3.1 Setting Up Mailagent
Date: 18 Apr 1995 00:03:10 GMT
From: Raphael Manfredi <ram@hptnos02.grenoble.hp.com>
1] First, make sure mailagent is available on your system. The easiest
way to do this is to run:
mailagent -V
which will print the mailagent version if it is available, or the shell
will issue an error message "mailagent: not found" or something like it.
2a] Now you need to understand the MTA (Mail Transport Agent, the program
that delivers the mail; usually, sendmail) will NOT deliver to mailagent
directly, rather to an intermediate (small) filter program. Two versions
are available: shell or C -- refer to the mailagent(1) manpage to choose,
but I recommend you use the C version first, and move to the shell
version if you can't run a binary from your .forward file.
2b] Locate the filter program (it will be filter or filter.sh depending on
whether you choose the C or the shell version respectively) under some
directory like /usr/local/lib/mailagent. From now on, we'll assume we
use the C filter and that it is located under /usr/local/lib/mailagent.
3] Copy the file /usr/local/lib/mailagent/mailagent.cf as ~/.mailagent and
edit it to configure your system correctly. You will see two distinct
sections in that file and you need to set-up properly the first one, the
"Configuration section".
If you have a version of mailagent that is recent enough (at least 3.0 PL32)
then you can create an initial configuration very easily and quickly by
running the following command:
mailagent -I
which will set-up an almost ready-to-use ~/.mailagent file. All you
need to do at this point is go through its configuration section
to make sure mailagent made the right choices...
The minimal set of variables that MUST be correctly set (i.e. for which
you cannot rely on the default set in the file) are (# introduces comment
in a shell-like manner, which run up to the end of the line):
home: # Your HOME directory, as reported by "echo $HOME", usually.
path: # Path to be used to locate mailagent and perl, at least.
p_xxx: # Path to be appended to "path" above when running on machine "xxx".
user: # Your login name.
name: # Your name.
level: # Logging level. I recommend you raise it to 20 for testing.
The mail folder directory is ~/Mail by default, but it can be changed
only from within your rule file by putting
maildir = ~/mail;
at its top, for instance, to make it ~/mail. The rule file is defined as
the "rules" parameter, and is set to ~/.rules by default.
4] Ensure the directories configured in your ~/.mailagent under "logdir" and
"spool" and "queue" do exist. If you use the standard setting, this
requires the following commands:
cd
mkdir var
cd var
mkdir log mailagent
cd mailagent
mkdir queue
5a] Create a rule file (named ~/.rules by default) for testing:
cd
vi .rules
5b] Enter the following in ~/.rules
Subject: /test/ { SAVE testing };
The meaning of that rule should be pretty obvious: If we receive a mail
whose subject line contains the word "test", then we save that mail in
a folder named "testing", under the default folder directory (~/Mail).
5c] Create a ~/.forward file as follows:
"|exec /usr/local/lib/mailagent/filter >> /export/home/ram/.bak 2>&1"
The meaning of that line is the following: every mail should be piped
(hence the leading "|" character) onto the filter program, and any
output from that program (i.e. errors) are appended to some file in
your home directory, with stderr following stdout (2>&1) in traditional
sh syntax.
== IMPORTANT NOTES ==
* Your .forward is always processed by sh, regardless of your login shell.
* Replace /export/home/ram with your proper login directory full path.
That's a part that makes your .forward unique (for zealous optimizing
sendmail that are dead wrong about optimizing!) and that can save
you a lot of trouble if anything goes wrong! Just look at your ~/.bak!
* Replace /usr/local/lib/mailagent/filter with the proper filter path
on your machine.
5d] Note that on many systems, you need to ensure your .forward will be
readable by everybody, and that your home directory has the "x" bit set
for everybody (i.e. can be part of a lookup path) so that sendmail can see
and parse your .forward file. To ensure this, type:
cd
chmod a+r .forward
chmod a+x .
6] Send yourself two test messages: one with "test" in the subject, and one
without "test" in the subject.
7a] Checkout your ~/.bak file: it should be empty!
7b] Checkout your ~/var/log/agentlog file to see what really happened to your
messages. Watch out for any ERROR or WARNING logs. This assumes you have
left at least the default logging level (9) in the ~/.mailagent file (the
"level" variable). But for testing, that level should be raised to 20 to
help you diagnose what's going on.
7c] Look out in ~/Mail/testing. You should find there the message whose Subject
line contained the word "test". Then make sure the other message has been
delivered to your regular mailbox. (Since no match occurred in your rule
file, the mail is left in your mailbox by default).
