| < generate-files | Russ Allbery > Software > control-archive | update-control > |
(Process Usenet control messages to maintain a database)
process-control
process-control processes Usenet control messages, applying the rules from a control.ctl file in the same format used by INN, and maintains a database of active groups with descriptions. It also archives control messages in gzip format in a central archive and logs all of its actions to a log file.
It expects to receive on standard input file names and message IDs, one pair of file name and message ID per line. The message ID is optional, but is used for logging until the message has been parsed and therefore omitting it may result in log messages that have no associated ID. The file is deleted when it has been successfully processed.
Messages are considered invalid and are discarded (with a log message) if:
The message is larger than 256KB, or the message is not a checkgroups control message and is larger than 64KB.
The message is syntactically invalid. This means that it's empty or it contains a line in the headers that doesn't fit the syntax of the headers (a sequence of printable, non-space ASCII characters followed by a colon and at least one space, or a line beginning with whitespace indicating a continuation of the previous header).
Any of the headers Approved, Control, From, Message-ID, or Subject are missing, or any of the headers Approved, Control, Date, From, Message-ID, Subject, Sender, or X-PGP-Sig are duplicated.
The message is not a newgroup, rmgroup, or checkgroups control message, or the Control header doesn't follow the correct syntax for that type of control message. An rmgroup control header must have exactly one argument, the name of the newsgroup to remove. A newgroup control message may have two or three; the second must be the name of the newsgroup to create and the third, if present, must indicate the group mode.
The message is a newgroup control message and the newsgroup mode does not
begin with either y (unmoderated) or m (moderated).
The message is a newgroup control message and the group name to be created
is longer than 80 octets, contains a character other than those from the
set [a-z0-9+_.-], begins or ends with a period, does not contain a period,
contains two consecutive periods, begins with a character other than a
lowercase letter, starts with control., example., or to.,
contains a name component (delimited by periods) that is entirely numeric,
contains a name component (delimited by periods) equal to all or ctl
(these have special meaning for older servers), or contains a name
component that begins with a character other than a digit or letter.
All other messages will be archived. newgroup and rmgroup control messages will be archived in a file named hierarchy/group.gz, where group is the affected group and hierarchy is its first component (the part up to the first period in the name, if any). checkgroups control messages will be archived in other.ctl/checkgroups.year where year is the current year.
Messages will also be checked against the rules from a control.ctl file and messages that, according to the rules in that file, should be acted on will result in updates to a database of active groups. That database is a Berkeley DB hash file with keys equal to the group names and values of the group mode, a single space, and then the group description. It is locked against concurrent updates.
Messages that control.ctl indicates need to be checked using PGP will be passed through GnuPG using the same algorithm as pgpverify and will only be processed if the signer matches the required signer in control.ctl.
newgroup messages must have a For your newsgroups file: line in the
body, followed by a line suitable for a checkgroups message (the group
name, some whitespace, and the description, containing no control
characters). (Moderated) will be added to the description if not
present for a moderated group, and will be removed if present for an
unmoderated group. For a newgroup message to be acted on, it must also
not contain any component longer than 30 characters.
The same constraints on group names mentioned above are also applied to
the group names in the body of a checkgroups message, and any groups not
fitting those requirements will be ignored, as will any group that does
not have a description. In a checkgroups message, the moderation status
of the group will be determined by the presence or absence of
(Moderated) at the end of the description.
Even if a newsgroup already exists, its description and mode will be checked and updated by any applicable newgroup or checkgroups message.
All changes made to the database will be logged, starting with the word
ACTION:.
The database of active groups, updated as described above. It is locked
against simultaneous access by using fcntl locking on a file by the same
name but with .lock appended.
The root of the control message archive. It is locked against
simultaneous writers by using fcntl locking on a file named .lock at
the top level.
The control.ctl file used to determine which control messages should be applied to the active newsgroups database.
The keyring used to verify PGP-signed control messages. The user IDs on the keys stored in this keyring must match the user IDs expected according to the rules in control.ctl.
Where actions are logged. %Y is replaced by the current four-digit year and %m by the current two digit month.
Used for temporary files for PGP verification.
Written by Russ Allbery <eagle@eyrie.org>, based very heavily on controlchan by Marco d'Itri. Portions based on pgpverify by David Lawrence.
export-control(1), update-control(1)
This script is part of the control-archive package. The control-archive web page at <https://www.eyrie.org/~eagle/software/control-archive/> will have the current version of the package.
| < generate-files | Russ Allbery > Software > control-archive | update-control > |