requested changes / are these difficult?

[Harry Putnam <reader@newsguy.com>]

It seems that scoring works on a whole group at once.  That is, you
cannot limit the view one way or another and score just that view.

Some kinds of scoring (on head and body) are currently not allowed in
agent mode (unplugged).  Can that be changed with out major revamping?

These suggestions would only work well in agent mode with the articles
being manipulated already stored on disk.

A further stipulation would be that this kind of scoring could be done
off line.  And not really be connected to the current scoring calls.

If scoring could be aimed at a 'limited' view, or at least only the
articles present when an <RET> is done on a group, then the usefulness
of limited views would increase 10 fold. (actually it would
theoretically increase infinitely - but 10 fold is a good start) : )

You could limit the view to anything scorable by using the 
'<number> / v' .  I think this would be quite an advantage.

This would be a temporary scoring not initiated when a group is opened
like the current scoring does.   But something that could be invoked 
by the user when in a group.  And would not use the current score
file but a temp. file created for each session, if desired.  Once this
temp file was called up, all the score writing commands would apply to it. 

Apply a similar mechanism as re-search-forward to the scoring of 'body'
or 'head' strings.  Currently those kinds of scoring are not possible
unplugged.

All this would add up to a way to have a limited view of a regexp
search on the body, or an arbitrary header.

These functions are already possible when plugged.  The biggest problem
is that scoring is carried out on-line even if the articles showing are 
cached.  Consequently, scoring on the body is brutally slow.

Being illiterate with lisp and most other things pertaining to computing,
make me imminently  *un*qualified to know if this line of reasoning is
barking up the wrong tree, so somebody stop me if it is off base.

Here briefly is how I envisage using such a setup:

Enter a group (comp.unix.wizardlike.commands), by-pass the groups
'.SCORE' file.  Run an empty .TEMP score file with V R to reduce all
scores to 0.

Pick a subject you are craving to know all about.  In this case, you've
forgotten the unix 'find' commmand that allows you to run 'find' in such a
way that it tears through your disk and finds every last instance in
every common file where people are instructed to forcibly pour milk down
the throats of there pets. You remember it has the characters {} in it.

Use the 'I' (increase) pick 'b'= body,  type of search p=regexp (no
permanence choice will come up in this scheme) 
At the 'raise:' prompt we insert: 'find.* [{}]'<RET>

The scoring tears through the 800 articles showing, and elevates about 50
of them.  You think hard and remember the word 'grep' should be in there
so now limit the view to the first scored articles (50), and run another
higher score where you enter 'find.*grep.* [{}]' 

Sure enough, this turns up several examples where you are able to
construct a command:

find / type -f -exec grep 'milk' -il {}\;

Finally rerun the real .SCORE file or just wait till you open the group
again and it happens anyway.

Of course, uses may vary.  : )
 
-- 

Harry Putnam  reader@newsguy.com