DSPAM integration

[bofh]

Hello,

Ife see dspam is used with hash driver at the Moment.

I would recommend the following:
DSPAM as primary antispamfilter
Use with Mysql Driver instead of hash
Better preconfig of dspam itself
User based Configuration in the account
Groupbased Configuration on admin or domainadmin level.

With dspam its possible to have an absolute individual configuation of the Filter which is for Service Providers a must have. Its currently possible to set the most important Settings per User out of the Box.

Also things like spamtraps are possible and groupbased (for example domainbased) common trainings and spamtraps and config for spamtraps out of the ui which work awesome on dpsam


The current issues in Dspam (standalone) from my point of view are the quarantine box and the webui.
Both wont be needed in Zimbra since we have a Spam folder which used instead of quarantine and autodeletion after x days.
the webui is also not needed if those options (only a few mysql rows needed for that so not a real effort)

only groups are in an txt file
Rest will work out of the box and would solve a lot user whises i read here including of course my own :-)

That way every user can decided which kind of trainings, how agressive it should be, turn spam checkin off or not. which mails should be trained which not (user behaivior) if autowhitelist should be used or not

would be a big whish for me and a lot of other people - support by the dspam community is granted for such an implementation

[phoenix]

would be a big whish for me and a lot of other people - support by the dspam community is granted for such an implementation

If you wish to see any changes in Zimbra then you should file a bug report (vote on it and post the bug number here so thers can see it and vote), that will get to the attention of the developers. Of course, Zimbra is Open Source and you could always implement that configuration yourself.

[bofh]

this was posted under development which is meant to discuss such things

@phoenix
why dont you stop giving such useless comments without anything about the topic and stop moving threads without a reason

i file now a complain about you

[phoenix]

@phoenix
why dont you stop giving such useless comments without anything about the topic and stop moving threads without a reason

The correct place for a request to improve or extend Zimbra functionality is in bugzilla as a RFE (Request For Enhancement), that's what will get attention of the developers not posting a comment in the forums. You're certainly welcome to post it in the forums for additional comments but bugzilla should be the final resting place for your RFE. If you don't find advice on the correct place to post your 'request' then that's fine with me.

i file now a complain about you

Fine, feel free to do so.

[Dirk]

Speaking as a server admin rather than a developer I'd say this sounds great, anything that offers 'more' of something is a good thing..... generally (!)

From a developer side, I've no idea how it would be implimented, or even if it's sensible to do so. Most, if not all of the work to impliment this would be done by Zimbra employee's, most of whom dont spend much time reading the public forums (that's these forums).

The dev's do however, read and use the bugzilla, which is why phoenix pointed you in that direction. Discussing it here is fine, and encouraged, but the discussion of it wont lead to an implimentation unless it's logged as an RFE.

Hope that makes sence.
Cheers.

[uxbod]

Agreed Dirk ..

Would be nice to understand how it would play with Postgresql. I have been using it more and more instead of MySQL and pretty impressed with its speed and functionality.

I would be nice if Zimbra could support either

[phoenix]

Would be nice to understand how it would play with Postgresql. I have been using it more and more instead of MySQL and pretty impressed with its speed and functionality.

My experience is that current releases of PostgreSQL are far better than MySQL and Standards Compliant.

I would be nice if Zimbra could support either

At the risk of sounding repetitive - there's an RFE in bugzilla for it if you'd like to vote or give some input to it.

[uxbod]

For completeness Bug 2363 - support PostgreSQL

[ArcaneMagus]

As a side note bofh, the Developers forum is more for discussing working on Zimlet code or the Zimbra source directly. Topics like this are more for the Administration of a Zimbra server. I think the topic of better integration with DSPAM has even been discussed here before...

Oh and I'm voting on the PostgreSQL Bug right now

[bofh]

About Postgres and dspam :-)

It doesnt matter if Postgres or Mysql as a DSPAM backend.
Actually the postgresdriver for dpsam got rewritten and should be in the current stable or in the next one.
It has improved performance and less overhead..

reason why i said mysql is because dspam already use mysql.
all tokendata is stored in Xsql Db also userpreferences (except hashdriver this is another animal and just kept for small scale installations of dpsam and historical compatibility)

i think many people dont know much about dpsm and differences to spamassasin so here a shot note:

dspam works absolute different to the common spam filters. it works with statistical data instead of scores. it support different methods/algorythm for this (all well documentated).
it analyse (based on the method) every mail and build a token data of it - its like a checksum but not only for the complete mail also for parts of the mail.

that way its not nessesary to update frequently with new definition files and stuff it build its own - userbased - or group - systemwide.
the only small downside is that its database can grow a but big (1gig and more are seen out there 700 meg are common depending on the algorythm)

besides that the performance is very good. and it really do not matter so much if you use mysql, postgres or another sql backend. it fetches only a couple of rows per mail and write back a couple. beeing that a bottleneck is not reported by anyone (and its common used in bigscale installations)
because the effort of an installation was a bit high and the project not so well known there are more big scale (10k mailboxes and more) out there


so about the the integration.
its not nessesarly needed todo by zimbra team itself. if we cant get and admin ui its possible todo the userinterface by an zimlet.

the only thing then is authentication. all whats needed (for personal preferences) is stored in one simple sql table
userid/setting/value


ps: its clear that no developer will follow this here. first i wanted a discussion and ideas about it bevore we file a bug featurerequest
to the request is as complete and detailed as possible - so if we can make the life of developers easier its might help a bit to get it done :-)
and of course people vote for it too

hey make a deal i vote for postgres and all postgresfans vote for dspam hehe