A User Agent Configuration Mechanism

For Multimedia Mail Format Information

Status of This Memo

This memo provides information for the Internet community. It does not specify an Internet standard. Distribution of this memo is unlimited.

Abstract

This memo suggests a file format to be used to inform multiple mail reading user agent programs about the locally-installed facilities for handling mail in various formats. The mechanism is explicitly designed to work with mail systems based Internet mail as defined by RFC's 821 (STD 10), 822 (STD 11), 934, 1049 (STD 11), 1113, and the Multipurpose Internet Mail Extensions, known as MIME. However, with some extensions it could probably be made to work for X.400-based mail systems as well. The format and mechanism are proposed in a manner that is generally operating-system independent. However, certain implementation details will inevitably reflect operating system differences, some of which will have to be handled in a uniform manner for each operating system. This memo makes such situations explicit, and, in an appendix, suggests a standard behavior under the UNIX operating system.

Introduction

The electronic mail world is in the midst of a transition from single-part text-only mail to multi-part, multi-media mail. In support of this transition, various extensions to RFC 821 and RFC 822 have been proposed and/or adopted, notably including MIME [RFC-1521]. Various parties have demonstrated extremely high-functionality multimedia mail, but the problem of mail interchange between different user agents has been severe. In general, only text messages have been shared between user agents that were not explicitly designed to work together. This limitation is not compatible with a smooth transition to a multi-media mail world.

One approach to this transition is to modify diverse sets of mail reading user agents so that, when they need to display mail of an unfamiliar (non-text) type, they consult an external file for information on how to display that file. That file might say, for example, that if the content-type of a message is "foo" it can be displayed to the user via the "displayfoo" program.

This approach means that, with a one-time modification, a wide variety of mail reading programs can be given the ability to display a wide variety of types of message. Moreover, extending the set of media types supported at a site becomes a simple matter of installing a binary and adding a single line to a configuration file. Crucial to this scheme, however, is that all of the user agents agree on a common representation and source for the configuration file. This memo proposes such a common representation.

Location of Configuration Information

Each user agent must clearly obtain the configuration information from a common location, if the same information is to be used to configure all user agents. However, individual users should be able to override or augment a site's configuration. The configuration information should therefore be obtained from a designated set of locations. The overall configuration will be obtained through the virtual concatenation of several individual configuration files known as mailcap files. The configuration information will be obtained from the FIRST matching entry in a mailcap file, where "matching" depends on both a matching content-type specification, an entry containing sufficient information for the purposes of the application doing the searching, and the success of any test in the "test=" field, if present.

The precise location of the mailcap files is operating-system dependent. A standard location for UNIX is specified in Appendix A.

Overall Format of a Mailcap File

Each mailcap file consists of a set of entries that describe the proper handling of one media type at the local site.

For example, one line might tell how to display a message in Group III fax format. A mailcap file consists of a sequence of such individual entries, separated by newlines (according to the operating system's newline conventions). Blank lines and lines that start with the "#" character (ASCII 35) are considered comments, and are ignored. Long entries may be continued on multiple lines if each non-terminal line ends with a backslash character ('\', ASCII 92), in which case the multiple lines are to be treated as a single mailcap entry. Note that for such "continued" lines, the backslash must be the last character on the line to be continued.

Thus the overall format of a mailcap file is given, in the modified BNF of RFC 822, as:

         Mailcap-File = *Mailcap-Line
         
         Mailcap-Line = Comment / Mailcap-Entry
         
         Comment = NEWLINE  /  "#" *CHAR NEWLINE

NEWLINE = <newline as defined by OS convention>

Note that the above specification implies that comments must appear on lines all to themselves, with a "#" character as the first character on each comment line.

Format of a Mailcap Entry

Each mailcap entry consists of a number of fields, separated by semi-colons. The first two fields are required, and must occur in the specified order. The remaining fields are optional, and may appear in any order.

The first field is the content-type, which indicates the type of data this mailcap entry describes how to handle. It is to be matched against the type/subtype specification in the "Content-Type" header field of an Internet mail message. If the subtype is specified as "*", it is intended to match all subtypes of the named content-type.

The second field, view-command, is a specification of how the message or body part can be viewed at the local site. Although the syntax of this field is fully specified, the semantics of program execution are necessarily somewhat operating system dependent. UNIX semantics are given in Appendix A.

