Hi, I'm not sure if this is valid ml to report issue. If not, sorry.
Recently I noticed some Japanese emails are detected as spam, and it is false positive actually. The cause of it seems to be TVD_SPACE_ENCODED and TVD_SPACE_RATIO_MINFP. Both of rules uses __TVD_SPACE_RATIO, and it matches Japanese characters easily, because Japaneses doesn't have spaces as word separator. Well, so, the question is, TVD_SPACE_RATIO checks !__ISO_2022_JP_DELIM with __TVD_SPACE_RATIO in rule. But TVD_SPACE_ENCODED and TVD_SPACE_RATIO_MINFP uses __TVD_SPACE_RATIO without !__ISO_2022_JP_DELIM. Is it intended one? BTW, if I added "&& !__ISO_2022_JP_DELIM" check to __TVD_SPACE_ENCODED and TVD_SPACE_RATIO_MINFP rules in .spamassassin/user_prefs, it seems to prevent the false positive for me. Is there recommended way to prevent false positive? meta __TVD_SPACE_ENCODED (__TVD_SPACE_RATIO && __SUBJECT_ENCODED_B64 && !__SUBJECT_UTF8_B_ENCODED && !__ISO_2022_JP_DELIM) meta TVD_SPACE_RATIO_MINFP __TVD_SPACE_RATIO && !__LCL__ENV_AND_HDR_FROM_MATCH && !__SUBSCRIPTION_INFO && !__RCD_RDNS_MAIL && !__SUBJECT_ENCODED_QP && !__THREADED && !__TO_EQ_FROM_DOM && !__BOTH_INR_AND_REF && !__X_CRON_ENV && !__HAS_THREAD_INDEX && !__HDRS_LCASE_KNOWN && !__ISO_2022_JP_DELIM -- OGAWA Hirofumi <hirof...@mail.parknet.co.jp>
