[7912] in www-talk@info.cern.ch
Re: Common Log format
daemon@ATHENA.MIT.EDU (Paul Phillips)
Thu Apr 13 01:59:12 1995
Date: Thu, 13 Apr 1995 01:45:49 +0500
Errors-To: procmaster@www19.w3.org
Reply-To: paulp@cerf.net
From: Paul Phillips <paulp@cerf.net>
To: Multiple recipients of list <www-talk@www10.w3.org>
On Wed, 12 Apr 1995, Brian Behlendorf wrote:
> I was going to say this is something you could do with a Perl script, one
> that parsed your config files so it knew accesses to /dirname/ =
> /dirname/index.html, but that information isn't necessarily available at
> log-analysis time if your site changes frequently.
Right.
> The real question is - what do you do when the requested object and the
> object actually delivered differ? This can be because of short cuts
> (DirectoryIndex in httpd, soft links in the file system, etc) or now content
> negotiation, where a request for /dirname/ could return /dirname/index,
> /dirname/index.html, /dirname/index.html3, or /dirname/index.cgi.
> However, there are good debugging-related reasons for knowing the actual
> request was.
I'm prepared to log both. My server will provide multiple logging
levels, and I will include perl scripts to turn any format into the
common log format for statistics packages. If it were one or the other, I
think most people would prefer knowing what data was sent to what was
requested.
> $HOST $IDENTD $AUTHUSER [$MDAY/$MN/$YR:$HR:$MN:$SC $GMT] "$REQUEST" $ERRCODE $LENGTH
This is a nice idea. The trick is getting support from the statistics
package writers and the server writers, because each would wait on the
other, I'd imagine.
-PSP
