SmartList mailinglist management -------------------------------- Copyright (c) 1993-2000, Stephen R. van den Berg, The Netherlands. <srb@cuci.nl> NOTE: Run "flist -v" to find out more about your installed version of SmartList. Contents: 1. Creating and removing mailinglists or archive servers 2. Remote maintenance of mailinglists 3. Customisation 3a.Digest processing 3b.Restricting subscriptions 3c.Restricting submissions 3d.Auto subscription on first submission 3e.Autosending files to new subscribers 3f.Moderated lists 3g.List of lists 3h.Positively discriminate certain daemons 3i.Default help text replies 3j.Unsubscribe assistance 3k.Exploding other (non-SmartList) lists 3l.Making the subscriber list public 3m.Schematic overview of what goes on behind the scenes 4. The archive server 4a.Sending files to the archive server 4b.Restricting access to the archive server 5. The format of the dist file 6. Multigram and the thresholds in rc.init/rc.custom 7. Choplist & sendmail 8. FTP addresses & the SmartList mailinglist 9. The FAQ $Id: Manual.html,v 1.3 2006/12/06 22:38:48 hu Exp $ 1. Creating and removing mailinglists or archive servers ----------------------------------------------------- Make sure that the .bin directory is in your PATH. Now you can issue commands like: createlist testing createlist testing joe@somewhere.edu createlist -a testing joe@somewhere.edu removelist testing The first command creates a mailinglist with two useful addresses: testing testing-request The second command does the same, but it also specifies joe@somewhere.edu to be the responsible contact person for this list. The third command does the same, except instead of creating a mailinglist, it creates an archive server. The fourth command removes all traces of the "testing" mailinglist again. There are four other convenience-utilities that can be used: led A wrapper around your editor, should be used instead of calling up your editor directly whenever editing a SmartList managed file. It automatically takes care of proper locking and performs attribute checks. delink It will unlink a file from its hardlinked counterpart(s) (without deleting it). showlink It will display what groups of files are linked together. donatelist It will put a list under complete and exclusive control of the maintainer (a local user). See below. If you are running several lists maintained by separate maintainers, you can give a maintainer complete and sole control over his or her own list without the need for them to have user list or group list rights. For this to work, you simply have to "donatelist the_maintainer his_list" the whole tree that contains his list to him (her). Make sure that the group id of all necessary files in the tree are still group-writable by "list", because that's the access privilege the mailinglist will be running under. The maintainer has to be careful to use an umask of 007 while editing in his mailinglist directory. This allows the mailinglist-programs to function while still limiting access to all mailinglist files to *one* person only (the maintainer). N.B. The person a list is being donated to is able to gain access to the SmartList account, if he tries hard enough. To get secured multiple lists, install SmartList several times. 2. Remote maintenance of mailinglists ---------------------------------- To facilitate remote maintenance of some mailinglists by their maintainers I have created the .bin/x_command script. It parses mails sent to the -request address and can execute some administrative commands. The mail should be sent to the -request address of a mailinglist and should contain a field in the header looking like this: X-Command: joe@somewhere.edu password command "command" can be anything of the following: subscribe mailaddress unsubscribe mailaddress checkdist mailaddress To multigram-match mailaddress to the list (showing the eight best matches) showdist To list the distfile showlog To list the log wipelog To clear the log help To show this command summary info Ditto The exact fieldname defaults to "X-Command", but can be customised to whatever you want. The password defaults to "password", but can/should be changed. The "joe@somewhere.edu" is always the mail address of the maintainer. Note that this has to match what was specified on the command line of "createlist" when the list was created. Note that the X-Command: field has to be part of the header, when it's in the body of the mail, it has no effect. Anytime an X-Command: mail has been processed, the results will be mailed back to the maintainer of the list, and the X-Command: field will have been renamed to X-Processed:. Although this remote-facility is convenient, some might argue that it presents a security hole. Well, in order to make this hole as small as possible, you can keep the password secret. Also, the exact mailaddress of the maintainer might not be publicly known. You can simply change the X-Command field into something else like X-MyCommand. Above all, since faking mail is a well known possibility it would be ridiculous to take more precautions than these. Besides, if someone indeed manages to sneak in a bogus X-Command:, it will never go unnoticed since the mailing list maintainer (and only the maintainer) will always receive the X-Processed: mail. For your convenience, a sample script "doxcommand" is present in the SmartList/examples directory. It can be used to easily generate these X-Command mails. Do remember to read-protect this script once the password has been changed. 