Re: new package: wip/dcc (Distributed Checksum Clearinghouses)

On Tue, Dec 23, 2008 at 09:16:10AM -0500, Greg Troxel wrote:
> 
> Petar Bogdanovic <[email protected]> writes:
> 
> > On Tue, Dec 23, 2008 at 08:08:52AM -0500, Greg Troxel wrote:
> >> 
> >> Fredrik Pettai <[email protected]> writes:
> >> 
> >> > On Dec 23, 2008, at 3:21 AM, Petar Bogdanovic wrote:
> >> >> On Mon, Dec 22, 2008 at 08:24:42PM -0500, Greg Troxel wrote:
> >> >>>
> >> >>> (...)
> >> >>>
> >> >>> I added the dcc plugin line in spamassassin - will see how it goes.
> >> >>
> >> >> Don't forget to add:
> >> >>
> >> >>   dcc_dccifd_path /var/dcc/dccifd
> >> >>
> >> >> to your SA-configuration. Otherwise SA will use dccproc.
> >> >
> >> > Why not add that information to wip/dcc/MESSAGE?
> >> 
> >> Sure, or create /usr/pkg/share/examples/dcc/spamassassin.conf with the
> >> right lines to add (and then perhaps point to it via MESSAGE).
> >
> > I'll add a recommendation to MESSAGE (and rewrite MESSAGE on that
> > occasion), since that line is not crucial for DCC.pm. Using dccifd
> > instead of dccproc is just more efficient.
> >
> > All the other DCC.pm options (dcc_{body,fuz1,fuz2}_max, dcc_timeout,
> > dcc_options..) have pretty sane defaults IMHO, so if someone wants to
> > change them, he should read dcc(8), dccproc(8) or dccifd(8) first.
> 
> That works, but I suggested an example file so one could just cp or
> symlink it (I think, or insert it) instead of using an editor.  I have a
> bias against MESSAGE since I tend not to notice it, but I look for
> examples with pkg_info -L.
> 
> I'll amend my suggestion to
> 
> /usr/pkg/share/examples/dcc/dcc-spamassassin.cf
> 
> with contents
> 
> loadplugin Mail::SpamAssassin::Plugin::DCC
> dcc_dccifd_path /var/dcc/dccifd
> 
> since DCC is disabled in spamassassin by default.

Right, I forgot that SA will

	``first read files ending in ".pre" in
	  lexical order and then it will read files
	  ending in ".cf" in lexical order''

in various SA-configuration directories. So adding an example file could
really spare you an editor session.


> And then does one need to start dccifd?

Yes, otherwise the dccifd socket won't exist and SA will fall back to
bin/dccproc.

------------------------------------------------------------------------------