The optional fields, which may be given in any order, are as follows:

-- The "compose" field may be used to specify a program that can be

used to compose a new body or body part in the given format. Its intended use is to support mail composing agents that support the composition of multiple types of mail using external composing agents. As with the view-command, the semantics of program execution are operating system dependent, with UNIX semantics specified in Appendix A. The result of the composing program may be data that is not yet suitable for mail transport -- that is, a Content-Transfer-Encoding may need to be applied to the data.

-- The "composetyped" field is similar to the "compose" field, but is

to be used when the composing program needs to specify the

Content-type header field to be applied to the composed data. The "compose" field is simpler, and is preferred for use with existing (non-mail-oriented) programs for composing data in a given format. The "composetyped" field is necessary when the Content-type information must include auxilliary parameters, and the composition program must then know enough about mail formats to produce output that includes the mail type information.

-- The "edit" field may be used to specify a program that can be used

to edit a body or body part in the given format. In many cases, it may be identical in content to the "compose" field, and shares the operating-system dependent semantics for program execution.

-- The "print" field may be used to specify a program that can be

used to print a message or body part in the given format. As with the view-command, the semantics of program execution are operating system dependent, with UNIX semantics specified in Appendix A.

-- The "test" field may be used to test some external condition

(e.g., the machine architecture, or the window system in use) to determine whether or not the mailcap line applies. It specifies a program to be run to test some condition. The semantics of execution and of the value returned by the test program are operating system dependent, with UNIX semantics specified in Appendix A. If the test fails, a subsequent mailcap entry should be sought. Multiple test fields are not permitted -- since a test can call a program, it can already be arbitrarily complex.

-- The "needsterminal" field indicates that the view-command must be

run on an interactive terminal. This is needed to inform window- oriented user agents that an interactive terminal is needed. (The decision is not left exclusively to the view-command because in some circumstances it may not be possible for such programs to tell whether or not they are on interactive terminals.) The needsterminal command should be assumed to apply to the compose and edit commands, too, if they exist. Note that this is NOT a test -- it is a requirement for the environment in which the program will be executed, and should typically cause the creation of a terminal window when not executed on either a real terminal or a terminal window.

-- The "copiousoutput" field indicates that the output from the

view-command will be an extended stream of output, and is to be interpreted as advice to the UA (User Agent mail-reading program) that the output should be either paged or made scrollable. Note that it is probably a mistake if needsterminal and copiousoutput are both specified.

-- The "description" field simply provides a textual description,

optionally quoted, that describes the type of data, to be used optionally by mail readers that wish to describe the data before offering to display it.

-- The "textualnewlines" field, if set to any non-zero value,

      indicates that this type of data is line-oriented and that, if
      encoded in base64, all newlines should be converted to canonical
      form (CRLF) before encoding, and will be in that form after
      decoding.  In general, this field is needed only if there is
      line-oriented data of some type other than text/* or non-line-
      oriented data that is a subtype of text.

-- The "x11-bitmap" field names a file, in X11 bitmap (xbm) format,

which points to an appropriate icon to be used to visually denote the presence of this kind of data.

-- The "nametemplate" field gives a file name format, in which %s

will be replaced by a short unique string to give the name of the temporary file to be passed to the viewing command. This is only expected to be relevant in environments where filename extensions are meaningful, e.g., one coulld specify that a GIF file being passed to a gif viewer should have a name eding in ".gif" by using "nametemplate=%s.gif".

Any other fields beginning with "x-" may be included for local or mailer-specific extensions of this format. Implementations should simply ignore all such unrecognized fields to permit such extensions, some of which might be standardized in a future version of this document.

Some of the fields above, such as "needsterminal", apply to the actions of the view-command, edit-command, and compose-command, alike. In some unusual cases, this may not be desirable, but differentiation can be accomplished via separate mailcap entries, taking advantage of the fact that subsequent mailcap entries are searched if an earlier mailcap entry does not provide enough information:

       application/postscript; ps-to-terminal %s;\ needsterminal
       application/postscript; ps-to-terminal %s; \compose=idraw %s

In RFC 822 modified BNF, the following grammar describes a mailcap entry:

         Mailcap-Entry = typefield ; view-command
                             [";" 1#field]
         
         typefield = propertype / implicit-wild

propertype = type "/" wildsubtype

         implicitwild = type

wildsubtype = subtype / "*"

         view-command = mtext
         
         mtext = *mchar

mchar = schar / qchar

schar = * <any CHAR except ";","\", and CTLS>

qchar = "\" CHAR ; may quote any char

field = flag / namedfield

namedfield = fieldname "=" mtext

flag = "needsterminal" ; All these literals are to

/ "copiousoutput" ; be interpreted as

/ x-token ; case-insensitive

         fieldname =    / "compose"      ;Also all of these
                        / "composetyped" ;are case-insensitive.
                        / "print"
                        / "edit"
                        / "test"
                        / "x11-bitmap"
                        / "textualnewlines"
                        / "description"
                        / x-token

Note that "type", "subtype", and "x-token" are defined in MIME. Note also that while the definition of "schar" includes the percent sign, "%", this character has a special meaning in at least the UNIX semantics, and will therefore need to be quoted as a qchar to be used literally.

