[LINK] Introducing geoloc

[Jon Seymour]

On Mon, Jan 17, 2011 at 8:33 PM, Paul Brooks
<pbrooks-link at layer10.com.au> wrote:
> On 14/01/2011 1:59 PM, Jon Seymour wrote:

> "at their usual point of presence" is a bit of a stretch. Personally, every time I've
> enabled my android hotspot has been well away from my usual location, due to their
> being no other wifi network available. This feature is more likely to muddy the
> SSID+MAC-to-location database, as all these 'hotspots' become indexed at such a wide
> variety of places that it will resemble the song "I've been everywhere, man". Trying
> to determine a location on the basis that one of these is within range is likely to
> become extremely misleading.
>

It might have been more accurate to state: at their usual point of
WiFi presence.

In my own case, it pinned me at my work location, and then, when I was
on holidays at the cafe down the road from where I live.

Here are two maps which show the movements of two different types of
WiFi access point (according to Google) over a ~15 day period.

The first is for a type of pocket wifi device of the type I own:

   http://keepingthebastardshonest.net/geoloc/db/maps/change/

[ mac addresses derived by replacing the last 2 digits of my MAC
address with alternatives ]

For reference, the movement of my device is from Surry Hills to Pyrmont.

The second is for a type of base station offered by an Australian ISP:

   http://keepingthebastardshonest.net/geoloc/db/maps/bobbs-change/

[ mac addresses derived from observation while wandering around the
city (and also Melbourne), and restricting to a particular OUI prefix
The movements towards the cluster indicate that the devices were
present in the clustered location when I first observed them, but the
location was only updated in the last two weeks. ]

I was surprised that Google had updated the devices in a relatively
short period of time. For example, on the first map, there
is a WiFi device in Dampier, WA whose location changed slightly over
that period. I haven't worked out how this happened, but it
strikes me unlikely that Google happened to have street view car way
out in Dampier in the last two weeks. If it wasn't a Google street
view car that provided this location adjustment, one wonders on what
basis the location was updated.

I switched on an old access point recently and discovered that Google
has discovered it in the last 2 weeks.

That said, I have some counter examples: a friend has some access
points at his house in Wahroonga that have been there for months, yet
Google still lists them as being in Pymble and Melbourne, where they
had been previously.

It is possible that the reason my own devices and devices near me have
their locations updated relatively frequently is that I live not so
far from a Google office and so are more likely to encounter a street
view vehicle than the average person.

I am currently running a little experiment to get a better handle on
how quickly WiFi devices in my neighbourhood are discovered by Google
and how often their positions are refreshed.

jon.