Re: WWWVL: gathering stats (usage)
On December 7, you (Keith Instone) wrote
:>Here is a proposal, off the top of my head, and
:>probably not the best way to do it:
:>At the same level as the "home page" for each area
:>of the Virtual Library exists a file called "stats".
:>The URL for this file can be easily computing by looking
:>at the URL for the home page. The URL for ALL of the
:>stats can easily be accessed by scanning thru the main
:>WWW VL page itself.
:>So, cern runs a program nightly that gathers up info
:>from all of the stats files and generates the appropriate
:>(sorted) HTML files as an optional way to see the topics
:>in the VL.
:>Usage should be one of these stats. Probably averge number
:>of downloads of the home page per day (take the weekly total
:>and divide by seven, so we don't have weekday/weekend problems).
:>Later, usage will probably have to be listed in terms of HOURS,
:>but we can deal with that when the time comes.
For info.cern.ch, we really need to process day per day...
Results are, from December 2 to December 6:
- for the top page Overview.html
40 000 to 45 000 requests on week days
20 000 to 25 000 requests on week-ends
- for its "summary" Overview2.html
1 500 to 2 500 requests per day
:>Other stats could be # of links and # of megabytes, perhaps. For
:>each stat, the cern program would generate a different variation
:>of the VL list of topics. Even fancier stats can be dreamed up, such
:>as the number-of-links-traversed (since the real usefulness of a
:>home page full of links is not who downloads it, but who finds
:>something worth jumping to from it).
:>The stats file should be really simple, so that someone can edit it by
:>hand, and so that different platformns can easily write programs
:>to generate it automatically. Maybe something like:
I don't mind wether we use a stat file, or comments inside the file.
But I remember we had a previous discussion on this:
Steve Waterbury says (Aug 18):
> It sounds good, but wouldn't it be better to formalize
> certain aspects of the structure of the WWW VL pages so
> that the information could be derived directly from them
> without creating another html document?
This isn't an issue for me: both solutions are easy to implement in perl.
An inclusion to the existing file would be:
<!-- Usage: 350 -->
<!-- Links: 765 -->
<!-- Cited: interactions, July 1994, p 10 -->