http://bugzilla.spamassassin.org/show_bug.cgi?id=3021





------- Additional Comments From [EMAIL PROTECTED]  2004-02-10 02:38 -------
I would like to implement this but I would like to get some feedback from
developers before I do any coding.

I'm leaning against making it a blacklist option for the following reasons: 1)
It logically fits in with the whitelist entries, so would require a blacklist
entry for each address that is whitelisted; 2) If an email address can be sent
through more than one mail server the syntax to specify the negation of them
would be ugly; 3) More options mean for more complexity for the user.

What I would like to do is have something that is generated automatically by the
whitelist_from_rcvd and def_whitelist_from_rcvd entries.

Here is one proposal: When checking for a whitelist match, if the result is not
successful and while searching it is discovered that a sender email address
matched the address portion of a whitelist_from_rcvd or def_whitelist_from_rcvd
and the mail server failed to match, then trigger a FORGED_SENDER or a
DEF_FORGED_SENDER rule.

For example, if there is def_whitelist_from_rcvd [EMAIL PROTECTED] amazon.com 
and you
get a mail From [EMAIL PROTECTED] sent through [EMAIL PROTECTED] that
would trigger DEF_FORGED_SENDER.

The advantages are that it works automatically to provide a good spam test for
every whitelist entry, and can be implemented with just a minor change to
check_whitelist_from_rcvd to make it instead of returning only 0 or 1, possibly
return -1 if there was no full match but there was an email address match
without a server name match.

The disadvantages are:

1) Say you have a friend [EMAIL PROTECTED] who sometimes sends mail through the
mail.example.com server and sometimes is on the road sending the mail through
various Internet cafes. You want to be able to whitelist mail from
[EMAIL PROTECTED] when it is sent through mail.example.com, but you want to be
neutral regarding mail from [EMAIL PROTECTED] sent through elsewhere, as it may 
or
may not be legitimate. This syntax provides no way to express that.

2) If there is a def_whitelist_from_rcvd entry for [EMAIL PROTECTED] 
paypal.com, for
example, you would not be able to add a whitelist_from_rcvd for [EMAIL 
PROTECTED]
ebay.com. For that to work the test would have to check both the
def_whitelist_from_rcvd list and the whitelist_from_rcvd list at the same time.
On the other hand, this is a pretty farfetched case that could be handled by
requiring you to use undef_whitelist_from_rcvd.

I think we could ignore the second problem, but what about the first? Is it
acceptable to not be able to whitelist_from_recvd without penalizing mail snet
through different servers? Would we have to add a syntax to express that such as
whitelist_from_optional_rcvd, or adding an "optional" keyword to
whitelist_from_rcvd?

Any thoughts about this before I submit a patch to implement it?




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to