>> You might want to look into the old Chickenpox rule. On 04/17/2010 03:04 PM, Alex wrote: > Yes, big help. That did it, using the default scores. This was > written a number of years ago. Do you think it's still safe to use > the default scores?
NO! I put some of the (previously) better-performing chickenpox rules into my sandbox a while ago to investigate this. It's still there: Now: http://ruleqa.spamassassin.org/?rule=/CHICKENPOX 2004: http://wiki.apache.org/spamassassin/MasscheckChickenpox They are abysmal; the best S/O was 0.339, which means it hit more ham than spam. > I still wish I had a better grasp on regex so I could write a > correct rule to catch these, as I think that is probably the best > approach. Maybe someone knows of a list of all the URL shorteners to > be used in a combo uri/meta rule? > > Since the whole point is to shorten the URL, I bet I could write > something that categorically checks for a URL that's short -- small > host part plus small pathname... Somebody on this list wrote a parser to actually parse shorteners to their obscured URLs. You're looking at something far simpler, which we can certainly try. I've checked in a test at r935257 http://tinyurl.com/sa-r935257 (using a shortened link seemed appropriate here). This adds two rules, URL_SHORTENER (which detects a known URL shortening service) and SHORT_URL (which notices a particularly short ccTLD link that does NOT use a known shortening service).
signature.asc
Description: OpenPGP digital signature