Local DCC 'server' and MS?

[Malcolm Ray]

> Peter Bates <Peter.Bates at LSHTM.AC.UK> wrote:
> >
> >Reading this
> >http://news.spamassassin.org/modules.php?op=modload&name=News&file=article&sid=56&mode=thread&order=0&thold=0
> >
> >and also upgrading to SA 2.60, which mentions this when doing a
> >'-D --lint':
> >
> >debug: DCCifd is not available: no r/w dccifd socket found.
> >debug: DCC is available: /usr/local/bin/dccproc
> >
> >... are people out there running their own DCC servers/peers, or know
> >what this 'dccifd' is?
>
> SpamAssassin 2.60 now supports two ways of using dcc: the old one involves
> running the dccproc program for each message; the new one talks to the
> ddcifd daemon via a socket. This should be more efficient because of the
> reduced startup costs.
>
> I haven't finished investigating dcc properly, so I don't know about
> running a local dcc server.

In case it's not clear, running dccifd is not the same as running your
own DCC server.

> Given dcc's principles, it would make sense
> if checking a message was the same as reporting it, but it isn't
> clear to me if this is how it actually works.

AFAIK, by default dccproc reports as well as queries.  I haven't tried dccifd,
but it's designed to be a simple alternative to using dccproc, so I believe
it's the same in this respect.  When I was testing it with SpamAssassin
2.55, I found that by default it supplied the -H and -R flags to dccproc;
I used dcc_options in spam.assassin.prefs.conf to add -Q to avoid reporting
(since I didn't want to taint the database with a bunch of old spam and
generated mail.  Pretty anal, since my tests would have made very little
difference).

DCC is plain weird, IMO.  The documentation is poor and the interface is
strange.  I never did figure out how to get it working with SpamAssassin
in the way I wanted, which was to just gather data for a while without
blocking anything.  For example, DCC's idea of logging seems to be to keep
a copy of the entire message.