tlug: Interesting statistic

["Stephen J. Turnbull" <turnbull@example.com>]

>>>>> "jdh" == John De Hoog <washi@example.com> writes:

    jdh> This statistic caught my eye today.

    jdh> (Assuming the results are meaningful, I wonder if this means

Before interpreting, I'd want to know who uses the Hitometer.  I'd bet
www.linux.org does not, for example.  In general, I would expect
"content providers" to be the most likely to use turn-key web sites,
including the "Hitometer", while the academic/research-institution-
based sites that I would imagine Linux users are relatively more
likely to visit are more likely to use homebrew log-analyzer scripts
or not care.

Also, the numbers added up to 100%.  Do all browsers announce their
OS?  Netscape and MSIE do, but how about Lynx, Arena, w3.el, etc?
httpget (or whatever it's called) and other background URL fetchers?
The latter are almost surely not from Windows.

What about firewalls and caches?  As I recall, one of the big
objections that providers had to caches was that this would distort
their marketing data.  Here, this works the wrong way as far as I can
tell: people behind corporate firewalls are more likely to use Windows
than the other OSes, and those firewalls are often enough set up to
keep them from browsing on company time (or the records that firewalls
often keep inhibit them), both tending to reduce the hit rate from
Windows machines.

On the other hand caches and mirrors _may_ be far more popular at
academic/research sites, reducing hit rates from both Linux and Mac
users.

Also, what is a "pageview"?  Is that necessarily a user-initiated GET
following a link (or hand-inputting an URL), or is it going to include
concomittent image-fetches (which would drastically lower the hit-
rates for lynx and w3.el users) and/or "push" fetches from Javascript
and Macromedia (both of which I suspect are less common in Linux
browsers)?  I would assume not---otherwise why not use cat httpd.log | 
wc?---but how close to eliminating bogus hits does Hitometer get?

Even if it is the best case (only user-initiated GETs are counted), I
would not be surprised to find that Linux users often follow somewhat
fewer links to get the information that they want and that their pages
average longer than those viewed by Windows users, reducing the number
of GETs.  (Based only on my own browsing experiences; Linux-related
sites tend to be shallow with long pages compared to corporate and
content-oriented sites.)  For download-and-saves, I imagine Linux
users are far more likely to use ftp than Windows users are, reducing
the number of fetches they do by browser.

    jdh> Linux users are really that small a proportion of the online

I suspect this sample is biased against Linux, but the theme of
overwhelming Windows dominance and small, even tiny, presence for
Linux-based browsing is certainly correct.

    jdh> community, or whether they are hiding their Linux identities,

Probably not, although several Linux browsers provide the ability to
spoof, so that they can trigger content meant for other browsers.

    jdh> or whether they just don't Web browse very much. I also

Linux users probably spend proportionately more time on email, news,
ftp, and even M-bone when "online" than do Windows users.  Or do you
define "online" purely in terms of "Web browsing"?

    jdh> expected Mac users to be more plentiful, and am surprised by
    jdh> the relatively high Win98 figure.)

To me, both of these suggest some, perhaps substantial, sample bias
toward sites catering to Windows activists in the Hitometer data.

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences        Tel/fax: +1 (298) 53-5091
---------------------------------------------------------------
Next Meeting: 10 October, 12:30 Tokyo Station Yaesu central gate
Featuring the IMASY Eng. Team on "IPv6 - The Next Generation IP"
Next Nomikai: 20 November, 19:30  Tengu TokyoEkiMae 03-3275-3691
---------------------------------------------------------------
Sponsor: PHT, makers of TurboLinux http://www.pht.co.jp