Feature request: exclusion sets

[johnhtodd]

For each rule subset (basically: everywhere a matching string can be specified) we need to have an exclusion capability.

Example: matching for "paypal" often ends up with "people" as a matching element, either nyiis, or soundex, or metaphone or possibly other ways like levenschtein and others if they're loose enough. We should exclude any matches for "people" but that becomes a manual process of tuning.

1. We need to have a way to include strings that never trigger a match even though a rule might match on them, either in the file itself (an array) or a call-out to an external file that bulk-loads the strings on startup or HUP of the process.

This would be done after the matching process but before counters.

2. There also needs to be a metric which counts how many times matching elements have been dropped due to the exception strings, so we can understand when load is dramatic but output is small.

[johnhtodd]

And the most obvious use case: we need to exclude the "owner" of the target string as well (in this theoretical example, we would add "paypal.com" to the exclusion list. HOWEVER... this raises another point. Should we have this be a regexp? We would always want to exclude "paypal.com" at the end of the string, but NEVER want to exclude it somewhere in the middle of the matched string (foo.bar.paypal.com.baz.xyz for example)