On Tue, 1 Feb 2000, Matt Morgan wrote:
> I have been doing an experiment with testing analog's accuracy for someone
> who is pretty demanding about having exactly right answers (I keep telling
> him it's a comparative science . . .). In general, when I try to use
> manual methods to count up the same values that analog is reporting on, I
> get different answers than analog does. Of course, I could be leaving out
> a lot of steps. I'm hoping someone can give me some ideas why my method is
> wrong.
>
> For example, I first ran a grep -v [ip-address] on a logfile to remove
> requests from the company's firewall--to ignore all requests from internal
> sources, which they like to do. In analog, we achieve the same thing with
>
> HOSTEXCLUDE [ip-address]
>
> Then, if I run a wc -l on the output to count the total number of entries,
> I invariably get a higher count than analog's total requests. It varies
> from about 5% to 10% higher.
>
> I don't think it's corrupted log file entries; usually those run 1%-2% of
> the total requests (normally because of referrers with commas in them,
> unfortunately).
>
> What else could be going on? Or--how should I be using manual methods to
> verify analog's numbers?
>
First, I have never seen two log analysers produce exactly the same figures.
You should get much closer than that though.
In successful requests (most reports) analog only counts lines with status
codes 200-299 & 304. This probably accounts for the discrepancy.
All the gory details are in docs/defns.html.
--
Stephen Turner [EMAIL PROTECTED] http://www.statslab.cam.ac.uk/~sret1/
Statistical Laboratory, 16 Mill Lane, Cambridge CB2 1SB, England
"We can ask you to pay the full amount which you owe us if you:
(a) become bankrupt; or (b) die." Egg Credit Card Agreement
------------------------------------------------------------------------
This is the analog-help mailing list. To unsubscribe from this
mailing list, send mail to [EMAIL PROTECTED]
with "unsubscribe" in the main BODY OF THE MESSAGE.
List archived at http://www.mail-archive.com/[email protected]/
------------------------------------------------------------------------
