Hints, Tips, and Warnings related to scoring.

Warnings:

Scores out of order in article scan mode:

	If not all the articles in a newsgroup are scored, strn will
usually wait until either a key is pressed or until all the articles
are scored.  If you enter article scan mode before all the articles
are scored, strn may have to score articles when they are placed on a
screen "page".  This may result in scores which are out-of-order on
the screen.
	This problem occurs because strn attempts to fill pages with
articles before determining a score.  (Unscored articles are placed
at the end of the article list.)  I hope to have a solution for this
problem soon.

Editing score rule files while running strn:

	* Strn caches scoring rule files in memory while running.  You
can use the article level 'e command or the article scan mode
control-E command to edit scoring rule files while running strn.
If, however, you use an external editor while strn is running, strn will
not recognize that the scoring rule file has changed, until you edit a
scoring rule file (which flushes the cache), or restart strn.

Subjects with TAB characters or consecutive spaces:

	* In strn, subjects are stored in a "compressed" format.  The
compression changes TAB characters to spaces and reduces multiple
consecutive spaces to a single space.  For instance, the subject:
foo	bar  baz	    misc
is stored as
foo bar baz misc

	* This compression has a few consequences:
	1.  The %-expansions that are replaced with a subject use the
article's true subject, and not the compressed form.  This means that
they are not suitable for use in strn score files.
	2. The 'S' abbreviation (using the '"' command) will write out
the compressed format of the subject line.
	3. To be completely safe in scoring rule files, use patterns
like "foo\b*bar" for any subject with TABS or multiple consecutive spaces.
(The \b* will substitute for any pattern of blanks or tabs.

Speed tips:

	* In general, strn scoring will take as long as the slowest
rule.  Here is a rough ordering of speeds:
1.  (fastest) Cached header lines in a database
	 (From:, Subject, Message-ID lines.)
2.  The Xref: header line.
	(If your site uses the NOV database, and that database
includes Xref, then Xref will be as fast as From, Subject, or IDs.
Even if Xref is not in the database, it will be faster than others
since it is cached.)
3.  Any other header line rules.
	(Standard headers (Organization:, Keywords:, etc...) may be
slightly faster than non-standard ones (X-Moderator-Review), but the
difference will be minimal.)
4. (slowest) NewsClip scoring.
	(NewsClip is much slower over an NNTP link than with a local
spool, since the NNTP version has to make a full copy of each article.)

	* For fastest processing, use only the Subject:, From:, or
Message-ID: header lines.  For NNTP users or (many) NOV sites, the
Xref: line will be faster than other header lines.  If you use *any*
other header lines, strn will load the entire header into memory.
After the header is loaded into memory, all header line queries are
equally fast.  This means that if you use just *one* extra header rule,
you might as well use hundreds.

	* If you want to give crossposted articles a score, use the
Xref: header instead of the Newsgroups: header.  The Xref: header is
stored in News OverView (NOV) databases, and is cached in memory.
Using Xref: instead of Newsgroups: is *much* faster when using NNTP.
Here's an example:
#Give crosspostings to alt.flame a low negative score.
-50 Xref: alt.flame
