On 04.09.2014, at 20:18 , Felix Baumann <[email protected]> wrote: > I thought MLS had no interest in pure cell tower locations? > > I once brought up a website that people are using to publish/collect exact > cell tower locations to you and you said that MLS isn't interested in them... > > Pls clarify your opinion on this ;) > > Or weren't you interested in the idea of making a separate database for that?
I changed my mind a bit on this. For this year or as long as we are still focussed on getting reach/quantity our main goal is to be better than GeoIP and usually our clients only send a single cell network in their requests. In those cases it’s more helpful to know where people typically are inside each cell area and that’s a good enough accuracy. For next year or when we get to look a bit more on the quality side, we probably want to look at trilateration a bit more. Today we do it, but based on the N points where most people are inside each cell. The average of those might be worse, than taking N true antenna positions and averaging them. Or at least it should get easier to reason about the accuracy of those results. I haven’t thought about this in detail. But I think for doing single cell lookups our current approach works better. And for lookups based on multiple cells, having better antenna position estimates is probably advantageous. So in the future we might want to maintain position estimates for both. Hanno _______________________________________________ dev-geolocation mailing list [email protected] https://lists.mozilla.org/listinfo/dev-geolocation
