https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7500
--- Comment #6 from Adrian Pepper <[email protected]> --- In the two sample messages in the first attachment (in future I'll give better names and now I believe I can add further attachments after the bug has been created), in addition to the later headers tracing subsequent message history, an X-Originating-IP gets generated before the header I show you. Would those, in particular the X-Originating-IP affect the RDNS_NONE evaluation? As I said elsewhere, in our context (some of our contexts) simply changing the IPv6 link-local addresses to 172.16.* IPv4 has the effect of both removing RDNS_NONE and evaluating to "ALL_TRUSTED". (The obviously(?) second message in the "attachment" is simply the first with that change made). That is, the first message gets "RDNS_NONE and no "ALL_TRUSTED", while the second (with the IPv6 manually edited to IPv4) gets no "RDNS_NONE", but does "ALL_TRUSTED". That also jives with what I observe; users sending from thunderbird (or in my case, heirloom-mailx) using the recommended relay get IPv4 versions of that specific header, while users using Microsoft Exchange, possibly via the Outlook Web Application/Access (owa) get the IPv6 version. The latter users are essentially penalized about 7 points compared to the former. Starting therefore with 2.x or sometimes 3.x points, it's actually not difficult to throw in a few favourite spamassassin key words (or other things, such as lots of HTML, and attachment types) and get classified as spam. If we can decide on a transfer method, I should be able to send to you both messages without my obfuscating changes (which changes primarily consisted of substituting bogus list names for the real list server list names). -- You are receiving this mail because: You are the assignee for the bug.
