[Mailman-Users] Mailman bounce counters and delivery of spam

Albert ARIBAUD albert.aribaud at free.fr
Sun Jan 31 19:10:06 CET 2010


Mark Sapiro a écrit :
> Albert ARIBAUD wrote:
>> Well, I can't argue that in this instance free.fr does a better job of
>> detecting spam than the list, but a) free.fr does not provide mailing
>> lists beyond the most basic functionality, and b) I'm not in position to
>> suggest moving the list as I'm not the owner, only a member.
>>
>> Regarding your suggestion of better spam detection at the earliest
>> stage, I agree wholeheartedly. But while I might be able to suggest
>> mailman settings to the list owner, heavier actions such as changing the
>> spam detection methods in place on the list are out of question.
> 
> 
> So you agree that the problem is that the management of the Mailman
> installation and/or the list is not effective in keeping spam off the
> list, but you ask us change Mailman to solve it.

Uhm... Precisely no, I don't ask that. I even specifically asked for any 
solution *except patching and recompiling* mailman! What I am after is 
advice on *configuring* a mailman setup so that at least spam rejections 
don't increase bounce count.

Note also that this does not imply that spam rejections should be 
ignored altogether; if someone can advise on a solution where spam 
rejections do not increase bounce count *and* get signalled to the 
mailman list/server owner (so that he knows his own spam protection 
goofed), then I'll be quite happy with it!

> Note that if the list server had spam detection as effective as yours,
> there would be no problem. Perhaps you can change your own spam
> detection to just drop rather than reject a message with Precedence:
> list.

I'm afraid I cannot influence my provider's spam rejection at SMTP 
session time. I have some control mon mail that was accepted at SMTP 
time, but that's too late wrt my bounce counter problem.

> Anyway, a request to change Mailman to mitigate the effects of
> ineffective management is not necessarily unreasonable, but I think it
> is in this case. The definition of a 550 in both RFCs 2821 and 5321 is:
> 
>    550  Requested action not taken: mailbox unavailable (e.g., mailbox
>       not found, no access, or command rejected for policy reasons)
> 
> You ask that any 550 (or 5xx) that mentions spam be ignored by Mailman,
> but what about all those 550s generated with text like "Service
> Unavailable" or other generic reasons which I suspect are as common in
> rejections of spam as reasons that mention spam specifically.

Uhm... No, I don't ask that *all* 5xx rejects be ignored, only that 
*those 5xx rejects that mention spam* be ignored, not that *any 5xx* be 
ignored--see my original post. Indeed, ignoring all 5xx rejects would 
render the whole bounce counter idea useless, so I'm definitely not 
talking about this.

> Mailman's bounce recognition is already overloaded with kludges and
> heuristics. Adding even more for this is not going to happen.
> 
> And, my suggestion to change your spam detection is only intended to
> encourage you to think about what a can of worms you'd be opening when
> you try to do things like this.
> 
> As far as suggesting changes to the owner goes, what is the list's
> policy for non-member posts? Discarding or Holding them blocks a lot
> of spam. Rejecting is bad because it will bounce spam to the From:
> address which is certainly forged, and Accepting just invites spam
> on the list.

They're discarded as far as I know -- the owner is aware -- as I am -- 
of 'spam fighting 101' rules such as the "no open list" and "no bouncing 
spam to From:" ones you raised here.

> Also, if the list allows subscription without admin
> approval, new members should be moderated by default.

I'm not sure about moderation. I will forward this point to the list owner.

> If the list is active (at least 1 post per day) and spam is only
> delivered every second or third day, setting bounce_info_stale_after
> to 1 and possibly reducing the threshold score may solve the problem
> without impacting bounce processing for dead addresses.

Thanks. Seeing as the list in question has had on average fifty messages 
per day during the last ten days, the conditions you state are met. I'll 
forward your suggestion to the list owner.

Amicalement,
-- 
Albert.


More information about the Mailman-Users mailing list