Investigate high accuracy mode alternatives

[anssiko]

The highAccuracy flag is a hint, and as a consequence, has known interop issues. Investigate whether this flag could be dropped from the modernized API. Are there any use cases that would support porting it over?

It seems a better alternative would be an API that addresses the following use case properly, especially for a one-shot geolocation request (via https://github.com/WICG/geolocation-sensor/issues/2):

return me a position as soon as the accuracy is within Z meters.

This suggest a web developer settable accuracy thresholds of some sort.

[yell0wd0g]

Re. part 1, UAs had quite the latitude in implementing the previous Geolocation API, resulting in a soup of sensors including WiFi and GPS, with a potential low accuracy IP mapping as a fallback. Developers tended to resent the lack of feedback as to what choices were being done without proper communication, so perhaps we could consider a list of supported/active providers of geolocation like Android seems to have.

Use case two reminds me a bit of Geofencing of which there was an API that seems to be put on ice, @mkruisselbrink might be able to provide some insights.

[martijnthe]

For our use cases it's desirable to be able to let the client of the GeolocationSensor specify "desired accuracy" (in meters), to enable the "sensor backend" to balance low power consumption and the needs of the sensor clients.

I think it's best to let the client express the accuracy in absolute terms (meters) rather than a relative concept like "high accuracy".

This is related to, but different from, https://github.com/WICG/geolocation-sensor/issues/13

[alexshalamov]

@anssiko If I remember correctly, we were experimenting with threshold parameter in the options that was meant to be used for such use cases. The specification already defines such concept, however, it is not controllable from the client side.

@pozdnyakov WDYT?

[pozdnyakov]

threshold could be provided either as a part of the generic SensorOptions construction parameters or specifically for the GeolocationSensor class. Note, that threshold parameter can only set the minimum value for threshold, the actual threshold might be bigger (depending on the actual device sensor and the underlying software platform)

[martijnthe]

Having both a threshold (meters) and desiredAccuracy (meters) in the constructor parameters object makes sense to me.

[anssiko]

This issue seems to be a duplicate of #13.

I submitted a proposal in https://github.com/WICG/geolocation-sensor/issues/13#issuecomment-366701610 and suggest we continue discussion in that issue.

[tomayac]

This issue seems to be a duplicate of #13.

…which in turn IMHO can be merged into #25. Are you OK with that?

[anssiko]

There seems not to be good use cases for highAccuracy bit, also implementation experience from Geolocation API demonstrated interop was hard. Better abstractions to explore as alternatives are accuracy and latency.