https://bz.apache.org/SpamAssassin/show_bug.cgi?id=8296
Bill Cole <billc...@apache.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
CC| |billc...@apache.org
Resolution|--- |INVALID
--- Comment #2 from Bill Cole <billc...@apache.org> ---
Users are free to do as they wish with their configuration, but it is not a
good idea to ignore anything that is likely to have common features that
correlate to the ham/spam classification. The specifics of the domain,
selector, and body hash in a DKIM signature are likely to correlate.
As for the others you mention, I don't think those are common enough for the SA
project to judge whether or not they are "noise" in regards to Bayes
classification. In my corpora of half a million messages, mixed ham and spam,
that pattern is nowhere to be found. Whether those are noise or useful, we have
no way of knowing because we have no information as to what they mean or why
they exist. They are likely to be the product of one site's filtering process
and so are likely to be useful for Bayes, IF they are present.
A default configuration change is not a bug unless the existing configuration
is causing concrete identifiable errors. If you want to discuss these changes
with the user & developer community, the place to do that is the SpamAssassin
Users mailing list. See
https://cwiki.apache.org/confluence/display/SPAMASSASSIN/MailingLists for how
to subscribe.
--
You are receiving this mail because:
You are the assignee for the bug.
