Elaborating is my favorite pastime :)
I mentioned the minimum score choice because while most of what we do is looking for ways to add points, sometimes we also want to subtract them in order to give credit, or alternatively, we sometimes don't want to subtract more than a certain number of points. So for the completeness of options, it made sense to include (though again, it's not nearly as useful as the maximum weight feature because of how filters are mostly used).
This might have been useful for instance in my FOREIGN/TLD set, where the TLD-[Region] filters are scored in the Global.cfg as 3 points, and then for each hit (which should be unique in this case) one point is subtracted, like so:
----- Global.cfg -----
TLD-ASIAN filter C:\IMail\Declude\Filters\TLD-Asian.txt x 3 0
----- TLD-Asian.txt ----- MAILFROM -1 ENDSWITH .af HELO -1 ENDSWITH .af REVDNS -1 ENDSWITH .af
I didn't want to score this too high because there are many cases where a reverse DNS entry is missing from a valid sender, but alternatively, I could have coded it to credit back more points then the score given the Global.cfg and upped the score of the Global.cfg like so:
----- Global.cfg -----
TLD-ASIAN filter C:\IMail\Declude\Filters\TLD-Asian.txt x 4 0
----- TLD-Asian.txt ----- MAILFROM -1 ENDSWITH .af HELO -2 ENDSWITH .af REVDNS -2 ENDSWITH .af
So a credit of up to 5 points could be deducted, and currently that would give a score of -1 if all three hit, but I might not want to give back 5 points and limit the credit to 4 points with a MINWEIGHT -4 entry (figuring that the points in the Global.cfg would then be added to the score from within the filter). This would allow a sender with a MAILFROM and HELO, or a MAILFROM and REVDNS to net only one point, but I could add 3 points for just a MAILFROM which matched, which might be beneficial in this instance. This would be useful in a many-to-many matching system for both positive and negative scoring.
I could see other uses such as pseudo whitelists which make use of negative weighting inside of the filter and may track various types of information, so this would protect from crediting back more points than was desired, but at the same time allow less credit than the total defined by the MINWEIGHT.
As far as one-to-one negative weight matches go, these would benefit from the functionality where the filter is stopped from processing after reaching a certain value, thus saving processing time as you described.
I don't see any immediate reason why you would need a MAXWEIGHT and MINWEIGHT in the same filter if that helps.
I must admit that it's kind of hard to come up with perfect examples since I have been trying to work within the current framework, however I would imagine that over time, there would be even better uses for limiting the negative weights applied within a filter.
Regardless of that, I would give my left nut just to have the MAXWEIGHT feature as you expanded on it along with the other things :) I'm pretty confident that this would increase my capacity by 25% with the addition of having the test stop on a MAXWEIGHT as well as an END.
Thanks again,
Matt
R. Scott Perry wrote:
Please note the minimum score in addition to the maximum one (I'm not sure if you got that, though it's not nearly as important).
I did see that -- could you elaborate on that one a bit?
--- [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]
--- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com.
