https://issues.apache.org/SpamAssassin/show_bug.cgi?id=7067

Kris Deugau <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[email protected]

--- Comment #5 from Kris Deugau <[email protected]> ---
(In reply to AXB from comment #4)
> why is this a FP?
> 
> score TO_NO_BRKTS_NORDNS_HTML               0.001 0.001 0.001 0.001
> score TO_NO_BRKTS_HTML_ONLY                 0.001 1.308 0.001 1.308
> 
> the rules do what they're designed for.

As of when the message was originally processed (Jul 8), and as of the rules
update from ~Friday or so:

72_scores.cf:score TO_NO_BRKTS_NORDNS_HTML    0.001 2.620 0.001 2.620
72_scores.cf:score TO_NO_BRKTS_HTML_ONLY      0.001 2.499 0.001 2.499

> Your sample clearly shows this and if anything, the senders should fix their
> sloppy msg generation as well do the right thing and get fcrdns

I don't have time to bully thousands of small senders like this into fixing
formal RFC-correctness things like this (no < > on the To: address). 
Apparently they don't send notices to GMail or AOL accounts or they would
probably have proper rDNS already.

I *have* actually tried contacting people about issues like this in the past,
and gotten one of:

1) No response, no change.  (No surprise.)
2) "uhhhh....   whut?"

Usually because the only point of contact I can find is the hosting customer
whose site the widget is on...  and they are NOT systems/server folks, so they
don't have a clue what needs fixing.

> If this msg was tagged as spam, then there were more rules involved or the
> threshold has been heavily modified

I probably should have included the full hit report originally.

I don't run with a modified threshold systemwide;  just a few on a per-user
basis.  (To be exact, 5 with a threshold under 5;  one at 5.5 due to.... yep,
rules like this causing otherwise legitimate - and unwhitelistable - website
form mail to get tagged;  5 domain accounts at 7;  a legacy hosting role
contact at 7.5;  and 5 more with a threshold of 8 which at one point or another
had FP issues.)

Content analysis details:   (5.6 points, 5.0 required)

 pts rule name              description
---- ---------------------- --------------------------------------------------
 0.0 HTML_MESSAGE           BODY: HTML included in message
 0.7 MIME_HTML_ONLY         BODY: Message only has text/html MIME parts
-1.5 BAYES_05               BODY: Bayes spam probability is 1 to 5%
                            [score: 0.0493]
 1.0 BROKEN_TEXT_1          RAW: 20+ short lines
 0.8 RDNS_NONE              Delivered to internal network by a host with no
rDNS
 2.6 TO_NO_BRKTS_HTML_ONLY  To: misformatted and HTML only
 2.0 TO_NO_BRKTS_NORDNS_HTML To: misformatted and no rDNS and HTML only

No custom rules hit, no modified scores except for Bayes (as of original
processing), not even a funky Bayes hit.  I notice that if it had hit BAYES_00
instead, and I had not modified the BAYES_* scores, it *still* would have been
tagged - the stock BAYES_00 score is currently -1.9;  I have it at -3.

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

Reply via email to