SA 3.1.3 + DCC 1.3.

Gary Mills
Thu Oct 26 01:24:05 UTC 2006

On Wed, Oct 25, 2006 at 08:54:22AM -0600, Vernon Schryver wrote:
> > From: Gary Mills 
> > >   3. If the local whiteclnt file whitelist a mail message, should the
> > >     X-DCC header say so?  Except for upward compatibility worries, I think
> > >     so.  I'm thinging of making the X-DCC header be something like this
> > >     in that case:
> > > 
> > >        X-DCC--Metrics: - 0; ok Body=0 Fuz1=0 Fuz2=0
> > The absence of an X-DCC header on whitelisted messages has been a
> > constant source of confusion here.  People think that DCC somehow
> > skipped that message.  It would be quite helpful if people could
> > tell when a message was whitelisted.  Rather than overloading the
> > `X-DCC-*-Metrics' header though, I'd recommend using a new one that
> > still begins with `X-DCC-*-'.  As well, it should clearly say that
> > the message was whitelisted.  `0' means nothing to most people.
> Why a completely different X-* header?  If that's really desirable, I
> don't much mind the extra code.  (I agree that putting the brand name
> in the field name of the header was a mistake.  For that matter, the
> brand name was a mistake.  The use of the DCC didn't turn out as expected,
> with servers all in the hands of DCC service providers.)

I think you can assume that the X-DCC prefix belongs to DCC.

> The '0' server-ID is not significant; it's the "ok" that matter.  In
> certain rare error cases now, the DCC client client emits headers like
>     X-DCC-brand-Metrics:  0; Body=0
> with a missing DCC client host name and bogus server-ID of 0.  I've
> changed and extended that case of X-DCC-*-Metrics header to include
> the client host name and to cover cases where the DCC server failed
> to answer but the local whiteclnt files blacklist the message with:
>     X-DCC-brand-Metrics: clientname 0; bulk Body=many
> I figured that would still be matched by SpamAssassin and other patterns
> that look for "bulk" or "many".
> The smallest change to cover whiteclnt blacklisting seemed to be
> to replace "bulk" with "ok". 

I'd prefer "whitelisted" or some other word signifying that DCC
has determined that the message should not be rejected.

> For whitelisting, would "OK" be beter than "ok"?  If all else were
> equal, I'd rather not add more pollution to the X-* namespace of SMTP
> headers.

This is a fundamentally different case.  Normally people look at
the counts in an `X-DCC-*-Metrics' header and assume that when one
count reaches some limit, the message will be rejected.  In the
whitelisted case, this will never happen.  Using a different header
makes sense in this case.  There are no metrics.

-Gary Mills-    -Unix Support-    -U of M Academic Computing and Networking-

More information about the DCC mailing list

Contact by mail or use the form.