On Wed, 15 Jan 2020 07:55:59 +0200
Henrik K wrote:

> On Tue, Jan 14, 2020 at 10:42:48PM +0000,
> [email protected] wrote:
> > https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7366
> > (In reply to RW from comment #8)  
> > > It's not about anyone seeing them, it's about avoiding name
> > > collisions between the core rules and local rules. At the moment
> > > there's no standard way of doing this. Someone people use long
> > > prefixes, but these reduce readability and bloat meta-rules. And
> > > there's no guarantee that someone wont pick the same prefix and
> > > contribute their rules to core.
> > > 
> > > L_  seems like the ideal candidate for a reserved prefix.  
> > 
> > Along with __L_ for subrules.
> > 
> > +1 from me. We can even add build tooling to ensure none such ever
> > get added to the base rules.  
> 
> Guys please continue on list.
> 
> If you want some L_ policy that's fine.  Not something I personally
> would use.

I doubt I would as as I've already gone too far down the route of
not caring, but there is a lot of advocacy for local prefixes.

 
> It would seem more productive to actually make spamassassin --lint
> output info messages (not errors) when rules are redefined.  And
> perhaps add a new tflag "redefine" (suggestions?) to suppress those
> warnings for intentional redefines.

That requires actual coding, and it only partially works. 

Let's say I have local rules:

body  __FOO ...
meta  FOO   __FOO ...
score FOO  0.001

and then in the middle of the night sa-learn downloads new rules:

header  __FOO ...
meta    BAR   __FOO ...
score   BAR  3.0

My high-FP informational version of __FOO then gets used in place of
the core version in a high scoring meta rule.







Reply via email to