Cary:

More good news - thanks!

I don't know about other countries, but statistics based on server "hits"
are becoming pretty useless in the UK because of the very widespread use
of system caches. The whole of the UK national academic network (JANET)
has a multi-giga one which turns over once every few days and most larger
institutions have another big central cache as well.  Typically, the
disks on these are kept full at all times, with the oldest downloads being
automatically over-written as new pages or versions are used.

Consequently, if I for example download the ICOM Code of Ethics that would
be recorded as just one hit in Stockholm.  However a copy of that file
with its web address would be kept on our own University central cache for
perhaps 4 or 5 days in term time and possibility as long as 2 - 3 weeks
during the summer vacation when the traffic is drastically reduced, and
might be used by perhaps 30 students several times over during that
period.

In addition, calling that file up from our University involves JANET -
which in turn will keep a copy on the national cache. From this
potentially literally millions of users in a couple of hundred
institutions call down hundreds of thousands of web documents a day
without knowing whether they are getting this from the original source,
from the national cache or from an institutional cache. Similarly, most if
not all of the independent/commercial Internet Service Providers (ISPs) in
Britain each have their own central web caches.

One indication that this is happening are home page on-screen web
counters: I have seen the VLmp counter "stuck" at the same number for days
on end - presumably because I'm being served up with the same cache copy.

This is clearly a general problem in Britain, though I understand than in
some other European countries efforts have been made to outlaw, or at
least limit, third-party caching of this kind?

Two final points.

1.  In fast-moving areas users may be getting seriously out of date
information.

2.  If the UK situation is being repeated in other heavy use countries
then the real web use statistics for the more popular pages at least may
be several times greater than anyone is recording.



Patrick Boylan


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Change ICOM-L subscription options and search the archives at:
   http://home.ease.lsoft.com/archives/icom-l.html