Assessing the geographic resolution of exhaustive tabulation for geolocating Internet hosts

Geolocation of Internet hosts relies mainly on exhaustive tabulation
techniques. Those techniques consist in building a database, that keeps the mapping
between IP blocks and a geographic location. Relying on a single location
for a whole IP block requires using a coarse enough geographic resolution. As
this geographic resolution is not made explicit in databases, we try in this paper
to better understand it by comparing the location estimates of databases with a
well-established active measurements-based geolocation technique.
We show that the geographic resolution of geolocation databases is far coarser
than the resolution provided by active measurements for individual IP addresses.
Given the lack of information in databases about the expected location error
within each IP block, one cannot havemuch confidence in the accuracy of their location
estimates. Geolocation databases should either provide information about
the expected accuracy of the location estimates within each block, or reveal information
about how their location estimates have been built, unless databases have
to be trusted blindly.