3. Customisation ------------- The mailinglists can be customised in several ways: - For all the lists: - Since all the lists initially share the same help.txt, subscribe.txt, unsubscribe.txt, rc.init, rc.submit and rc.request files (hardlinked), any change to them will affect all lists. - Since all the lists have the .bin directory in their PATH, any change to one of the Bourne shell scripts in there will affect them all. - Per list: - Every list directory contains an "rc.custom" rcfile which can be edited to your hearts content to customise certain parameters for this list only. - Small local customisations can be realised by uncommenting one or more of the RC_LOCAL_* assignments in rc.custom. You then have to create the appropriate rc.local* file in which you can put any commands you'd like (e.g. adding a general signature or disclaimer to every outgoing submission). - For graver customisation you can remove the hardlink (using .bin/delink for example) to any of the files in a list directory and provide that list with its own copy in order to edit that to taste. - Since the current directory is in the PATH before the .bin directory you can create per-list copies of any of the Bourne shell scripts in .bin which can then be changed without affecting the other lists. - Per group of lists: - The same applies as when customising per list, but you should then hardlink the appropriate files among the group of list directories. By default the scripts create and use hardlinks in various places. You are completely free to change some or all into symbolic links instead (or substitute "ln" with "ln -s" in some scripts). Some editors have a habit of moving the file you were editing to a backup name and writing out a new copy in its place. This can cause problems if the editor is unaware of the symbolic or hard links in place. You should make sure that the editor is aware of the link and preserves it, even after the file has been edited (for people using emacs: try setting backup-by-copying-when-linked to true). By using the "led" script instead of calling your editor directly you will be timely warned of anything your editor broke. If you are not using the remote-maintenance facility and you start editing or customising scripts/files by hand, then you should make sure that there doesn't arrive any mail to those lists that are affected by your changes. The best way to do this is by using the command "led" whenever you want to edit a SmartList governed file. Led will take care of all the necessary locking for any file, led can also be used to edit non-SmartList files. If you don't use "led" but still would like to put incoming mails on hold temporarily, then you can do this: - for all the lists by creating the file: .etc/rc.lock - only for one list by creating the file: rc.lock in the list directory of that list. The .bin/flist command checks to see if these rc.lock files exist AND are not older than 17 minutes before delivering the mail. So, if you create an rc.lock file, mails to that (or all) lists will stall for the next 17 minutes. If you need more time, touch the file every so often. You should remove the rc.lock files again after finishing your editing. If you would like to change the -dist alias (used to distribute the mail to the subscribers) into something less well known, go right ahead, but remember to change the corresponding assignment to listdist. For completeness sake one should also correct the createlist and removelist scripts in that case. If you are using choplist to expand the dist-file, then you don't even need a -dist alias at all. 3a.Digest processing ----------------- You can configure a list to send out digests of accumulated submissions. In order to do so, simply uncomment the appropriate assignment to digest_flag in rc.init (if you want all lists to be digested) or rc.custom (if you only want this list to be digested). Digests are then sent out every so often depending on size and age of the accumulated messages. The conditions for sending out a digest are checked during the arrival of every submission. If, however, traffic on the list sometimes is very low (i.e. less often than the maximum age of a digest) a digest could be laying around for longer than the specified maximum period (3 days by default). In order to make sure that the digest gets sent out anyway, you should be running the .bin/cronlist program every so often. The recommended procedure is to create a cron job (under the list account) that contains something like the following entry: 0 6 * * * /home/slist/.bin/cronlist The cronlist script can be customised to taste (maybe you'll need to adjust the setting of the PATH variable). Beware: call cronlist with an absolute or relative path, do not rely on PATH to find it for you (cronlist uses $0 to find the location of the lists it is responsible for). By default, cronlist will run the flush_digests program. By using the above listed crontab entry, you will ensure that