Re: [Pan-users] Ignore author in watched thread

[Duncan]

David WE Roberts posted on Fri, 21 Feb 2014 10:15:52 +0000 as excerpted:

> Thanks - my personal view is that the GUI should remove the need to
> modify the score file directly apart from advanced users. Ignoring
> someone who posts to a watched thread doesn't seem that advanced.

Actually, the way watching and ignoring works, that's integral to 
definition of watched/ignored, so it's not /just/ the GUI, altho the GUI 
contributes to the issue as it's not as flexible as direct-editing.  But 
a lack of understanding of the mechanisms and definitions involved is the 
real problem.

And direct-editing is /always/ going to allow finer control.  It's the 
nature of the beast, particularly with advanced functionality such as 
scoring that no GUI can fully cover, without being even more complicated 
than simply editing the file manually.

How pan scoring works is like this:

Normal scoring is relative: Each score adds or subtracts X scoring points 
on the article, and in the absence of an absolute score, the end score is 
simply the sum of all those additions and subtractions.

Thus, by normal scoring, you could say any post in this particular thread 
gets a +9999 (which by itself equates to watched), while this particular 
poster gets a -9999 (which by itself equates to ignored).  If said poster 
posts to that thread, the two would cancel out and (lacking any other 
applicable scores) said poster's posts to the watched thread would thus 
come thru as simply zero-scored -- normal posts.

But of course relative scoring allows you the flexibility of defining 
that poster's negative score to say -99999 (another nine) instead, if you 
REALLY hate that poster's posts THAT much, thus overriding a whole SERIES 
of other scores that would make normal posts watched.

Or, if you wish, you could assign a -10000, by itself just one point 
worse than ignored, so that a single watched would still leave it 
negative scored (-1, not ignored, but below normal), but that single 
watched plus just one point from some other score (say a subject line 
score, for example) would bring it back to zero/normal, and a watched 
plus just two points would bring it back positive (barely, +1).

*BUT*...

There's also absolute scores, and as implemented in pan's GUI, that's 
exactly what watched/ignored actually use, =9999 for watched, =-9999 for 
ignored.

The way absolute scoring works is that when pan sees one of those, it 
immediately sets that as the score, regardless of any previous relative 
score, and stops processing further scoring for that article.

So the first absolute score that matches always applies.  And since new 
scores are always appended to the file, a previously existing ignored 
(=-9999) when a thread is watched (=9999) will always appear first in the 
file and thus those posts will always be ignored, since if the previously 
existing ignored matches, pan will never get to the later watched for the 
thread for that post.  But if you watch a thread, THEN try to ignore a 
poster who happens to post IN that thread, the absolute watched (=9999) 
score will appear before the ignored (=-9999), and pan will see the =9999 
first and stop processing further scoring for that post, so pan will 
never see the =-9999 ignored score for that author, since it was added 
AFTER the watch-thread score.

Which is why the first paragraph says it's integral to the definition of 
watched/ignored, since as pan's GUI defines them, they're absolute 
scores, the first one that matches wins and nothing further down gets 
processed at all.

If you don't want it to work that way, the simple solution is to use 
relative scoring instead of the absolute scoring of pan's watched/ignored 
GUI.  Because if you tell pan to use an absolute score, ignoring anything 
else further down the list, that's /exactly/ what pan does.

IOW, it's a case of the computer doing /exactly/ as it was told to do, 
instead of what you /intended/ to tell it to do. =:^)


Of course there's two ways to arrange things so pan does what you 
/intended/ instead.

a) Edit the scorefile directly, so the absolute scores appear in the 
order you intend, regardless of the order you added them.  Basically, 
prioritize your absolutes.

b) EITHER using the GUI to set a relative score, OR editing the scorefile 
directly, change the absolute scores as necessary to get the same effect 
as the ranked absolutes.  In this case, the GUI option would involve 
setting a (relative) +9999 score on the thread you intend to watch, 
instead of using pan's watch-thread function to set an absolute score.  
That way, the absolute ignored scores would be the first absolute scores 
pan would see, and thus the ones it would use, regardless of any other 
relative or absolute scores that would otherwise apply.


> As with so many things, the rules engine and GUI seem to lack that final
> 5% - however given the limited developer resource and the general good
> functionality of Pan it is hard to grumble convincingly.

One could convincingly argue that the watched-thread function should set 
a relative +9999 instead of an absolute =9999 for that reason.  However, 
one could EQUALLY convincingly argue that in the presence of other 
relative scores (say a -100 for subject or whatever, which would take the 
thread back below watched, as long as nobody changed the subject so the 
subject score didn't apply) doing so subverts the purpose of WATCHED, to 
ALWAYS set it watched, absolute (tho of course with the caveat that no 
previous absolute score applies).

IOW, you'd have people who actually use scoring for /scoring/, but still 
want a way to set an absolute watched thread, up in arms then.


> One other thing that I think I have noticed  - rules can be limited to
> one month, six months or forever.
> 
> However I have not seen that rules which have lapsed are cleared out of
> the rules file.
> 
> This may be a feature (allowing old rules to be re-activated) or a
> deficiency which allows uncontrolled growth of the rules file.
> 
> For the GUI, I would tend towards the latter.

The slrn scoring functionality, which pan adopted, is simply too advanced 
to be properly covered in a GUI.  The GUI thus gives you limited starter 
functionality, but the intention is clearly to simply expose some of the 
simplest functionality via GUI, with the GUI option to edit the scorefile 
itself there for those who want something more advanced.

And I guess expiring scores are considered "advanced".  Tho it /would/ be 
useful to at least have a pan option, probably found in preferences in 
the GUI and on by default, to delete expired scores.

> Still, I am using Pan because I can't find a better alternative.

Indeed.  Tho on MS, I know there were quite some choices at least in the 
past.  Forte's Agent is of course the most famous in that regard, the 
reference by which all others certainly must be measured.  Are they no 
longer around?

And on Linux/Unix there's other options too, including the respected gnus 
emacs mode and the previously mentioned slrn, altho neither one have 
quite the combination of both GUI mode and text AND binary functionality 
that pan has, which is what has kept me a loyal pan user for over a 
decade now. =:^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman