|
| 1: create a patch that makes it works
| 2: add that patch to bugzilla
| 3: wait
|
| :)
An do a Wiki entry :-))
|
| > the question behind it the best way to invoke dspam by amavis in
| > the
| > zimbra setup
|
| amavisd is not tested with dspam, if it works its lucky, but you will
| get better results using dovecot-antispam integration with dspam
| relearning
i know thats barley testet, however
1. theres no way around for zimbra amavis is the way togo, and zimbra is atm
the only real alternative to exchange
at least on a professional level.
2. it should not make a big difference. its just another layer in between.
amavis doenst do much more then postfix would do except it can forward
something to dspam, take it back, forward it again to something else take it
back and so on
more or less (ok i know this description isnt technically precise but just to
paint a picture)
so now, the question behind is just which parameters do we really need in amavis
and if we need to process or just classify.
in TOE it should not matter.
TOE doenst have a learning process by default, it need reclassification anyway
- which is done nightly by feeding all reclassified corpuses anyway.
so a dspam signature isnt needed - i think - i guess - i think so but dont know
for 100%
however in TUM or TEFT we may profit of em a bit more or dont we?
does the hole corpusfeeding make signatures absolute obsolete or may they be
helpful?
also the source , should we use "source = error" by default? in the zimbra
amavis setup we always have only one dspam user
i also dont see a way or think its even usefull to use virtual users other than
parsing the to field
i dont think ususal virtual users are possible with zimbra atm because at the
state where the mail is processed the users isnt know yet.
as far i know it goes trough the filters first then it goes into zimbra core
where it will be determined which user it is (among other things).
- but thats ok for now - for now we forget regular virtual users.
so from my understanding and perspective in that setup source=error should be
always fine and just enables optional the parsetoheader to virtually create
pseudo users/groups
so again what i really wanna know
1. which amavis parameter we should use - should zimbra use by default
1.2 is process needed in there or is classify fine too?
1.3 shouldn we use source=error to optionally use the parsetoheader?
1.3.1 or should we use source=error only if we want to use parsing option
2. anyone using dspam with amavis other than with zimbra?
any expiernces? suggestions? drawbacks?
3 anyone any idea why in hell many amavis docus tell us to use the trainingmode
and features at all? all those parameters are already set in dpsm conf , they
should not differ, and if they not differe i dont see a reason to specify em
again in amavis
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Dspam-user mailing list
Dspam-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dspam-user