7d] TROUBLESHOOTING
* If your mail was not properly delivered, please make sure your rule file
and configuration file are correct by issuing the following command:
mailagent -d
* If the previous command does not output the single rule you should have
put in ~/.rules, then please make sure step 3 and 4 above were correctly
performed (those being the crucial steps ensuring a proper configuration).
* Check the ~/.bak file for error messages.
* Check for typos in any of these files:
~/.forward
~/.mailagent
~/.rules
* Check the file and directory permissions of your .forward (set in step
5c above).
Type... In order to...
------- --------------
cd Go to your home directory.
ls -l .forward Check the permission: it should say -rw-r--r--
ls -ld . Check permission of home dir: it should say drwx?-x?-x
The ?'s may be r's or hyphens or one of each (i.e.,
drwx--x--x, drwxr-xr-x, drwxr-x--x, drwx--xr-x).
* If none of the previous hints helped you identify problem, and you can't
figure it out from the output in ~/.bak or in the ~/var/log/agentlog file
(or whatever file you have configured for logging within your ~/.mailagent
file, variables "log" and "logdir"), then make sure your mail is not
waiting in the MTA's queue: this might be the case if the agentlog file
is empty. If you are using sendmail as MTA, you can run:
/usr/lib/sendmail -bp
to print out the queue.
* As a last resort, please look at the mailagent(1) manual page under
the section "Testing Your Installation" for more tips and things
to look at.
8] Once you have successfully tested mailagent in steps 6 and 7 above,
you're on! Mailagent is ready to process your mail. All you have to
do is extend the ~/.rules file to add more rules.
For instance:
To Cc: www-talk { SAVE www-talk };
To Cc: agent-users { SAVE agent-users };
Those two rules filter the two mailing lists www-talk and agent-users into
their respective folders, whether the mailing list address appear in the To
OR Cc header.
Since rules are not qualified as a pattern match (contrary to our test
above), they match on logins in the address, i.e. they will match things
like www-talk@chip.com or chip!www-talk, or a plain simple www-talk if
this is a local alias. (This implicit matching on logins works only for
some selectors like To, Cc or From which are know to contain addresses).
If you wish to sort on patterns appearing in the Subject of messages for
instance, then you must use a pattern matching syntax, as in:
Subject: /star trek/ { SAVE star-trek; };
to save in a folder "star-trek" all the messages whose subject contains
the words "star trek". Case matters, but keep on reading...
9] As an advanced topic, since mailagent is written in Perl, you have all the
power of Perl's regular expressions at your disposal. Which means you can
write things like this:
To Cc: agent-users { REJECT AGENT };
<AGENT> Subject: /^\w*subscribe/i { DELETE };
<AGENT> * { UNIQUE -a; SAVE agent };
The second lines makes use of that perl extended regular expression syntax:
\w matches an alphanumeric character plus "_", while the trailing "i" option
requests a case-insensitive match.
You also note you have a real automaton at your disposal. You can enter
a special state (AGENT in our example) and continue parsing by only
scanning for rules tagged by this mode. The first match stops the automaton,
unless you REJECT to continue processing. When not restricted by a mode list,
a rule is always taken into account. For example, assuming the automaton
is in the state "NEWS", it will not consider rules tagged <AGENT>, as in
the above example. The automaton begins in mode "INITIAL".
The "UNIQUE -a" action followed by a SAVE ensures only one copy per
Message-ID will ever end-up in your agent folder -- no duplicates!
Also note you can have more than one action per rule, and that the last
one uses '*' to match anything, i.e. its action part between {} will
always be executed in AGENT mode, when reached by the automaton.
Also, since in Perl regular expression syntax, \b matches a word-boundary
and \s any space or tab character, we can write our Star Trek message
sorting into a much more robust form:
Subject: /\bstar\s+trek\b/i { SAVE star-trek; };
(\s+ matches one or more white spaces, while \s* would match zero or more,
see the Perl manual page for a complete description of regular expressions.)
which will match on various subject strings like "Last Star Trek season"
or "I am addicted to Star trek", but not on "Tristar treks" -- whatever
that means :-)
All in all, the filtering automaton syntax is pretty much intuitive and
easy to read. You have to learn which actions are possible and what they
mean, naturally.
Parent document is top of "Filtering Mail FAQ"
Previous document is "3.0 Mailagent"
Next document is " 3.2 Tracking Your Incoming Mail"