To send a message to one or more people, mail can be invoked with arguments which are the names of people to whom the mail will be sent. You are then expected to type in your message, followed by a ‘control-D' at the beginning of a line. Any flags following the list of recipients, will be passed, together with their arguments, directly to sendmail(1). For example to change your From address to somebody@somewhere.net you can specify: To prevent multiple copies of a message being sent to the same address, duplicate addresses (after alias expansion) are removed from the bcc-addr, cc-addr, and to-addr lists. In addition, addresses on the cc-addr and to-addr lists are removed if they occur on the bcc-addr list and addresses on the cc-addr list are removed if they occur on the to-addr list. If the to-addr list is empty after these deletions, most systems will insert the line “To: undisclosed recipients:;”. The section below Replying to or originating mail, describes some features of mail available to help you compose your letter.

In normal usage mail is given no arguments and checks your mail out of the post office, then prints out a one line header of each message found. The current message is initially the first message (numbered 1) and can be printed using the print command (which can be abbreviated p). You can move among the messages much as you move between lines in ed(1), with the commands + and - moving backwards and forwards, and simple numbers.

After examining a message you can delete (d) the message or reply (r) to it. Deletion causes the mail program to forget about the message. This is not irreversible; the message can be undeleted (u) by giving its number, or the mail session can be aborted by giving the exit (x) command. Deleted messages will, however, usually disappear never to be seen again.

Many commands (e.g., delete, from, and print) accept a list of messages as an argument. Messages may be specified by their message number, by a range of messages, or by a pattern string matching certain fields in the header as described below. These message “specs” may be combined by the usual binary boolean operations ‘&', ‘|', and ‘^', which denote, respectively, a logical “and”, “or”, and “xor”. Logical expressions may be grouped with parentheses ‘(' and ‘)' and negated with ‘!'. If the binary operator is missing between two message specs, it is assumed to be a ‘|'. This is for simplicity, backwards compatibility, and also to to facilitate using the ‘|' symbol to denote a pipe. (See enable-pipes.) Besides the obvious (base10) message numbers, the characters ‘^', ‘-', ‘.', ‘+', and ‘$' denote, respectively, the first message, the message before the “dot” (the current message), the “dot” message, the message following the “dot”, and the last message. A “message range” consists of two message numbers separated by a ‘-'. A ‘*' denotes all messages and is equivalent to ‘^-$'. A pattern is a string (not beginning with any of the above special characters). If it does not begin with a ‘/', it is compared with the senders address. If it begins with a ‘/', and searchheaders is not defined, the remainder of the string is compared with the subject field. (See searchheaders for searching other header fields or the message body.) If regex-search is not defined, then the comparison is a simple case insensitive substring match. (See regex-search for regular expression matches.) A list of messages may be restricted by a “colon-modifier” string, i.e., a ‘:' followed by one or more of the characters: If there are no address specifications other than colon-modifiers, the colon-modifiers apply to all messages. Thus “from netbsd :n” would display the headers of all new messages with ‘netbsd' in the sender's address, while “from :!r” and “from :nu” would both display all new and unread messages. Multiple colon-modifiers may be specified and a single ‘:' with no letters following indicates the colon-modifier from the preceding command. For example: would display the headers from messages 1, 3, 4, 5, and 12. would display all headers that had ‘anon' in the sender's address and either ‘foo' or ‘bar' in the subject line. Generally, commands cannot select messages that are not displayed, such as deleted or hidden messages, the exception being the undelete command.

You can use the reply command to set up a response to a message, sending it back to the person who it was from. Text you then type in, up to an end-of-file, defines the contents of the message. While you are composing a message, mail treats lines beginning with the character ~ specially. For instance, typing ~m (alone on a line) will place a copy of the current message into the response right shifting it by a tab stop (see indentprefix variable, below). Other escapes will set up subject fields, add and delete recipients to the message, and allow you to escape to an editor to revise the message or to a shell to run some commands. (These options are given in the summary below.)

You can end a mail session with the quit (q) command. Messages which have been examined go to your mbox file unless they have been deleted in which case they are discarded. Unexamined messages go back to the post office. (See the -f option above).

It is also possible to create a personal distribution lists so that, for instance, you can send mail to “cohorts” and have it go to a group of people. Such lists can be defined by placing a line like in the file .mailrc in your home directory. The current list of such aliases can be displayed with the alias command in mail. System wide distribution lists can be created by editing /etc/mail/aliases, see aliases(5) and sendmail(1); these are kept in a different syntax. In mail you send, personal aliases will be expanded in mail sent to others so that they will be able to reply to the recipients. System wide aliases are not expanded when the mail is sent, but any reply returned to the machine will have the system wide alias expanded as all mail goes through sendmail(1).

See mailaddr(7) for a description of network addresses. mail has a number of options which can be set in the .mailrc file to alter its behavior; thus “set askcc” enables the askcc feature. (These options are summarized below.)

Here is a summary of the tilde escapes, which are used when composing messages to perform special functions. Tilde escapes are only recognized at the beginning of lines. The name “tilde escape” is somewhat of a misnomer since the actual escape character can be set by the option escape.

Options are controlled via set and unset commands. Options may be either binary, in which case it is only significant to see whether they are set or not; or string, in which case the actual value is of interest. The binary options include the following:

A MIME message is (recursively) divided into a series of MIME parts that can be thought of as sub-messages, each with a header and body. When MIME support is enabled (by setting mime-decode-message), mail splits a message into a series of its smallest MIME parts and processes those parts as if they were messages themselves, passing the header and body through a pipeline of the form: The MIME-decoder decodes ‘base64' or ‘quoted-printable' encoding and is enabled according to the ‘Content-Transfer-Encoding' of the part. The MIME-hook is an external program to further process the part (see below). The pager is the program that pages the message (see PAGER). Any of these intermediate pipe stages may be missing and/or different for the head and body of each MIME part. Certain ‘Content-Types' may disable the entire pipeline (e.g., ‘application/octet'). The MIME-hook stage is not present unless one of the following variables is set: where TYPE and SUBTYPE are the ‘Content-Type' type and subtype (respectively) of the MIME part to which the hook applies. If the “-SUBTYPE” is missing, any subtype is matched. The value of these variables has the format: where the command is expected to read from stdin and write to stdout, and the possible flags are If your command begins with one of these flags, precede it with a space to signal the end of the flags. WARNING: automatically running a program is a potential security risk if that program has bugs, so be careful what you run. Examples: View all ‘Content-Type: image/jpeg' parts with xv(1) (assuming it is installed): Decode all ‘Content-Type: images/*' blocks with uudeview(1) (assuming it is installed), placing the results in /tmp: Read all ‘Content-Type: text/html' parts using lynx(1) (assuming it is installed) and add this support to ‘multipart/alternative' blocks: