rSpamD questions
Problem reported by Douglas Foster - Today at 3:26 AM
Submitted
I am trying to assess the potential benefit of moving to rSpamD, and the effort required to get there.   I have read the documentation, but it did not provide enough information.   Hoping an rSpamD user can add some clarity.

Goal 1:   Authenticate the SMTP Mail From domain and assign a disposition result:   
Assume:  
Methods I use to authenticate the SMTP Mail From domain
  1. SPF Pass, based on the SPF policy for example.com
  2. DKIM Pass, based on a DKIM signature for example.com or aligned with example.com
  3. Google Signature for example_com.*.gapssmtp.com, which i interpret to mean that Google has authenticated the address
  4. Table lookup on host name:   Local policy trusts that example.com is hosted on google.com servers, so the combination is authentic as long as the server name is authenticated by forward-confirmed DNS.  (Either HELO or Reverse DNS can be used for matchng.)
  5. Table lookup on IP address: Local policy trusts that example.com is hosted on IP address a.b.c.d 
  6. Exemption:   SMTP Mail From authentication is not required if the message From domain is authenticated by DKIM Pass for a signature aligned with that domain.
Interaction of authentication and reputation:
  • Reputation lookup says block:   Block the message, authentication does not matter
  • Reputation lookup says quarantine:   Quarantine the message, authentication does not matter.
  • Reputation unknown:   Quarantine the message, authentication does not matter
  • Not authenticated:  Quarantine the message
  • Authenticated and acceptable:   Allow the message, subject to other tests
  • Authenticated and highly trusted:   Whitelist the message   
I have similar requirements for authentication and reputation checks on the message From address and on the Reply-To address.

Making it work in rSpamD
It is clear that rSpamD can do the SPF Pass test easily, but everything else is blurry.

How many different workers will I need to write from scratch to make this work?    If more than one, is there a parsed object model of the message that is passed from one worker to the next, or does every worker need to do its own parsing?

Any insight is appreciated.

Reply to Thread

Enter the verification text