[Foundation-l] Where do our readers come from? Q&A

[Ronald Beelaard]

I read all kind of confusions about funny correlations between language
versions and countries where visitors are coming from.

As I (privately) communicated with Erik, the following flaws are in the
current analysis:

* The country code AU is often used (by apnic in this case) as a placeholder
for ranges that are pre-reserved. For instance to allocate parts of that
very big range in bits and pieces to countries in the area (e.g. JP)
* Similarly Ripe is doing that for the country code EU (not to be confused
with the language code eu)

Other misinterpretations may occur because there are some conflicts between
country and language codes. An example of this is for instance SL (Sierra
Leone) and sl (Slovenian) and I guess UA (Ukraine) and uk (Ukrainian?) is a
similar case. But there are certainly more.
See also: http://meta.wikimedia.org/wiki/Language_codes/Conflicts, although
imo this list is not comprehensive.

Another cause of problems might be the fact that the assignments of IP
ranges continuously change. That happens on a small scale (e.g. re-assigning
a block of 65536 or much smaller), but also on a larger scale. The result is
that you can't fully trust a so-called geo-IP database (like MaxMind). I
don't know how quickly such a database is outdated, but have noticed major
shifts of ranges of more than 16 million addresses within half a year
(concerning the AU - JP confusion).
Structured lists do not exist, so the only way is continuously checking the
data in such a database via the Regional Internet Registries. That is a
complicated, but also a very time-consuming process.

So don't draw conclusions in the case of small countries and/or languages.

Rgds Ronald