Re: location providers

On Wed, Jan 21, 2009 at 10:00 PM, Thomson, Martin
<Martin.Thomson@andrew.com> wrote:
> Two sources, one result, how?
>
> Alright.  Let's go with altitude.  I take each source and quantify its error.  The barometric altimeter is subject to pressure variations so it is off less than 200m, 95% of the time; I probably have to work this out ahead of time.  The GPS directly provides altitude uncertainty, which I scale to the same confidence.
>
> I then have two values:
>
>  BAlt = 300 +/- 200m
>  GPS = 400 +/- 63.4m
>
> As a device implementer, I can combine these using some sort of agreement algorithm (a+b/2 works fine, but there are definitely more sophisticated methods), OR I can pick the one that reports the smaller uncertainty.
>

Yes, I completely agree. Altitude isn't the only property that needs
to be treated this way. In Gears we already have a multitude of
location providers (multiple network-based providers, GPS) and we
already have algorithms that arbitrate between the various results and
return the ones with the least uncertainty back to JS. I understand
that it is difficult to compare altitude readings measured using
different datums and methods, but it is certainly possible, as Martin
pointed out. Unless other implementers strongly argue differently, I
think it is time to conclude this discussion and stick to the current
simple API design.

Thanks,
Andrei

Received on Thursday, 22 January 2009 15:19:46 UTC