[WikiEN-l] Wikipedia growth since February 2004 greatly slowed

[Richard Gallagher]

On 15 Jul 2004 at 8:09, Jimmy (Jimbo) Wales wrote:

> Pete, I think you're not reading the evidence correctly, but I think
> that we need to make sure we get this right.  Forecasting the future
> is tricky enough, and we can only do it well if we understand the
> past.
> 
> So I invite real scrutiny of all these numbers, because I really need
> to know the right answer.
> 
> > 1) The Webalizer pages: (http://wikimedia.org/stats/en.wikipedia.org/)
> 
> Look at pageviews per day... 
> February 2004 3.01 million
> June 2004, 5.28 million
> 
> For the intervening months, there are significant problems with the
> averages due to some missing data.  (For example, several days in
> April are in the dataset and part of the denominator but clearly have
> missing data, for example showing 94 pageviews in a day.)
[snip]
> Notice in particular the dramatic spike in June which is continuing
> unabated into July.  This spike, I believe, corresponds with the
> acquisition of new servers.

I'm with Pete in the view that Wikipedia overall is growing very slowly at the moment 
and has been since February/March this year.  The big jump in hits seems to have 
occurred in the last week (week 22) of May 2004.  See the monthly and yearly charts at
http://wikimedia.org/stats/live/org.wikimedia.all.squid.requests-hits.html

Now this big jump coincides with the upgrade to monobook.  Looking at the chart you 
will see the traffic ramps up over the space of about a week or so, this corresponds with 
the piecemeal fashion that the different languages were upgraded to monobook in the 
last day on May.

Before monobook each page load without graphics caused four hits - the page itself, 
the logo graphic, a style sheet and a javascript file.  With monobook this number has 
increased the exact number depends on the browser configuration.  But will include 
several stylesheets, one or two javascript files, the logo image, the background image, 
the user icon, the list item icon, the gfdl and media wiki logos.

Just as a test loading the main page on English Wikipedia under Internet Explorer 6.0 
results in 24 hits to the server.   Now of course some client side caching will occur so it 
will not a have a massive effect on the stats, but I still believe it would be big enough to 
account for the 2 to 3 times increase that happened.

I believe the actual number of visits to Wikipedia since about March has stayed the 
same or grown very little.  Since June both the Webalizer and cache stats are very flat 
except for the standard weekly cycle (low weekend, high mid week).

As for the Webalizer "pages" statistic, I believe that some non-pages are incorrectly 
being counted as pages.  But without analysing Webalizers configuration and the log 
files themselves I can't say for sure what exactly is happening.

Richard