Something bad seems to have happened to scoring

Something bad seems to have happened to scoring

[Karl Kleinpaste]

My score rules are behaving differently than they ought, or used to,
not that long ago.

I'm looking at my own outgoing archive of Usenet-posted messages.  A
message is marked `Y' and italicized (scored down).  I wonder about
this, and mumble `V t', which gives me a *Score Trace* buffer showing:

("vanillaknot.com>" 10 nil s)  ->  /home/karl/News/all/SCORE
("cinnamon.vanillaknot\\|mesquite.charcoal" -10 nil r)  ->  /home/karl/News/usenet/SCORE
("kleinpaste" 10000 nil s)  ->  /home/karl/News/all/SCORE
("kleinpaste" -10000 nil s)  ->  /home/karl/News/usenet/SCORE

Short summary of triggered score rules:
- I score up on things that are followups to myself, globally, and
  this article was one in a longish discussion chain.
- But I cancel that in my "usenet" group.
- I score up myself globally, so I can easily find my own messages,
  and so that threads in which I'm participating sort to the top..
- Again, I cancel that in my "usenet" group.

Hm, it ought to be score zero.  But `V S' says "-10".

In a similar vein, I'm looking at articles in list.ding:  

O   28-Jan > [   6: Harry Putnam           ] Was I dreaming...?
R   28-Jan +---> [   7: -> ding@gnus.org       ] 
R + 28-Jan |   `---> [  10: Josh Huber             ] 
R   28-Jan |       `---> [  21: Harry Putnam           ] 
R   28-Jan |           `---> [  13: ShengHuo ZHU           ] 
O   28-Jan |               `---> [  14: Josh Huber             ] 

Josh's article, following up to mine, scores up, based on References
mentioning me.  But Harry's, ShengHuo's, and Josh's 2nd are not; `V S'
on each says "0", yet the `V t' *Score Trace* says:

("vanillaknot.com>" 10 nil s)  ->  /home/karl/News/all/SCORE

So it clearly ought to have score 10.

What gives?

Re: Something bad seems to have happened to scoring

[Frank Schmitt]

Karl Kleinpaste <karl@charcoal.com> writes:

>("vanillaknot.com>" 10 nil s)  ->  /home/karl/News/all/SCORE
>
>So it clearly ought to have score 10.

You didn't have by accident started using adaptive scoring?

-- 
One Ring to rule them all, One Ring to find them,
One Ring to bring them all and in the darkness bind them
In the Land of Mordor where the Shadows lie.
19. Dezember 2001

Re: Something bad seems to have happened to scoring

[Karl Kleinpaste]

Frank Schmitt <usereplyto@Frank-Schmitt.net> writes:
> You didn't have by accident started using adaptive scoring?

No, gnus-use-adaptive-scoring is nil.

Re: Something bad seems to have happened to scoring

[Harry Putnam]

Karl Kleinpaste <karl@charcoal.com> writes:

> In a similar vein, I'm looking at articles in list.ding:  
>
> O   28-Jan > [   6: Harry Putnam           ] Was I dreaming...?
> R   28-Jan +---> [   7: -> ding@gnus.org       ] 
> R + 28-Jan |   `---> [  10: Josh Huber             ] 
> R   28-Jan |       `---> [  21: Harry Putnam           ] 
> R   28-Jan |           `---> [  13: ShengHuo ZHU           ] 
> O   28-Jan |               `---> [  14: Josh Huber             ] 
>
> Josh's article, following up to mine, scores up, based on References
> mentioning me.  But Harry's, ShengHuo's, and Josh's 2nd are not; `V S'
> on each says "0", yet the `V t' *Score Trace* says:

Not sure what baring it may have but in the case above I put the
process mark on yours and Joshs reply then pressed F.  I think that
may cite only the first in references or something.  Yes it seems to.

For example in this little thread you started:

 3R. 28-Jan+[  41: Karl Kleinpaste     ] Something bad...
 2R. 28-Jan     [  14: Frank Schmitt       ] 
 1R. 28-Jan+        [   5: Karl Kleinpaste     ] 

If I put the process mark on all three I get only the first in
References:, when pressing F  -- I think that might qualify as a bug.

This doesn't do much to explain your problem but it may complicate it
a bit.

Re: Something bad seems to have happened to scoring

[Karl Kleinpaste]

I figured out _what_ caused the problem.  I still haven't a clue as to
_why_ it's a problem.

I score up on followups to myself, by detecting my domain name within
References.

(("references"
  ("vanillaknot.com>" 10 nil s)))

I've been using that for eons.  Recently, it had occurred to me that
it would be nice if I could /stop/ scoring up if my contribution to a
thread was too far in the past, so I had added this to the above
scorefile:

  ("vanillaknot.com>[ 	][^ 	]+[ 	][^ 	]+[ 	][^ 	]+[ 	][^ 	]+" -10 nil r)

That was an attempt to say "my domain, followed by whitespace, then
non-whitespace, then whitespace, then non-...", so that in the case of
at least four more entries in References, the mild 10 point score-up,
caused by a match in References in the first rule, would be canceled
by this rule due to the reference being too far back.  The bracketed
portions are SPC TAB pairs (and negated SPC TAB pairs).

It doesn't seem to work, but I don't understand why not.  More
importantly, however, is that the mere presence of this rule causes
all those other scorings to screw up.  The variation provided by this
rule is -10, which is the amount of error in the previous `V t'
listings.  But `V t' doesn't list this rule as _ever_ being triggered,
even in cases where it seems certain it should..

I am deeply confused by this.  If anyone has a clue, I'd be interested
in hearing it.  I've simply removed the rule for now, and everything
is as it should be.

--karl

Re: Something bad seems to have happened to scoring

[Paul Jarc]

Karl Kleinpaste <karl@charcoal.com> wrote:
>   ("vanillaknot.com>[ 	][^ 	]+[ 	][^ 	]+[ 	][^ 	]+[ 	][^ 	]+" -10 nil r)
...
> The bracketed portions are SPC TAB pairs (and negated SPC TAB
> pairs).

Your tabs are probably matching the field-delimitor tabs in the NOV
lines; any whitespace in a References field should just be plain
spaces.  What happens if you remove all the tabs from the above, with
or without brackets around the non-negated spaces?


paul