Acknowledgements

The author wishes to thank Malcolm Bjorn Gillies, Dan Heller, Olle Jaernefors, Keith Moore, Luc Rooijakkers, and the other members of the IETF task force on mail extensions for their comments on earlier versions of this draft. If other acknowledgements were neglected, please let me know, as it was surely accidental.

Security Considerations

Security issues are not discussed in this memo. However, the use of the mechanisms described in this memo can make it easier for implementations to slip into the kind of security problems discussed in the MIME document. Implementors and mailcap administrators should be aware of these security considerations, and in particular should exercise caution in the choice of programs to be listed in a mailcap file for automatic execution.

Author's Address

   Nathaniel S. Borenstein
   MRE 2D-296, Bellcore
   445 South St.
   Morristown, NJ 07962-1910
   
   EMail: nsb@bellcore.com
   Phone: +1 201 829 4270
   Fax:  +1 201 829 7019

Appendix A: Implementation Details for UNIX

Although this memo fully specifies a syntax for "mailcap" files, the semantics of the mailcap file are of necessity operating-system dependent in four respects. In order to clarify the intent, and to promote a standard usage, this appendix proposes a UNIX semantics for these four cases. If a mailcap mechanism is implemented on non-UNIX systems, similar semantic decisions should be made and published.

Location of the Mailcap File(s)

For UNIX, a path search of mailcap files is specified. The default path search is specified as including at least the following:

   $HOME/.mailcap:/etc/mailcap:/usr/etc/mailcap:/usr/local/etc/mailcap

However, this path may itself be overridden by a path specified by the MAILCAPS environment variable.

Semantics of executable commands

Several portions of a mailcap entry specify commands to be executed. In particular, the mandatory second fie ld, the view-command, takes a command to be executed, as do the optional print, edit, test, and compose fields.

On a UNIX system, such commands will each be a full shell command line, including the path name for a program and its arguments. (Because of differences in shells and the implementation and behavior of the same shell from one system to another, it is specified that the command line be intended as input to the Bourne shell, i.e., that it is implicitly preceded by "/bin/sh -c " on the command line.)

The two characters "%s", if used, will be replaced by the name of a file for the actual mail body data. In the case of the edit adn view-command, the body part will be passed to this command as standard input unless one or more instances of "%s" appear in the view-command, in which case %s will be replaced by the name of a file containing the body part, a file which may have to be created before the view-command program is executed. (Such files cannot be presumed to continue to exist after the view-command program exits. Thus a view-command that wishes to exit and continue processing in the background should take care to save the data first.) In the case of the compose and composetyped commands, %s should be replaced by the name of a file to which the composed data should be written by the programs named in the compose or composedtyped commands. Thus, the calling program will look in that file later in order to retrieve the composed data. If %s does not appear in the compose or composetyped commands, then the composed data will be assumed to be written by the composing programs to standard output.

Furthermore, any occurrence of "%t" will be replaced by the content- type and subtype specification. (That is, if the content-type is "text/plain", then %t will be replaced by "text/plain".) A literal % character may be quoted as \%. Finally, named parameters from the Content-type field may be placed in the command execution line using "%{" followed by the parameter name and a closing "}" character. The entire parameter should appear as a single command line argument, regardless of embedded spaces. Thus, if the message has a Content- type line of:

         Content-type:  multipart/mixed; boundary=42

and the mailcap file has a line of:

         multipart/*; /usr/local/bin/showmulti \
           %t %{boundary}
   
   then the equivalent  of  the  following  command  should  be
   executed:
   
        /usr/local/bin/showmulti multipart/mixed 42

If the content-type is "multipart" (any subtype), then the two characters "%n" will be replaced by an integer giving the number of sub-parts within the multipart entity. Also, the two characters "%F" will be replaced by a set of arguments, twice as many arguments as the number of sub-parts, consisting of alternating content-types and file names for each part in turn. Thus if multipart entity has three parts, "%F" will be replaced by the equivalent of "content-type1 file-name1 content-type2 file-name2 content-type3 file-name3".

