pkg://kwebwatch-0.71-1.i386.rpm:66846/
opt/
kde/
share/
doc/
HTML/
en/
kwebwatch/kwebwatch-5.html
info downloads
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Final//EN">
<HTML>
<HEAD>
<TITLE>The KWebWatch Handbook: Accuracy</TITLE>
<META HTTP-EQUIV="content-type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="keywords" CONTENT="">
<META NAME="description" CONTENT="">
</HEAD>
<BODY BGCOLOR="#ffffff" LINK="#aa0000" TEXT="#000000" >
<FONT FACE="Helvetica">
<A HREF="http://www.kde.org/"><IMG SRC="logotp3.gif" BORDER="0" ALT="The K Desktop Environment"></A>
<HR WIDTH="100%" SIZE=2 ALIGN="CENTER" NOSHADE>
<P ALIGN="RIGHT">
<A HREF="kwebwatch-6.html">Next</A>
<A HREF="kwebwatch-4.html">Previous</A>
<A HREF="kwebwatch.html#toc5">Table of Contents</A>
</P>
<H3><A NAME="s5">5. Accuracy</A></H3>
<H3><A NAME="ss5.1">5.1 About Accuracy</A>
</H3>
<P>In cases where the web server is not configured to respond to HTTP "If-Last-Modified"
requests, KWebWatch cannot be confident of accuracy. These are the times when
the words "BEST GUESS" appear in the results field of the Properties Dialog.
For these URLs, you may have to do some experimentation to deduce whether or
not KWebWatch is guessing accurately.
<P>KWebWatch currently uses one of two algorithms for detecting changes in cases
where it has to guess. The default guessing method is a simple byte count.
If the byte count is changed, KWebWatch guesses an update has been found. This
is a very accurate method at the HTML level, but unfortunately it can fail at
the content level, particularly in cases where a site serves up random images
(often advertising) with each connection. While the image files may be different,
the actual content of the site is often unchanged, and KWebWatch reports a
"false positive."
<P>The second available guessing method may be useful in some cases where such
false positives are returned. To switch to this method, check the "Ignore tags
when making best guess" box on the Properties Dialog for the URL in question.
This will cause KWebWatch to alter its byte count so that everything inside
HTML tags goes uncounted. This may give more accurate results in some cases,
but also watch out for "false negatives", caused by faulty HTML which throws
the byte counting off.
<P>
<H3><A NAME="ss5.2">5.2 Accuracy Strategies</A>
</H3>
<P>A recommended course of action is as follows:
<P>- For URLs that report results without the "BEST GUESS" prefix, do nothing.
You should feel confident of the check's accuracy.
<P>- For URLs that show "BEST GUESS" in the result field, first do some
experimentation, looking for false positives. Namely, mark the page as
read, and then manually check for updates repeatedly. If KWebWatch
reports "BEST GUESS: Unchanged", and the content does not appear changed,
leave things as they are.
<P>On the other hand, if rapid, successive checks cause KWebWatch to always report
"BEST GUESS: Update Found", you are likely getting false positives. In this
case, try checking the "Ignore tags when making best guess" option and see
if it helps.
<P>
Unfortunately, there are some URLs that KWebWatch will not be able to
accurately monitor, no matter how things are configured.
<P>
<P ALIGN="RIGHT">
<A HREF="kwebwatch-6.html">Next</A>
<A HREF="kwebwatch-4.html">Previous</A>
<A HREF="kwebwatch.html#toc5">Table of Contents</A>
</P>
<CENTER>
<HR WIDTH="100%" SIZE=3 ALIGN=CENTER NOSHADE>
</CENTER>
</FONT>
</BODY>
</HTML>
