*;xv %s; test=RunningX; \ edit=xpaint %s # Convert images to text using the netpbm tools image/*; (anytopnm %s | pnmscale -xysize 80 46 | ppmtopgm | pgmtopbm | pbmtoascii -1x2 ) 2>&1 ; copiousoutput # Send excel spreadsheets to my NT box application/ms-excel; open.pl %s
Usage:
auto_view
mimetype
[
mimetype
...]unauto_view {
*
|
mimetype
... }
In addition to explicitly telling Mutt to view an attachment with the MIME viewer defined in the mailcap file from the attachments menu, Mutt has support for automatically viewing MIME attachments while in the pager.
For this to work, you must define a viewer in the mailcap file which
uses the copiousoutput option to denote that it is
non-interactive. Usually, you also use the entry to convert the
attachment to a text representation which you can view in the pager.
You then use the auto_view configuration command to list the content-types that you wish to view automatically. For instance, if you set it to:
auto_view text/html application/x-gunzip \ application/postscript image/gif application/x-tar-gz
...Mutt would try to find corresponding entries for rendering attachments of these types as text. A corresponding mailcap could look like:
text/html; lynx -dump %s; copiousoutput; nametemplate=%s.html
image/*; anytopnm %s | pnmscale -xsize 80 -ysize 50 | ppmtopgm | \
pgmtopbm | pbmtoascii ; copiousoutput
application/x-gunzip; gzcat; copiousoutput
application/x-tar-gz; gunzip -c %s | tar -tf - ; copiousoutput
application/postscript; ps2ascii %s; copiousoutput
unauto_view can be used to remove previous entries from the auto_view list. This can be used with message-hook to autoview messages based on size, etc. “unauto_view *” will remove all previous entries.
The multipart/alternative container type only has
child MIME parts which represent the same content in an alternative
way. This is often used to send HTML messages which contain an
alternative plain text representation.
Mutt has some heuristics for determining which attachment of a
multipart/alternative type to display:
First, Mutt will check the alternative_order list to determine if one of the available types is preferred. It consists of a number of MIME types in order, including support for implicit and explicit wildcards. For example:
alternative_order text/enriched text/plain text \ application/postscript image/*
Next, Mutt will check if any of the types have a defined auto_view, and use that.
Failing that, Mutt will look for any text type.
As a last attempt, Mutt will look for any type it knows how to handle.
To remove a MIME type from the alternative_order list, use the unalternative_order command.
Generating multipart/alternative content is supported
via the
$send_multipart_alternative
quadoption and
$send_multipart_alternative_filter
filter script. The composed text/plain content
will be piped to the filter script's stdin. The output from the
filter script should be the generated mime type of the content, a
blank line, and the content. For example:
text/html <html> <body> Content in html format </body> </html>
A preview of the alternative can be viewed in the compose menu using the functions
<view-alt>,
<view-alt-text> and
<view-alt-mailcap>, bound to
"v", "Esc v" and "V" by default.
If you ever lose track of attachments in your mailboxes, Mutt's attachment-counting and -searching support might be for you. You can make your message index display the number of qualifying attachments in each message, or search for messages by attachment count. You also can configure what kinds of attachments qualify for this feature with the attachments and unattachments commands.
In order to provide this information, Mutt needs to fully MIME-parse all messages affected first. This can slow down operation especially for remote mail folders such as IMAP because all messages have to be downloaded first regardless whether the user really wants to view them or not though using Section 8.2, “Body Caching” usually means to download the message just once.
By default, Mutt will not search inside
multipart/alternative containers. This can be
changed via the $count_alternatives configuration
variable.
The syntax is:
attachments
{ + | - }disposition
mime-type
unattachments
{ + | - }disposition
mime-type
attachments
?
unattachments
*
disposition is the attachment's Content-Disposition
type — either inline or
attachment. You can abbreviate this to
I or A.
The first part of a message or multipart group, if inline, is counted
separately than other inline parts. Specify root
or R for disposition to count
these as attachments. If this first part is of type
multipart/alternative, note that its top-level inline parts are also
counted via root disposition
(if $count_alternatives is
set).
Disposition is prefixed by either a “+” symbol or a “-” symbol. If it's a “+”, you're saying that you want to allow this disposition and MIME type to qualify. If it's a “-”, you're saying that this disposition and MIME type is an exception to previous “+” rules. There are examples below of how this is useful.
mime-type is the MIME type of the attachment you
want the command to affect. A MIME type is always of the format
major/minor, where major describes
the broad category of document you're looking at, and
minor describes the specific type within that
category. The major part of mime-type must be literal text (or the
special token “*”), but the minor part
may be a regular expression. (Therefore,
“*/.*” matches any MIME type.)
The MIME types you give to the attachments directive are a kind of pattern. When you use the attachments directive, the patterns you specify are added to a list. When you use unattachments, the pattern is removed from the list. The patterns are not expanded and matched to specific MIME types at this time — they're just text in a list. They're only matched when actually evaluating a message.
Some examples might help to illustrate. The examples that are not commented out define the default configuration of the lists.
Example 5.2. Attachment counting
# Removing a pattern from a list removes that pattern literally. It # does not remove any type matching the pattern. # # attachments +A */.* # attachments +A image/jpeg # unattachments +A */.* # # This leaves "attached" image/jpeg files on the allowed attachments # list. It does not remove all items, as you might expect, because the # second */.* is not a matching expression at this time. # # Remember: "unattachments" only undoes what "attachments" has done! # It does not trigger any matching on actual messages. # Qualify any MIME part with an "attachment" disposition, EXCEPT for # text/x-vcard and application/pgp parts. (PGP parts are already known # to mutt, and can be searched for with ~g, ~G, and ~k.) # # I've added x-pkcs7 to this, since it functions (for S/MIME) # analogously to PGP signature attachments. S/MIME isn't supported # in a stock mutt build, but we can still treat it specially here. # attachments +A */.* attachments -A text/x-vcard application/pgp.* attachments -A application/x-pkcs7-.* # Discount all MIME parts with an "inline" disposition, unless they're # text/plain. (Why inline a text/plain part unless it's external to the # message flow?) attachments +I text/plain # These two lines make Mutt qualify MIME containers. (So, for example, # a message/rfc822 forward will count as an attachment.) The first # line is unnecessary if you already have "attach-allow */.*", of # course. These are off by default! The MIME elements contained # within a message/* or multipart/* are still examined, even if the # containers themselves don't qualify. #attachments +A message/.* multipart/.* #attachments +I message/.* multipart/.* ## You probably don't really care to know about deleted attachments. attachments -A message/external-body attachments -I message/external-body
Entering the command “attachments ?” as a command will list your current settings in Muttrc format, so that it can be pasted elsewhere.
Entering the command “unattachments *” as a command will Clear all attachment settings.
Usage:
mime_lookup
mimetype
[
mimetype
...]unmime_lookup {
*
|
mimetype
... }
Mutt's mime_lookup list specifies a list of MIME
types that should not be treated according to their
mailcap entry. This option is designed to deal with binary types such
as application/octet-stream. When an attachment's
MIME type is listed in mime_lookup, then the
extension of the filename will be compared to the list of extensions in
the mime.types file. The MIME type associated with
this extension will then be used to process the attachment according to
the rules in the mailcap file and according to any other configuration
options (such as auto_view) specified. Common usage
would be:
mime_lookup application/octet-stream application/X-Lotus-Manuscript
In addition, the unmime_lookup command may be used to
disable this feature for any particular MIME type if it had been set,
for example, in a global .muttrc.