Re: spamassass-milter-0.2.0+cvs

[Cassandra Lynette Brockett]

Dan is trying to use a system call that is available to all users who run
SA/Sendmail/Spamass-Milter.

This has caused me a few hicups with code modifications I've provided in
the past - I forgot to take into account how Solaris handles things.

Your solution would kill the alias support for most of the sites I've
installed spamass-milter at, thereby making me go back to the old
hand-modifications of spamd to allow it to parse alias's, which is not my
idea of a good time... before moving to the new CVS of spamass-milter I
used to spend about two hours after an upgrade of SA to fix the alias
support I'd patched in to it, not my idea of fun...

In one of my favourite testing environments (where at least one user
receives 4000+ emails a week, and there are 200+ users) the additional
overhead of running sendmail -bv isn't a big deal, and it is the only
option I found that is cross-platform compatible (for instance Redhat
Linux has a wonderful command to parse the alias lists and give you the
local aliases related to the user entered - works for all users, but it
does not exist on Solaris, or even Debian).  Also sendmail -bv parses the
virtualuserstable, while most other options do not - and in a hosting
environment (which is the main system I deal with), not using it means
letting a pile of spam through the wall...

I can't remember when this was all debated but I know it was a few months
back - basically Dan decided to use sendmail -bv as it is the most
cross-system compatible of the available options and it handled
everything.

So yeah, adding a switch to add in the extra options if you intend to run
spamass-milter as an unpriviledged user is probably not a bad choice, but
trying anything else will just cause incompatiblities between sites and
therefore reduce the effectiveness of the milter.

I'm sure we will see such a switch in the next week or so, knowing what
Dan is like at coding I expect it sooner, unless he's got a massive code
overhaul he's working on...

--
Regards,
   Cassandra

Derek J. Balling said:
>
> On Oct 19, 2003, at 1:14 PM, Dan Nelson wrote:
>> Yes, I'll probably have to do both.  Although compared to the cost of
>> running spamassassin at all, the time spent running the 2nd sendmail
>> copy is probably in the noise.
>
> You can use getsymval('{rcpt_addr}') which will give you the result
> after all the virtusertable/etc. mappings (which is, really, the
> complicated part), and then have the milter tie to the
> /etc/mail/aliases.db, do a little bit of recursive checking using the
> rcpt_addr value to find who the final recipient is.
>
> It's not compatible with LDAP and stuff like that, but for the majority
> of installations, it'd work just spiffy.
>
> D
>
>
>
> _______________________________________________
> Spamass-milt-list mailing list
> address@hidden
> http://mail.nongnu.org/mailman/listinfo/spamass-milt-list
>