On Thursday 03 March 2005 08:59 am, Nick Harring wrote:
> > > > > I don't think vdelivermail or vpopmail in general should be
>
> calling
>
> > > > > spamc/spamassassin.  Let that be handled elsewhere.  Let's stick
>
> to
>
> > > > > delivering mail and deciding where it goes.
> > > >
> > > > However, lets remember that if spam is only scanned at the MTA
>
> level,
>
> > > > SpamAssassin user preferences will not function if the e-mail is
> > > > addressed to more than one sender.  Scanning in vdelivermail, at
>
> the
>
> > > MDA
> > >
> > > > level, does not have this restriction.  For that reason I still
>
> think
>
> > > > there is value in scanning in vdelivermail.
> > >
> > > Obviously this is a current limitation in simscan
> >
> > no, it's a limitation in how SMTP works.  This has been discussed on
>
> the
>
> > simscan mailing list several times, please read my posts there.
>
> Simscan isn't replacing qmail-smtpd, so this isn't strictly an smtp
> limitation. Perhaps I'm just not getting it, but why wouldn't the
> following work:
> Email comes in for users A, B and C. A and B have an SA threshold of 5,
> C has a threshold of 9. The message scores at 7. Delete A and B from the
> recipient list when queueing the message, and tell qmail-smtpd to accept
> the message since at least one recipient will be receiving the message.
> Since the other two users consider it spam, they don't really care what
> the remote side thinks. Other scenarios are just as easy to work through
> in a way that'd work.

that would require queueing multiple messages from the same SMTP conversation.

what happens if on the 49th recipient of a 50 recipient message, the queueing 
fails?  Your 'solution' is ugly, and simply will not work.

> I know people think that it makes some huge difference in their spam
> receipt levels whether they reject spam during smtp or not, however I've
> not seen anybody actually try and prove it, so until then I'm skeptical.
>
> I would really like to be able to use this solution, but using server
> defaults or the first users preferences are just flat out wrong ways of
> handling email.

unfortunately, when scanning at the SMTP level, that's all you have.  If you 
want per-user filtering preferences, scan at delivery time.  Simple as that.

> They're not only wrong conceptually, but are a huge 
> breach of the users trust. They setup custom settings with the
> understanding those settings would be used.

yes, and again this has all been discussed on the simscan mailing list.  
Please read the several threads there, paying close attention to my posts.  I 
cover most of these issues in detail.

-Jeremy

-- 
Jeremy Kitchen ++ Systems Administrator ++ Inter7 Internet Technologies, Inc.
  [EMAIL PROTECTED] ++ www.inter7.com ++ 866.528.3530 ++ 815.776.9465 int'l
      kitchen @ #qmail #gentoo on EFnet IRC ++ scriptkitchen.com/qmail
         GnuPG Key ID: 481BF7E2 ++ jabber:[EMAIL PROTECTED]

Attachment: pgpDM9LOiJNEU.pgp
Description: PGP signature

Reply via email to