After a more detailed investigation consisting of perusing the tmda
source code, I realize that there isn't currently any way to do all of
the things I outlined in my recent post on the tmda-user list (quoted
below).
Therefore, I am proposing that I write an enhancement to tmda that would
work as follows:
A new configuration variable would be created, called
PENDING_ACTION_HOOK. If defined, it would contain the name of an
executable program or script that would get invoked right before any of
the following actions are taken on a pending message:
release
delete
whitelist
blacklist
The program would be invoked as follows:
PROGRAM /path/to/pending/message ACTION
... where /path/to/pending/message is the full pathname of the file
containing the pending message
... and ACTION is one of the four actions named above
... and all of the tmda configuration variables will be in the
environment
This program could then perform any desired action, including to cause
the message to be "learned" in the way I described in my original
message.
One nice feature of this is that it would work both within tmda-cgi and
tmda, itself.
I don't think that it would be very difficult to write this and add it
to the code base. However, before I come up with a patch, I'd like to
discuss this with the rest of you, to get feedback, suggestions,
complaints, etc.
What do y'all think?
Here's the text of my original post:
> I'm using both crm114[*] and tmda in the following manner:
>
> 1. An incoming message gets filtered through crm114. Based on this,
> one of the following headers gets added to the message:
>
> X-CRM114-Status: SPAM definitely spam
> X-CRM114-Status: GOOD definitely non-spam
> X-CRM114-Status: UNSURE not sure
>
> 2. The message then gets routed through tmda. Based on my filter
> rules, the following occurs:
>
> if status == SPAM then
> store message in my "Spam" folder
> else if status == GOOD then
> release message to my default INBOX
> else (status == UNSURE)
> if sender has been previously confirmed then
> *** learn message as non-spam
> release message to my default INBOX
> else
> send out confirmation request
> endif
> endif
>
> Note the line flagged with "***". This involves piping the message
> through one of the crm114 utilities in order to cause it to be "learned"
> as non-spam. Is there a way to cause this learning to automatically
> occur without my hacking the source code of tmda?
>
> Also, want to do more learning under three other cases. Each of these
> applies to messages in the pending queue; i.e., those which were
> previously marked as UNSURE by crm114:
>
> A. The message has been sitting in the pending queue for the
> PENDING_LIFETIME, and it is about to be automatically expired and
> deleted. During this automatic expiration process, I'd like it to
> be learned as spam by crm114.
>
> B. Via tmda-cgi, I am manually deciding to release a pending message.
> Therefore, I'd like it to be passed through the crm114 utility and
> learned as non-spam as it's being released.
>
> C. Via tmda-cgi, I am manually deciding to delete a pending message.
> Therefore, I'd like it to be passed through the crm114 utility and
> learned as spam as it's being deleted.
>
> Are there any hooks in tmda and tmda-cgi that I can use for performing
> the "***" item above as well as items A, B, and C ... without my having
> to hack the source code of tmda or tmda-cgi?
>
> Thanks in advance.
>
> [*] http://crm114.sourceforge.net
--
Lloyd Zusman
[EMAIL PROTECTED]
God bless you.
_________________________________________________
tmda-workers mailing list ([email protected])
http://tmda.net/lists/listinfo/tmda-workers
