----- Original Message -----
From: "Shawn Bayern" <[EMAIL PROTECTED]>
> > The same logging mechanism could be used for 'semantic logging' for
> > click-stream analysis and hooks into personalisation engines and the
> > like...
>
> Ha, I was just going to suggest something similar, along the lines of
> Apache's mod_usertrack. Since we get sessions for free, click-stream
> analysis can actually be quite simple; it's really nothing more than the
> 'semantic logging' you suggest.
Probably the same 'logging taglib' would do for all I think - its just a
question of where the data goes (file, database, JMS etc).
I haven't tried it yet - but I would have thought through log4j we could use
different 'Appenders' (in log4j speak) to allow different categories of
message to go to different output destinations. So benchmark 'categories'
could go to a large text file, sematic messages about the products &
services a particular user looked at could go to JMS etc.
> Any takers? I'd write it myself but will be a bit busy over the next few
> days. I can probably pick it up later if nobody else wants to.
>
> I haven't thought through the best way to let users analyze the data.
> Were you thinking (like me) that we could also provide a tool, or
> guidelines, to interpret the 'semantic logging' data we stored? Perhaps a
> tag-based interface to the data would also be a good idea, just to
> demonstrate the technique?
Trying to make it XML-friendly so any XML tool could process the output
would be highly desirable. There's a nice clever way in log4j to produce log
files that are parse-able using external entities here:-
http://jakarta.apache.org/log4j/javadoc/org/apache/log4j/xml/XMLLayout.html
<James/>
_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com
