Karsten Bräckelmann wrote: >> Here's my workaround. It involves some redundancy, but it does the trick: > > After some brief moment of head scratching... > > The "workaround" basically is just weighting sub-rules in the meta, and > works regardless if it is meant to be the individual sub-rules' scores > or not. It is useful, IFF one actually wants to weight sub-rules, which > in practice is very uncommon.
I agree. I use it for my DNSBL overlap rule to essentially say: If there are too many points from DNSBLs, subtract some. Uncommon, but sometimes useful. >> # supposing: >> score rule2a 1.3 >> score rule2b 1.6 >> score rule2c 0.8 >> # you can compare score sums like this: >> meta NAME rule 1 && (rule2a * 1.3 + rule2b * 1.6 + rule2c * 0.8 > 2) > > FWIW, the weights in this example are indeed redundant. Given the > weights, any 2 sub-rule hits is greater than 2 -- dropping the weights > from the equation and requiring >= 2 or > 1 instead does the same. :) That example may have been overly simplistic, but I thought it conveyed the idea. To see a real-world example, see KHOP_DNSBL_ADJ in http://khopesh.com/sa/khop-bl/khop-bl.cf (though please use the actual channel if you're going to use my rules, otherwise you won't get updates).
