http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5780





------- Additional Comments From [EMAIL PROTECTED]  2008-01-17 21:07 -------
(reply to comment #10)
> see if the stuff nearby even seems like words?

That would be difficult to get working right and could make it easy to hide a
URI from SpamAssassin by surrounding it with artistic garbage that leaves it
visually standing out. Also, it's overkill. We want to find hostnames so we can
look them up in RBLs. For that false positives are better than false negatives,
because an occassional extra RBL lookup is not a big deal. But we want to avoid
false positives for things like the WEIRD_PORT rule that could cause overall 
FPs.

Right now I'm leaning toward just a tiny change just to the code that handles a
supposed URI that has no scheme and does not begin with www\d*\. or ftp\. to
only parse up to the first [:/] character. That will keep most uses of URIs
unchanged while avoiding the problems stemming from parsing garbage as links.




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to