[milters] Archive

Lists Index Date Thread Search

Article: 1065
From: Ken A
Date: 2006-07-21 10:56:51 -0400
Subject: Re: Q: separate milter-link policy for +test-links?

Removal...........: milters-request@milter.info?subject=remove
More information..: http://www.milter.info/#Support
--------------------------------------------------------




Anthony Howe wrote:
> Removal...........: milters-request@milter.info?subject=remove
> More information..: http://www.milter.info/#Support
> --------------------------------------------------------
> 
> Alex Broens wrote:
>> On 7/21/2006 10:04 AM, Anthony Howe wrote:
>>> Daniel Krones has suggested to me that it might be better to have a 
>>> separate policy for +test-links disjoint from the URI BL policy. I'm 
>>> sort of on the fence with this one.
>>>
>>> Part of me says, sure why not? Might get more people to try 
>>> +test-links without weakening the URI BL policy.
>>>
>>> But then the purest in me says no. If a link is bad from a valid 
>>> sender, then rejecting that message would notify them that they 
>>> probably made a mistake and to try again. However, discarding as 
>>> Daniel does would probably wouldn't help notify the sender.
>>
>> I agree.. discarding hits you back when you least expect it. 
>> especially when URL FPs (depending on your traffic, user base, etc) 
>> are not very frequent but when they hit you, they hit you bad.
> 
> Yes, but the BL tests are vasty different compared to the test-links 
> option.

I've found milter-link's default rbl check catches quite a bit of real 
spam, so I'm using it to reject, with some whitelisting in access.db. 
We're an ISP, so we get support mail at a dozen different addresses. 
Things like "I can't email my mom at typodomain.com" would be bounced 
back to the sender using +test-links, right? I would not want to reject 
based on a 'broken link'. There are just too many real world examples of 
why this would be considered 'unhelpful'.

Does +test-links follow querystring too, or does it chop off any 
?address_good&id=user@domain type stuff? (Don't want to build address 
databases!)

What if domains used by spammers are at a registries that do the 
'Sitefinder' thing? Will all links resolve?

I would be testing it if it had a separate policy and I could really 
_see_ what it was doing.

Thanks,
Ken A.
Pacific.Net

>>> So I'm looking for the opinion of milter-link users. Simplicity or 
>>> flexibility?
>>>
>>>> Is there a way to have separate policies for milter-link?  One policy
>>>> for RBL check results and a second (different) policy for the
>>>> -test-links option?  I like them both by the way.
>>>>
>>>> I mostly trust the results from some of the RBL's and prefer to simply
>>>> discard the bad messages but don't completely trust the -test-links
>>>> test because it is too easy for a link to be accidentally broken by
>>>> the sender.  I would prefer the -test-links option policy to be a
>>>> modified mail header X-milter-link- something so I can put the ones
>>>> flagged as simply broken links into a user viewable spam folder.
>>
>> yep. and X-header which can be use to score and avoid eg: SA's slow 
>> lookups would, imo, be a plus.
> 
> milter-link already has the X-milter-link-Report: header for policy=tag 
> and policy=none. This would suffice for the purpose of SpamAssassin 
> scoring.
> 

Lists Index Date Thread Search