Semantics of the "test" field

The "test" field specifies a program to be used to test whether or not the current mailcap line applies. This can be used, for example, to have a mailcap line that only applies if the X window system is running, or if the user is running on a SPARCstation with a /dev/audio. The value of the "test" field is a program to run to test such a condition. The precise program to run and arguments to give it are determined as specified in the previous section. The test program should return an exit code of zero if the condition is true, and a non-zero code otherwise.

Semantics of the "compose" field

On UNIX, the composing program is expected to produce a data stream for such a body part as its standard output. The program will be executed with the command line arguments determined as specified above. The data returned via its standard output will be given a Content-Type field that has no supplementary parameters. For example, the following mailcap entry:

        audio/basic; /usr/local/bin/showaudio %t
          compose = /usr/local/bin/recordaudio

would result in tagging the data composed by the "recordaudio" program as:

Content-Type:

                      audio/basic

If this is unacceptable -- for example, in the case of multipart mail a "boundary" parameter is required -- then the "compose" field cannot be used. Instead, the "composetyped" field should be used in the mailcap file.

Semantics of the "composetyped" field

The "composetyped" filed is much like the "compose" field, except that it names a composition program that produces, not raw data, but data that includes a MIME-conformant type specification. The program will be executed with the command line arguments determined as specified above. The data returned via its standard output must begin with a Content-Type header, followed optionally by other Content-* headers, and then by a blank line and the data. For example, the following mailcap entry:

        multipart/mixed; /usr/local/bin/showmulti %t \
          %{boundary}; \
          composetyped = /usr/local/bin/makemulti

would result in executing the "makemulti" program, which would be expected to begin its output with a line of the form:

        Content-Type: multipart/mixed; boundary=foobar

Note that a composition program need not encode binary data in base64 or quoted-printable. It remains the responsibility of the software calling the composition program to encode such data as necessary. However, if a composing program does encode data, which is not encouraged, it should announce that fact using a Content-Transfer- Encoding header in the standard manner defined by MIME. Because such encodings must be announced by such a header, they are an option only for composetyped programs, not for compose programs.

Appendix B: Sample Mailcap File

The following is an example of a mailcap file for UNIX that demonstrates most of the syntax above. It contains explanatory comments where necessary.

         # Mailcap file for Bellcore lab 214.
         #
         # The next line sends "richtext" to the richtext
         program
         text/richtext; richtext %s; copiousoutput
         #
         # Next, basic u-law audio
         audio/*; showaudio; test=/usr/local/bin/hasaudio
         #
         # Next, use the xview program to handle several image
         formats
         image/*; xview %s; test=/usr/local/bin/RunningX
         #
         # The ATOMICMAIL interpreter uses curses, so needs a
         terminal
         application/atomicmail; /usr/local/bin/atomicmail %s; \
             needsterminal
         #
         # The next line handles Andrew format,
         #   if ez and ezview are installed
         x-be2; /usr/andrew/bin/ezview %s; \
            print=/usr/andrew/bin/ezprint %s ; \
            compose=/usr/andrew/bin/ez -d %s \;
            edit=/usr/andrew/bin/ez -d %s; \;
            copiousoutput
         #
         # The next silly example demonstrates the use of
         quoting
         application/*; echo "This is \"%t\" but \
            is 50 \% Greek to me" \; cat %s; copiousoutput

Appendix C: A Note on Format Translation

It has been suggested that another function of a mailcap-like mechanism might be to specify the locally available tools for document format translation. For example, the file could designate a program for translating from format A to format B, another for translating from format B to format C, and finally a mechanism for displaying format C. Although this mechanism would be somewhat richer than the current mailcap file, and might conceivably also have utility at the message transport layer, it significantly complicates the processing effort necessary for a user agent that simply wants to display a message in format A. Using the current, simpler, mailcap scheme, a single line could tell such a user agent to display A- format mail using a pipeline of translators and the C-format viewer. This memo resists the temptation to complicate the necessary processing for a user agent to accomplish this task. Using the mailcap format defined here, it is only necessary to find the correct single line in a mailcap file, and to execute the command given in that line.

References

     [RFC-822] Crocker, D., "Standard for the format of ARPA Internet
     text messages", STD 11, RFC 822, UDEL, August 1982.
     
     [RFC-1521] Borenstein, N., and N.  Freed, "MIME (Multipurpose
     Internet Mail Extensions) Part One: Mechanisms for Specifying and
     Describing the Format of Internet Message Bodies", RFC 1521,
     Bellcore, Innosoft, September 1993.