[Pan-users] Re: scoring on "References:" field doesn't appear to work

[Duncan]

cav posted <address@hidden>, excerpted
below,  on Sun, 10 Oct 2004 14:31:48 -0700:

> 
> I'm trying to filter out some spam in some newsgroups.  The From and
> Subject lines are totally useless to use for filtering, because as
> everyone knows, every single spam on usenet has random subject lines and
> from fields nowdays. When using the menu to create a score, it shows the
> References field as something to use.  This would be good, becasue the
> references field is generally something along the lines of
> <address@hidden>
> which as you can see, contains the actual server name used to post the
> message.  A given spammer generally uses the same news server to post
> their crap from, so this would be a good way to filter them.  However, I
> can't seem to manage to get scores to work on the references field.  When
> I create a score, (which ends up in the score file as         References:
> news\.usenetmonster\.com ) It never seems to match any of the articles I
> want to filter out.  Is there some special trick I need to use to get it
> to work?  Has anybody managed to get scoring to work on references?

At the risk of stating the obvious, the "references" header maps to the
message-ids of posts to which the current posts is a followup, IOW, to
upline posts.  While some spam follows up an initial post, and a score on
references there should score all but the initial post, this likely won't
be too successful in general because most spam either won't contain
references, will be spam followups to legit posts (which would mean legit
followups to the post would be scored as well), or would be just as fake
as the subject and author headers you mention, and therefore probably
rather useless in scoring.

This of course begs the question of why message-id isn't an available
choice.  Unfortunately, PAN's filtering and scoring capabilities are
/very/ limited by the fact that Charles only implemented them on a subset
of the overview headers subset of all headers for a post.  There's no way
to act on an arbitrary header, or the body, or the entire post.  While
message-ids /are/ indeed part of the overview, apparently, PAN's actions
can't yet match them, except as they appear in the references header, not
the message-id header.

So, yes, as the other poster mentioned, PAN /does/ use slrn compatible
score files, but it doesn't yet implement more than a subset of possible
scoring criteria.  (Note that I don't know what others implement, only
that even the much maligned OE could do better than /that/, several years
ago when I last used it!) 

This is probably my biggest frustration with PAN.  Sometime back when PAN
only had filters, not scoring, I asked about this, and Charles agreed that
yes, this was worthwhile to implement, but that it wouldn't be coming any
time "soon", as his only focus at that time was overview headers.  I read
this as saying it was something to consider before PAN 1.0, but something
that was, /like/ PAN 1.0, off somewhere in the fuzzy distant future.  Of
course, that was while PAN development was pretty active, as well.  Now
that the latest beta is getting close to a year old (well, January, so 9
months, but it's looking to be a year...), and real active development 
hasn't really occurred for another six months previous to that, PAN 1.0
looks to be scheduled perhaps about the time Linux overtakes MSWormOS on
the desktop, if even then.

Of course, I'm not a developer myself, and I won't throw stones at those
that are.  Charles' always stated that patches were welcome, provided they
fit the overall spirit of PAN (therefore, one could patch PAN to avoid
GNKSA, and Charles even told folks where to look to do so with various
requests, but they weren't going to be included in PAN mainline, he said,
because that was counter to PAN's spirit and goals).  If I had the skills,
I'd have coded this functionality long ago and submitted it.  I don't, and
apparently, it was never a high enough priority for those that do, to
bother with.

Actually, that's somewhat the situation with PAN overall, at present, as
well, it would appear.  While a number of developers submitted various
patches, only Charles and Chris could be said to be core team.  No other
developers seemed interested enough in PAN or graphic client binary news
functionality in general to create a larger core, so when Charles got busy
with "real life", PAN's development has inevitably languished.  It would
appear that coders with the necessary skills, simply don't prioritize a
GUI binary news client all that highly, so other development projects get
priority.  Those that /do/ tend to prioritize it highly, either aren't on
Linux/*ix, or aren't developers with the necessary skills.  I certainly
fall in the latter category.  Perhaps someday...

I've been thinking I need to go take a better look at KNode, once again. 
I haven't taken a serious look at it in years, and the nearly year and
a half PAN has been idle is a long time in open source development
terms...  Given that KDE is my X desktop environment of choice already,
save for PAN and XMMS...

-- 
Duncan - List replies preferred.   No HTML msgs.
"They that can give up essential liberty to obtain a little
temporary safety, deserve neither liberty nor safety." --
Benjamin Franklin