at six o'clock in the morning all the overdue digests will be sent out. Flush_digests normally checks the age of the digest and does not send it out until it is overdue. If you want to force flush_digests to send out a digest of a particular list, you can create the file ".digest.force" in that list's directory. During the next run of flush_digests, it will remove .digest.force and push out the digest (if any) regardless of its age. If you create a file named digest.admin in either the main directory of the digested list or in the archive/latest directory belonging to it, it will be picked up by the next flush_digests and included up front to the actual digest under the heading "Administrivia". The archive/latest/digest.admin file digested list will be automatically removed after the digest has been pushed out. The digest.admin file in the main directory of the digested list will not be removed and is included in every digest. If you want to give your subscribers the choice of receiving digests or not. This is what you can do: Create two lists. E.g. if your list would be called "thelist", then you have the `real' list called "thelist" (created and used like a regular list) and the `digested' list called "thelist-d". In the distfile of thelist you should include thelist-d as one of the subscribers. In the rc.custom file of thelist-d you should edit the assignment to undigested_list to read "undigested_list = thelist@$domain". After you've done this, you're all set. People that want digests simply subscribe to thelist-d and people that don't, subscribe to thelist. 3b.Restricting subscriptions ------------------------- There are three ways in which you can restrict who can subscribe to a list: - You can put the addresses of unwanted subscribers on the so-called reject- list (the `reject'-file). - You can create a program (e.g. a shell script) called "subscreen". It must be executable and will receive the mail address of the prospective subscriber as the first argument. If subscription for that address is allowed, the program must return with exitcode zero. If subscription is disallowed, simply return with exitcode one. A sample program is provided in the examples directory. - You can completely disable automatic subscription by uncommenting the appropriate "auto_subscribe" line in rc.custom. - You can completely disable automatic unsubscription by uncommenting the appropriate "auto_unsubscribe" line in rc.custom. 3c.Restricting submissions ----------------------- You can restrict submissions to people on the accept-list (the `accept'-file). Mail from anyone else will be passed on to the maintainer instead of being submitted. To enable this you have to uncomment the appropriate "foreign_submit" line in rc.custom. By default the accept file is hardlinked to the dist file (i.e. if submissions are restricted, only subscribers can do so). If you want to allow only an even more select group, delink the accept file and edit it to taste. If you'd like to have both the dynamic accept file and a static one, create a new file "accept2", it will be searched in addition to the regular accept file. Beware that using an accept file is incompatible with having an owner- alias for this list (procmail administered lists do not need the owner- alias anyway, so if you've never heard of such a thing, ignore this warning). If, in addition to notifying the maintainer you want an automated reply to be generated to the submitter which was not in the accept file, then you can accomplish this by simply creating an accept.txt file. Its contents will (like the contents of the help.txt file) be returned to the submitter. This feature is not the same as a moderated list, the two features can be used accumulatively. On a related note, SmartList, by default, tries to identify administrative requests that got mailed to the submission address instead of the -request address and diverts them to the -request address. If for some reason this is undesirable, you can uncomment the appropriate "divertcheck" variable assignment in rc.custom to disable this feature. 3d.Auto subscription on first submission ------------------------------------- Instead of rejecting submissions by people not on the accept (dist) list, you can enable "force_subscribe". This will cause people submitting mails to the list to be autosubscribed to the list if they were not in the dist file. 3e.Autosending files to new subscribers ------------------------------------ You can create a file named "subscribe.files". It can contain any number of archive-server commands. The results (i.e. the files requested) will be sent to the new subscriber. 3f.Moderated lists --------------- First create a file named "moderators", it should contain the fully qualified mail addresses of all the moderators for this list (i.e. just local usernames are not sufficient, at least include an @host or host! ). Then uncomment the appropriate "moderated_flag" line in rc.custom. From then on all mail that does not contain an "Approved: the_address_of_one_of_the_moderators" field is forwarded to all the moderators. One of the moderators should then resend the mail to the list after adding an "Approved: his_own_address" field to the header (and possible editing the contents of the mail). It will be no problem if several moderators resubmit the same submission concurrently, since the mailinglist will filter out duplicates anyway (i.e. only the first one will go out and be archived). 