Re: simscan/spamc with per-user spam config stored in ldap

On 07/12/2011 06:31 PM, Bob Miller wrote:
> Greetings,
>
> I expect this list is not the most appropriate place to ask this
> question, but I suspect I am most likely to find someone who knows a
> solution here...
>
> I am implementing simscan on a qmail-ldap server.  I wish to use
> per-user spam filtering settings and store them in the ldap database, so
> I have --enable-spam-user=y in my simscan config.  The logs show that
> spamc receives the argument of the recipient's email address.  In the
> ldap logs, I can see a search for the spamassassin attribute, and it is
> using a filter of ([email protected]).  Everything is exactly
> as the documentation says it should be.
>
> This particular server is expected to only have a single domain on it,
> and I plan on adding other ldap-based services to it later.  So when
> constructing the DIT, I did not set the uid to be an email address, just
> the local part of it (first.last instead of [email protected]).
>
> So the spamc lookup succeeds, in that it correctly identifies that there
> is no user by the uid of a full email address, and hence no per-user
> config, but that success is obviously not the desired behaviour.
>
> It seems I can tackle this problem from at least three different angles.
> I can try and make simscan pass only the correct portion of the email
> address to spamc; I can try and make spamc search ldap with a filter of
> mail= instead of uid=; or I can let spam be filtered with just global
> configs at the smtp level, and try to implement the per-user configs
> later - either when the lda has the mail or by scanning it after the
> mail has been delivered to a maildir folder.
>
> I think I have read all relevant documentation regarding the first two
> options.  I find nothing that indicates the values involved can be
> changed.  I also spent a few hours looking over code yesterday to see if
> I have what it takes to implement such changes, but my programming
> skills are very limited, and so far I have not made heads or tails of
> where/how I might accomplish that.  Unfortunate, because this seems the
> cleaner approach.
>
> I find my brain slightly opposed to the idea of doing additional
> scanning later.  It seems wasteful to scan and sort a message twice
> using the same program, and it introduces yet another spot in the chain
> where something might go wrong (or makes a situation where one thing can
> go wrong in multiple places).  In addition, it seems documentation on
> such an idea is extremely limited, and even there is mostly related to
> using postfix and procmail, so it may not be any less complicated to
> implement than my first two ideas.
>
> So, having read a great deal of this mailing list's archives, I did not
> find any solutions suggested.  But I have come to realize that there is
> more than one genius on this list.  Hopefully one of you has run across
> this already and would share your approach to this problem...
>

Hello Bob,

What are you going to do with the messages that are flagged as spam? 
Reject them at smtp talk time or just tag them (subject/message headers) 
and let the user decide what to do with them?

Regards,

Hugo Monteiro.

-- 
fct.unl.pt:~# cat .signature

Hugo Monteiro
Email	 : [email protected]
Telefone : +351 212948300 Ext.15307
Web      : http://hmonteiro.net

Divisão de Informática
Faculdade de Ciências e Tecnologia da
		   Universidade Nova de Lisboa
Quinta da Torre   2829-516 Caparica   Portugal
Telefone: +351 212948596   Fax: +351 212948548
www.fct.unl.pt                [email protected]

fct.unl.pt:~# _