GmsCore
GmsCore copied to clipboard
microg
Merge LocalNLPBackend into MicroG
Due to the latest change, it seems better to merge LocalNLPBackend into MicroG, cf. https://github.com/Helium314/Local-NLP-Backend/issues/18
LocalNLPBackend seems to be the best NLP backend to maintain a local database for network locations.
Local location learning added via https://github.com/microg/GmsCore/commit/16e37cbad87749c853394547c02a7dd7df3516e3
Pleased to see this development - however, one question (acknowledging, that this may still be 'work in progress'): In the above mentioned commit, the "Mozilla Location Service" functionality is made invisible (and I assume also deactivated?), if no "API Key" has been defined in the build properties. What is the intention/necessity of this? Will there be follow-on commits?
Custom builds without API key never could use Mozilla Location Service (the service doesn't allow access without such API key). The change just hides the setting in those cases, as users otherwise might be confused.
Local location learning added via 16e37cb
It might be nice to add some other LocalNLP features.
- Import and Export: for easy migration between devices and from the original LocalNLPBackend
- Kalman filter: for bad GPS
- Bad emitter discard: to drop some Wifi AP which is movable, such as Phone Hotspot, Wifi on car...
- Kalman filter: for bad GPS
- Bad emitter discard: to drop some Wifi AP which is movable, such as Phone Hotspot, Wifi on car...
The code already:
- Filters out GPS data with lower reported accuracy than 25m
- Ignores Wifi AP that seemingly are reachable in an area of more than 200m within 14 days
- Ignores cells that seemingly are reachable in an area of more than 10km within 14 days
- Ignores Wifi AP which by SSID are assumed to be movable (based on the lists at MovingWifiHelper.kt:25-66)
Not perfect for now, but I'd say it's good enough.
- Import and Export: for easy migration between devices and from the original LocalNLPBackend
Indeed a good idea