3g.List of lists ------------- If you want people to be able to get an overview of which lists are publicly available at your site, you can have your listmaintainers create a file called "info.txt" in their respective list directory. This info.txt file should contain a short description of the purpose and main topic of this particular list. You can then setup a command like examples/gatherinfo in crontab to collect all these various info.txt files once a day. The thus gathered info.txt files can be placed in a directory which can be accessed by a gopher or ftp server, or, you can put them into the archive directory of a procmail-managed mail-archive server (e.g. under the mail-alias "metalist"). 3h.Positively discriminate certain daemons --------------------------------------- SmartList usually does not accept submissions or subscriptions from daemons. If you'd like to make an exception for some, you can do this by tuning the daemon_bias variable. A sample template can be found in the rc.init file. This variable can of course be set in the rc.init, rc.custom or rc.local files. Instead of directly specifying a weight and a regexp, you can just specify a weight. You'll then have to make sure that the variable is set only when mail from your special daemons arrives. Beware that if you change this for a list that has a digested shadowlist, you change it for the digested version as well. 3i.Default help text replies ------------------------- By uncommenting the appropriate "auto_help" line in the rc.custom file the list will respond to every undecipherable request message as if it requested help. Messages that will still get through to the maintainer are those: - that seem to come from a daemon. - which look like a reply. Depending on the typical audience you have on the list, enabling this might not be a good idea. 3j.Unsubscribe assistance ---------------------- By uncommenting the appropriate "unsub_assist" line in the rc.custom file you can turn on the unsubscribe assistance. If the someone is trying to unsubscribe from the list, but his mailaddress could not be found, he will receive back a number of multigram matches (determined by the value of unsub_assist) between his unsubscribe message and the dist file. If you'd like to enable this feature, please keep the following points in mind: - People can get excerpts of the dist file this way. - Malevolent individuals might become encouraged to unsubscribe lots of people from your list. This will not go by unnoticed, of course. It will be logged and the innocent subscribers will receive a copy of the unsubscribe request they didn't send. Nevertheless, it can cause considerable inconvenience. 3k.Exploding other (non-SmartList) lists ------------------------------------- A SmartList list can easily be used to function as a local exploder for a larger mailinglist. The advantages over using a regular alias are threefold: - (Un)subscription is handled automatically. - Bounce messages go to the local exploder list (instead of the larger mailinglist which really is not interested in your mail problems with some local aliases). - Misdirected administrative requests are filtered out of the regular submission channel. If the larger mailinglist you are exploding locally is a SmartList list as well, then there are no special precautions to take at all. If the larger mailinglist is not managed by SmartList, misdirected administrative requests will be caught *and* handled by the local list; if this "handling" turns out to be a problem you can turn it off by uncommenting the appropriate "pass_diverts" variable in rc.custom (this will cause misdirected administrative requests to be caught but then passed on to the maintainer verbatim). 3l.Making the subscriber list public --------------------------------- If you want anyone to be able to retrieve the list of current subscribers to a list, simply make a link from the dist file to archive/subscribers. This enables anyone to send in a request to the archive server retrieving the file "subscribers". 3m.Schematic overview of what goes on behind the scenes ---------------------------------------------------- Suppose you have two entries in the aliases file, one for thelist@domain and one for thelist-request@domain. Whenever mail arrives for either address, the following happens: - flist is started suid root with thelist as its argument - changes its current directory to that of the (internally hardcoded) main list directory - changes its uid and gid to that of the (internally hardcoded) list account - changes its current directory to that of thelist - waits until both ../.etc/rc.lock and rc.lock are gone or are old enough (17 minutes) Then, if it was a regular submission to thelist@domain: - flist execs procmail with rcfile rc.submit - pulls in rc.init that sets up the defaults - pulls in rc.custom that overrides some defaults (if any) - checks the submission format - if needed it now checks: - the accept list - the moderators list (Approved:) - checks the msgid cache for duplicate