netnews@wnuxb.UUCP (Ron Heiby) (02/05/85)
Somewhere, I saw a question asking whether it was possible to determine what
news on a site is actually being read. I would love to be able to find out
that there are only 12-18 newsgroups actually used by my people. It would
make it much easier to be innocuous in terms of disk space and cycles used.
I am being pressured to reduce the space used by netnews. I already expire
several groups after 1 week instead of 2, but the net, mod, fa, chi, and att
groups account for about 30,000 blocks on my machine and various source and
build directories bring the total to about 50,000 blocks. So, I understand
the desires of the bill-payers to want to trim back.
At first, I thought to compare the modification times with access times for
the files in the spool hierarchy, but had to discard that, as expire opens
each file to check for an "expires" header line. I can't afford to stop
running expire for a week or two to do the experiment. This, on top of
the efficiency argument adds much weight to putting the "expires" header
information into the history file.
A built-in method of determining which newsgroups are being used would be
real nice. Now, before anyone gets steamed, I don't really care who is
reading what groups. I just care that *someone* is using some groups.
Perhaps a count could be kept somewhere of the number of times a new
group is "entered" in vnews, readnews, and rn. The only other method
that comes to mind is to have a program run in the shell script that runs
expire, before expire. The program would look at each file in each group
(backwards, maybe?) until it found a "hit". A "hit" would be any file
that was accessed more than five minutes since the latter of the file's
modification date and the last expire run. When such a file is found,
the "group used" count gets incremented and the program moves on to the
next group. It could probably be done with a moderately sophisticated
shell (/bin/sh) script.
Are there any volunteers to write such a thing? Has anybody already done
so? Do I have to get my wife p*ssed off at me for doing yet more "work"
at home? :-)
--
Ronald W. Heiby / ihnp4!{wnuxa!heiby|wnuxb!netnews}
AT&T Information Systems, Inc.
Lisle, IL (CU-D21)netnews@wnuxb.UUCP (Ron Heiby) (02/05/85)
I forgot that article files with more than one link (multiply posted) should
probably be ignored to avoid false "hits". After this runs for a week or
two, the candidate newsgroups should be examined and perhaps shared with the
user community to see if anyone can come up with a good reason to keep any
of them.
--
Ronald W. Heiby / ihnp4!{wnuxa!heiby|wnuxb!netnews}
AT&T Information Systems, Inc.
Lisle, IL (CU-D21)msb@lsuc.UUCP (Mark Brader) (02/06/85)
One basis for finding out what newsgroups are being read would be a program to seek out all the .newsrc files, and look at each entry in them. It would consider a newsgroup to be being read if [1] it was subscribed to, and [2] there is not a large count of unread items (determined by comparing with active). No, I haven't written one. Mark Brader
avolio@grendel.UUCP (Frederick M. Avolio) (02/07/85)
> Somewhere, I saw a question asking whether it was possible to determine what > news on a site is actually being read. I would love to be able to find out > that there are only 12-18 newsgroups actually used by my people. It would > make it much easier to be innocuous in terms of disk space and cycles used. Fred Blonder at umcp-cs wrote s subscriber shell -- reports how many subscribers each news group has by looking at everyone's .newsrc file. We run it once a week at night. Contact him (or maybe he'll (re-) post it?) at Electronic-Address: decvax!harpo!seismo!umcp-cs!fred Don't tell him I sent you. -- Fred Avolio {decvax,seismo}!grendel!avolio 301/731-4100 x4227
fred@gymble.UUCP (Fred Blonder) (02/08/85)
From: avolio@grendel.UUCP (Frederick M. Avolio)
. . . I saw a question asking whether it was possible
to determine what news on a site is actually being
read. . . .
Fred Blonder at umcp-cs wrote s subscriber shell -- reports how
many subscribers each news group has by looking at everyone's
.newsrc file. We run it once a week at night. Contact him (or
maybe he'll (re-) post it?) . . .
I just posted it to net.sources. It's article 60@gymble. This generates
a list of the form:
net.announce 30
net.general 28
net.announce.newusers 27
net.arch 9
net.ai 8
net.jokes 8
net.news.adm 8
.
.
.
The number on each line is the number of people who have that newsgroup
listed in their .newsrc file. It doesn't attempt to determine if they
actively read the newsgroup.
Don't tell him I sent you.
I'll get you for this. ;-)
--
Fred Blonder (301) 454-7690
Fred@Maryland.{ARPA,CSNet}
harpo!seismo!umcp-cs!fred