On 7/10/14 1:40 AM, Hanno Schlichting wrote:
We have some public telemetry data about "geolocation accuracy” for our nightly 
users. That tells us that more than 80% of those users get a GeoIP based result 
today, with the accuracy being well over 20km. Nightly users with activated 
telemetry are a very small percentage of all users, but I think for this purpose the 
number should be a good overall indicator.

I thought telemetry was enabled by default in Nightly and Aurora channels (and soon Beta)?

Hanno: are you referring to the telemetry data about geolocation accuracy on telemetry.mozilla.org? What are the "Start" and "End" columns measuring?

http://telemetry.mozilla.org/#filter=nightly%2F33%2FGEOLOCATION_ACCURACY


I think that most desktop users still have stationary devices, and their home 
or office WiFi networks aren’t in any database. If we added them to our 
database, I fear we’d just end up with many unconnected islands of networks, 
but nothing to connect those. We’d need someone with a GPS capable device to 
capture at least one of the networks, to anchor it with a good position. But if 
we have someone stumbling in the neighborhood, than that person will also 
likely pick up most of the networks our desktop users have seen.

I think you are right if we have someone using MozStumbler to thoroughly scan a neighborhood, but we only have a few hundred active MozStumblers. When we add more aggressive stumbling (but still opt-in) to Fennec, we might have tens of thousands of people submitting many short scans. If we have a large network of dark APs from desktop stumblers, then GPS positions reported by Fennec users might transitively "light up" a larger area. Keven Everets gave an example of someone scanning a street with MozStumbler and transitively providing position data for dark APs on side streets. But I agree that we would still have many disconnected islands of APs.


chris
_______________________________________________
dev-geolocation mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-geolocation

Reply via email to