submissions - if needed does digest processing (and stop) - archives the mail - munges the header of the mail to taste - fires off sendmail to send the mail to thelist-dist@domain - If the mail was an administrative request, it does not get passed on to the list, instead, procmail pulls in rc.request But, if it was an administrative mail for thelist-request@domain: - flist execs procmail with rcfile rc.request - pulls in rc.init that sets up the defaults - pulls in rc.custom that overrides some defaults (if any) - performs the necessary actions, depending on the content - if the content was undecipherable, it gets passed on to the maintainer of thelist If there are grave system failures during all this, the catch-all script rc.post will kick in and make sure that the mail is stashed away somewhere or forwarded to the maintainer, whatever works. This to ensure that no mail gets lost. 4. The archive server ------------------ All mail (except mail being forwarded from another mailinglist) sent to any of the lists is archived. The archiving is fairly straightforward. E.g. if you have a list called "scuba", then all submissions are archived in scuba/archive/latest/. The mails will be stored one-mail-per-file each. The files will be numbered. Now, normally, only the last two mails will be kept around, the others are periodically removed. This in order to keep down the archiving costs for people with limited diskspace. To change the size of the archive-history edit the rc.custom file. To get more sophisticated archiving, like grouping submissions monthly, you should either create a cron job or edit the .bin/arch_trunc file. The archive server can be accessed per mailinglist by sending mail to the -request address with the following Subject: Subject: archive The body of the mail or the rest of the subject line can then be filed with requests to the archive server. It basically understands five commands: get file ... ls directory ... egrep case_insensitive_regular_expression file ... maxfiles nnn help The archive server does a thorough check on the commands and the files that are requested. This to ensure that it does not access any files outside the "scuba/archive" directory. Any text-file that you put in the "scuba/archive" directory (or any of its subdirectories) can now be retrieved by the archive commands. If a file requested via the archive server starts with a header that begins with `Content-Type:', then the file is sent out as is, without encapsulation. This allows you to prepare the files in special formats that are directly supported by the recipient's mailuser agent. The leading Content-Type: and any immediately following fields will become part of the header. All other files are MIME-encapsulated before transmission. You should take a look at /home/slist/.bin/mimencap.local if you want to extend or customise the recognised file types. The MIME-encapsulation is automatic and depends on the availability of the metamail package in the PATH defined in rc.init. The programs from this package which need to be available are: mimencode and splitmail. If mimencode is not found on the PATH, the files will be sent out with a standard wrapper around them. You can switch from MIME-encoding to uuencoding by applying the appropriate patch from .examples/uuencode.dif. If you put links in the "scuba/archive" tree, you can allow the archive server to retrieve files from other parts of the filesystem. The whole archive server can be found in the .bin/arch_retrieve script. The archive server can be extended with arbitrary commands via the examples/retrieve.local script that comes with the distribution. 4a.Sending files to the archive server ----------------------------------- The archive server as installed with SmartList does not directly support the receipt and storage of files. What you can do is look in the at the rcfile script .examples/putfile. It provides you with all you'd need to receive and store files. 4b.Restricting access to the archive server ---------------------------------------- You can restrict archive access to people on the accept-lists (the `accept' and `accept2'-file). Mail from anyone else will be passed on to the maintainer instead of being passed to the archive server. To enable this you have to uncomment the appropriate "restrict_archive" line in rc.custom. 5. The format of the dist file --------------------------- You do not need to know this, unless you edit the dist file by hand or want to incorporate an existing list of addresses. In order to distribute incoming submissions the dist file is fed to sendmail with the regular :include: alias. So the format of this file must be in accordance with what sendmail would expect. In addition to that this file is searched and edited by multigram in order to find particular subscribers. The format which multigram expects is a bit more rigid than what sendmail allows. The following conditions apply: - One subscriber per line. - Empty lines are allowed. - The mail address of the subscriber must be the first word on the line. - Comments may follow the address (but separated from the address by at least one whitespace character). - Everything preceding the line containing: (Only addresses below this line can be automatically