unique readers

Russell S. Ireland irelandr at btv.ibm.com
Tue Jun 11 18:56:14 UTC 2002

On Tue, 11 Jun 2002 10:47:53 PDT Joe St Sauver wrote:
>>What ways are there to count the number of unique USERS of my news server?
>It would probably be helpful to know more about what you're trying to do 
>with the value you construct, e.g., justify an upgrade ("EVERYONE is reading
>news these days!") or associate traffic with particular users for accounting 
>purposes or ? Do you care about readers vs. posters?

The smart answer is that I am trying to satisfy my management...  In reality, 
there are three slaves of a master server carrying internal groups 
exclusively.  The slaves are accessed load "balanced" using RR-DNS.  In 
addition, there are (currently) two other "normal" peers of the master.  The 
original assignment was to collect the number of articles read from each 
server and the number of articles posted from each server for a given month.  
I posted here and got a response stating that I could combine the daily 
reports and run innreport on that file to get a month.  I plan to do that.

Then the mission creep came in.  The idea is being tossed around to collect 
the number of people in our organization reading articles.  At this point, 
there are no plans to collect per user stats, yet.  I think the purpose falls 
more along the lines of your first reason.  Readership justifies existence.  
The more readers we have, the more bang management is getting for their money.

>>The NNRP readership stats unfortunately don't appear to be good enough.  Thes
>>daily stats count readership by IP address or number (if it can't resolve).  
>>Unfortunately we experience name re-use due to some of the methods (VPN, 
>>dial-up,etc) that people are using to get to our intranet and then to my news
>>server.  And there are also the few multi-user machines as well.
>What operating system are you running under?
SuSe on s/390.  I just handle the inn portion.

> Would something like the 
>expanded logging available from TCP wrappers handle the majority of what
>you need (e.g., do your multi-user machines run identd)? Is it an issue
>(or a plus) that TCP wrappers aren't integrated into INN's logging and
I had not considered that angle and would need to expand my knowledge along 
those lines.

>And do you care about corner cases such as:
>   -- a single user who has opened multiple concurrent sessions at the same
>      time from the same address? (e.g., some of the more agressive binary
>      downloading tools)

Don't care right now.  Same "person" as far as I am concerned.

>   -- multiple users sharing a single system (classic example: shared machine
>      in a common area, or officemates or roommates sharing a single system)

If it is a true multi-user system or each even a single user with multiple 
accounts, then each person is represented by a unique login, or see below.  
This IS something I am trying to capture.  I want each login/user who uses the 
news services.  In our organization this is more rare than the case that 
triggered my post and the need to capture this point: users of DDNS and our 
VPN/remote access solutions.  During the course of a day, many different 
"machines" could be using the same "machine" name, as the IP's get assigned 
from a pool.
>   -- multiple users sharing a single "personally assigned" username? (you'll
>      know it is happening when the same user "logs in" from two geographicall

I don't care about this case as it is rare.  While it does happen, it is 
against our internal policy for users to share a single login/username.  Not 
everbody follows the rules, however, the physical realities imposed by our 
workstations make this inconvenient in many cases.

>      remote locations at the same time)

>Would an NNTP equivalent of HTTP's cookie state setting and preservation 
>technology accomplish what you want to do?
Possibly.  I'm not knowledgeable in that area.  Mmmmm...  Cookies....  Lots of 
potential for abuse?


Russell S.  Ireland 

More information about the inn-workers mailing list