The User Database

The User Database is a special database file that you create for use by sendmail. It causes sender and recipient addresses to be rewritten under control of an external database file. Ordinarily, any local address is first looked up in the aliases database. If it is not found there, that user's ~/.forward is next examined. If the User Database is enabled, the address is looked up in that database after aliasing and before forwarding.

Lookup can be local via a database file, remote via a User Database server, or via a Hesiod network service. Here, we describe the database file form. The others are described in UserDatabaseSpec (U).

Enable the User Database

The User Database is automatically enabled when you compile sendmail if you include support for NEWDB or HESIOD (see USERDB). To see whether a precompiled version of sendmail includes User Database support, run it with the -d0.1 switch:

% /usr/lib/sendmail -d0.1 -bt < /dev/null Version 8.8.4 Compiled with: LOG MIME8TO7 NETINET NETUNIX NEWDB SCANF USERDB XDEBUG -^ note

If USERDB is listed, User Database support is included.

Next you must declare the location of its database file with the UserDatabaseSpec (U) option (see ):

OU/etc/userdb.db <-in your cf file (V8) O UserDatabaseSpec=/etc/userdb.db <-in your cf file (V8.7) define(`confUSERDB_SPEC',/etc/userdb.db) <-in your m4 file

Here, the location of the database file is set to be /etc/userdb.db. You can also enable a default location for the database file that will take effect should the UserDatabaseSpec (U) option be missing by defining that location with UDB_DEFAULT_SPEC in compiling (see UDB-DEFAULT-SPEC).

Create the User Database

The User Database is a btree class (see ) database file created from a source text file using the makemap program:

% makemap btree /etc/userdb.db < /etc/userdb

Here, /etc/userdb is the source-text file that is input, and /etc/userdb.db is the database we are creating (the one defined by the U option in the previous section). [5]

[5] The db is added automatically if it is missing. We include it here for clarity.

The source-text file is composed of key and value pairs, one pair per line:

key value -^ whitespace

The key is a user's login name, a colon, and one of two possible keywords: maildrop or mailname. The keyword that is chosen determines the nature of the value. maildrop The value is the official delivery address for this user. If there are multiple official addresses, they should either be listed as a single compound value, with separating commas, as, for example:

root:maildrop sysadmin@here.us.edu,bill@there.us.edu

or be listed on individual lines:

root:maildrop sysadmin@here.us.edu root:maildrop bill@there.us.edu

This latter form requires you to use the -d command-line switch with the makemap(1) program (see ) when creating the database but has the advantage of being a simpler source file to manage. mailname The mailname keyword causes a "reverse alias" transformation, wherein the login name in the key is changed into the address in the value for outgoing mail. For example:

bob:mailname Bob.Roberts@Here.US.EDU

This causes mail sent by bob to go out addressed as though it is from Bob.Roberts@Here.US.EDU. [6] This transformation occurs in the header and envelope. But note that the sender-envelope is not rewritten by UDB unless the F=i flag (see F=i) is present in the delivery agent that is selected for the sender. Also note that the recipient headers are not rewritten by UDB unless the F=k flag (see F=j) delivery agent is selected for the recipient.

[6] Using full names in outgoing mail is probably not a good idea. Unlike login names, full names are not guaranteed to be unique. If current users expect to be able to receive mail under full names, future users with the same full name may be out of luck. Always weigh convenience against maintainable uniqueness when designing your mail setup.

Naturally, the maildrop and mailname keywords should occur in pairs. Each outgoing address that is created with mailname should have a corresponding maildrop entry so that return mail can be delivered. In the above example a reasonable pair might look like this:

bob:mailname Bob.Roberts@Here.US.EDU Bob.Roberts:maildrop bob

Here, outgoing mail from the user named bob will be addressed as though it is from Bob.Roberts@Here.US.EDU. Incoming mail (whether it is original or in reply to the outgoing) will be addressed as though it is to the name Bob.Roberts, which will be transformed into and delivered to the local user bob.

:default Outgoing Hostname

The mailname keyword allows the host part of outgoing addresses to mask the real hostname of the originating machine. This property can, for example, be used to convert the hostname into a firewall name:

bob:mailname bob@Firewall.US.EDU

Here, the canonical name of bob's machine is Here.US.EDU. The mailname keyword causes outgoing mail from bob to appear as though it is from the firewall machine (Firewall.US.EDU) instead.

Ordinarily, this transformation is not automatic. Each username that is to appear to be from the firewall machine will need an entry like that above in the User Database. To automate this process, you can use the special username :default in a mailname declaration:

:default:mailname Firewall.US.EDU

If a maildrop entry is found for a particular name, but no corresponding mailname record is found, the outgoing address is ordinarily unchanged. If, however, a default hostname has been defined with :default, that hostname replaces the local hostname for all addresses that lack their own mailname entry:

:default:mailname Firewall.US.EDU bob:maildrop bob@here.us.edu

In this example the user bob has a maildrop entry but lacks a mailname entry. Outgoing mail from this user will have the :default hostname used instead of the local hostname. The user sally, on the other hand, has neither a maildrop entry nor a mailname entry and so will not have her outgoing address rewritten.