removed) is write protected from changes by multigram (i.e. these addresses can never be automatically/accidentally unsubscribed). - If the line: (Only addresses below this line can be automatically removed) is not present at all, automatic unsubscriptions to this list are impossible. - Whenever multigram automatically removes an address from the list, it rewrites the dist file `in situ'. This means that the dist file will be contracted at that point. I choose to write in situ in order to avoid copying the dist file every time it changes (a real life saver if the list grows too big). - Multigram always adds new subscribers on the line immediately following the last filled entry in the dist file. Some sample entries (the preferred format): joe@some.where joe@some.where (some comment) joe@some.where (some comment) (some more comments) Deprecated, but allowed: <joe@some.where> <joe@some.where> some comment <joe@some.where> (some comment) Not allowed by multigram (although sendmail doesn't mind): (some comment) joe@some.where some comment <joe@some.where> 6. Multigram and the thresholds in rc.init/rc.custom ------------------------------------------------- The rc.init and rc.custom scripts define some threshold values: match_threshold, off_threshold, reject_threshold, submit_threshold. These values are fed to multigram as a cut-off value with which to decide if a certain mail address is on a list. The higher the threshold, the better the match must be. The thresholds have a scale from -16383 to 32766. This means that, for instance a threshold of 30730 can be used to find only mailaddresses that are almost verbatim on the list. A value of 24476 on the other hand allows for some error (like mailaddresses munged by gateways etc.) in finding matches to the list. The values 30730 and 24476 are somewhat arbitrary values which seem to work well for the particular problems at hand. To get a feeling for the values computed by multigram you can do the following test: Create a file with the same format as the distfile, fill it with any number of addresses you like (e.g. you could take an existing distfile). Now make a copy of this `distfile' and alter some of the addresses a bit (like omit one character, or add some gateway information, switch two words, change it into an uucp address, etc.). Next you should call up multigram with the following command line: multigram -l-16000 -b300 pseudo_distfile <altered_distfile Multigram will display up the 300 best matches it found after crossreferencing altered_distfile and pseudo_distfile. The output produced by multigram can be disected as follows: lineno. name1 goodness name2 Lineno. and name1 refer to the line number in pseudo_distfile which contains the mailaddress name1. Goodness is the metric that corresponds to the aforementioned threshold values, and name2 is the matching mailaddress from altered_distfile (which is usually the incoming mail). Once you get the hang of it you can play around a bit with the entries in altered_distfile by mutilating them more and more in order to see what multigram makes of it (try inserting some non- existing addresses as well). 7. Choplist & sendmail ------------------- There are two ways of distributing the message to the addresses listed in the dist file: either directly through sendmail and the :include: alias expansion, or indirectly, through choplist which does the expansion itself and calls up sendmail on suitable chunks. By default, choplist takes care of the expansion, uncomment the appropriate alt_sendmail environment variable in rc.init to revert to standard sendmail expansion. Choplist tries to make sure that even for big lists, the alias expansion goes as swiftly as possible and several sendmails will be delivering the mail to the subscribers concurrently. The various limits that can be imposed on this process can be tuned in the rc.init file. Most sendmails will do a worse job without choplist as a preprocessor. A side effect of using choplist as a preprocessor is that there is no need for a -dist alias anymore. 8. FTP addresses & the SmartList mailinglist ----------------------------------------- The latest version can be obtained directly from: http://www.procmail.org/ or ftp://ftp.procmail.org/pub/procmail/ There exists a dedicated mailinglist for SmartList users, send your subscription requests to: SmartList-users-request@procmail.org For procmail related questions not specific to the SmartList package you can also write to the procmail mailinglist, send your subscription requests to: procmail-users-request@procmail.org There is a readonly announcement list to stay informed about new versions and official patches of both procmail & SmartList, send your subscription requests to: procmail-announce-request@procmail.org SmartList-announce-request@procmail.org 9. The FAQ ------- If you have questions not answered by the above you should check the FAQ at: http://www.animalgenome.org/SmartList/ This FAQ was created and is maintained by various members of the SmartList community, most particularly jason@mindwell.com Werner Reisberger <werner@pure.ch> Peter Hartzler <ph@eainet.com> Zhiliang Hu <zhilianghu@gmail.com> Suggestions for the FAQ should be sent to <smartlist@lists.RWTH-Aachen.DE>.