On Thu, Jul 02, 2009 at 02:32:03PM +0200, Steve wrote:
> Hallo Tom,
> 
> 
> > Steve schreef:
> > > I see a bunch of issues:
> > > 
> > > 1) You should NOT use SBPH with MySQL driver. The only driver able to
> > handle SBPH without issues is the hash driver. Please use something else 
> > then
> > SBPH. I would suggest OSB.
> > > 
> > 
> > There seem to be multiple issues with some of these configuration
> > options, notably combintaaions between Algorithm, Tokenizer and the
> > various storage drivers. Maybe dspam could check for these and exit with
> > a "configuration error" when incompatible choices are made?
> > 
> They are not incompatible by definition. It is more a technical limitation 
> then a real hard incompatibility. I understand that a "configuration error" 
> would be nice and dandy but where to start and where to stop? I mean what 
> conditions do trap a "configuration error" and which one are not going to 
> trap a "configuration error"? The possibilities for making errors in the 
> configuration are endless and a procedure to check them all would be huge. 
> Hard configuration errors are anyway trapped and DSPAM will refuse to 
> run/work/process/whatever if you set some options. But the above mentioned 
> condition is more or less a technical limitation and as time passes it could 
> be easily happen that a new release of the RDBMS could cope with the load of 
> SBPH.
> I personally would not extend the current code to check each and every 
> (stupid) combination of options. Some things are better left to be done 
> outside by humans then internally by code. But that's my personal opinion. I 
> don't own DSPAM. It's a tool for us all. If the majority of the user base 
> would like to have that extended error checking then we could look to 
> implement it in the next release after 3.9.0. Depending on how the checking 
> is done it could be however that the checking is going to eat some processing 
> time when using DSPAM. For me that would be a big no-go. I am all for 
> trapping hard configuration errors but leaving the soft configuration errors 
> to be resolved by the user/admin using DSPAM.

I also agree that we should minimize big-brother coding unless it will
actually be a fatal problem. For example, sbph being restricted to the
hash driver would make it more difficult to work on making sbph work
with standard SQL drivers.

Cheers,
Ken

------------------------------------------------------------------------------
_______________________________________________
Dspam-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dspam-user

Reply via email to