[feature request]: add sharing abilities

[herrdeh]

I like this project and immediately installed it to my phone, it already greatly improved location accuracy. IMHO, the "self-learning" approach is the highway to heaven. As well I like taking WiFi and GSM together, it no longer makes sense to have distinguished WiFi and GSM backends. Therefore - thank you very much for your work.

But (sorry for the "but", but inavoidable): Why should the achieved results not be shared amongst the users - in the case they want to? And why should every user build his own database from scratch? (Nobody will do anyway, but use several location backends at the same time, not knowing how they will interact, as I do now - now I have 3 backends, which may not be the optimal solution)

So - these are my thoughts on enhancements (certainly nothing new to you):

• OPTION: provide a "to-begin-with" database, containing results from the openwlanmap.org project or some other open location project. Database size could be limited by just downloading certain postcode areas, landline phone region codes (my favourite) or just countries.
• OPTION: when phone is operating and GPS is active, check whether the given results can be improved in terms of accuracy -> if yes, mark these database entries and collect them for occasional transfer to the central server (when phone is connected to WiFi and AC power)
• OPTION: occasionally check the differences between phone database and central database and find out whether the phone can achieve improved data
• OPTION: offer an option to call GPS in regions with known-bad-coverage, with user-adjustable level of required minimum % of battery charge for that. (I'd be happy to do some GPS logging to improve the central database, but not if my phone is short of power...)

Sorry, this requires quite some controls and switches in the app. But as we're not iphone users, we should be able to handle them.

Chers, Wolf

[n76]

I really want to keep Déjá Vu simple for users. My hope is that with projects like LineageOS with microG we can start expanding the user base to less technically minded users.

Perhaps the way to meet you desires and/or those expressed in the import/export issue, would be:

• Add a service interface to the database access object. This would allow another app to access Déjá Vu’s database.
• Create another app (or set of apps) that is aware of the above interface to import/export data either for local backup or to share with others through a Internet based server.

This would allow power users or people who want to share data the ability to do so but still leave the Déjá Vu with no need for a confusing UI.

[gdt]

I am moderately onboard with the simple notion. But, I very strongly believe that a key property is privacy, and the current setup achieves that (modulo having a location history stored on the phone itself). Sending data to services like radiocells is sometimes ok and sometimes not, and that's difficult. I don't think it's going to be easy to encode the right logic into the app. This is exactly why, despite having some of the same thoughts as @herrdeh, I asked about export/import, because then I could experiment with whatever processing is needed, without having to adjust the basic DejaVu behavior.

For my own use, I'm ambivalent between a second app to export/import and options in the (missing) config screen.

The seed database idea makes sense, but I think you get that by using the openbmap backend also.

It would also be interesting to have another backend, that just turns on GPS, if battery run rules allow and some configurable filters find network data not good enough. But that seems out of scope of DejaVu.

[herrdeh]

Allright, a split into two parts may be a good idea.

But I'm quite certain - the dummy user will not like a location service, which, when installed first, will not provide any location service and then slowly improve over many months. And which will never deliver location services at places the user has never been before. So I still belive that a to-begin-with database has to be delivered at the very beginning.

Maybe making it easy is more about creating a smart setup wizard instead of leaving out features and controls?

[gdt]

Discussing user expectations is hard, but a key point of the location provider achitecture is that multiple providers can be installed at once. So there is no reason to install only DejaVu. In addition to the privacy property, another feature of DejaVu is that storage is only used for places your phone has been. Adding 1 GB of openbmap data would break that.

Actually, I find that DejaVu becomes useful very quickly, if one drives with GPS on. To me, a classic case is navigating someplace new with OsmAnd and then going inside; localwifi always did that well and this weekend DejaVu had good locations inside the building as well.

So perhaps it would be useful if you explained how you came to believe that people should install only one provider, rather than the approach (which is obvious to me, having years of UnifedNLP context) of installing both a downloaded-database provider and DejaVu together. Then we could add a hint to the documentation to explain this.

[herrdeh]

Thank you for your thoughts. Having been with UnifiedNlp now for years as well, to me the hardest part still is to decide between location providers. Actually, I only recently noticed that those I used do NOT have the self-improving feature.

To me, Deja-vue has got the potential to be "the one and only.", THE standard solution, with which most users can be happy out of the box. So all the other ones can be left to experienced users with special interests.

Core features to make it easy are:

• only one location provider necessary, because it deals as well with GSM AND WiFi (to me a really important one)
• self-improving without user interaction

So it just needs a to-begin-with database and it's ready to make quite a lot of people happy. Adding the sharing features wojuld be bonus for the more ambitious ones.

[herrdeh]

I created a very simple tasker profile (two working together, to be precise) in order to speed up the build-up-the-database process. It does not more than

• check whether I'm in my home zone & check whether I've more than xy% of battery charge & check whether the phone is shaked
• => in that case turn on gps every 20 sec and do a GPS location

Unluckily, I cannot save the already found good localisazions, nevertheless I already get good results at my daily routes. Hopefully, these tasks will be done in real software some day... (-;

Hints for improvement are appreciated.

tasker-deja-vue.zip