Hi there,
a simple download counter, given we can eliminate the proxy miscounts would be a good solution. All of the rest makes the system very intrusive or complex. All the 48h installation counting for example is supposed to prevent abuse, right? But if I want to abuse the system I can also always trivially fake the ‘48h have passed’ signal. IMHO not really worth it, given that it would require SOME kind of tracking system.
On the other hand, seeing downloads over time could give users a sense ofnan apps popularity.
@spaetz Thank you for the link to Fedora case: it tells, in particular, about importance of real metrics for sustainability of Fedora project…
Regarding “simple download counter” that you mentioned, my point is that such “simple metric”, like any others, which may be of developer’s interest, should be derived, i.e. calculated, based on raw data. And not calculated (counted) on the fly. This way in a case of any signs of wrong results it will be possible to filter out (or even delete) any errors/fake data and recalculate the “simple download counter” or any other report/metric. Without raw data the only way will be to reset the counter and thus lose all accumulated stats…
Of course it should be derived from raw data, we agree there. But as
others have pointed out, this is not trivial when web caches and proxies
come in the way, for instance are mirrors now provided by others which
makes download counts even more tricky to achieve.
As I wrote above, in order to protect user’s privacy AND have a way to link install / keeping / uninstall events of an application, I would assign some unique “application-download-id” to each installed application of every F-Droid client app instance (installed on a User’s device) and store that ID in the F-Droid client app instance. So the F-Droid client could inform a server about these events, tagged with the unique id. There will be no relation (stored on a server) between application-download-id (of each application) and a user, a device, or even between different applications, installed on the same device.
The “application-download-id” is linked to a concrete application version AND is unique (at least for this F-Droid repository, or maybe even globally for future scaling…). Thus each F-Droid Client instance will hold (and report to a server during installs and updates…) unique application-download-ids of applications, installed from F-Droid.
This way caches and proxies between an F-Droid Client and a server won’t influence the “evidences” (points in time, when each application-download-id is present on a device).
Mirrors and independent copies of F-Droid servers (repositories) will need to federate… Application packages are globally unique (and the same for the same application), version numbers are also the same, so no conceptual problems joining application data from different servers.
As I wrote in that conversation and repeat here:
That Conversation started with a developer’s screenshot, containing stats from Google Play. That stats obviously make the Developer proud of himself, his work, it generates positive feedback from his users… and that “Trending” stats don’t require “tracking users”, it is just clever anonymous algorithm.
Why fdroid cannot have similar or maybe even better statistics, so we could refer to it instead of to Google Play?
I was following a discussion on a Reddit Privacy sub where it was mentioned that F-Droid may introduce “trackers”. I almost posted that there was no need to worry about F-Droid doing this.
I don’t know, how exactly Google calculates this “performance indicator”, but earlier in this discussion I described, how I propose for F-Droid to collect information on application installs without tracking users and their devices.
I can see where this would be of use for developers. (I would also like to see a grading system or user reviews.) I’m all for it IF it can be implemented without users paying with their privacy. F-Droid is well respected. One taint, one perceived blemish, could do a lot of harm to a well-deserved reputation.
I agree that we should assess from User’s privacy point of view the concept and implementation of application installs data collection and its usage.
So far no blocking problems were found in my (draft) proposal here.
I think that when F-Droid team decides to implement such data collection, we will need to create separate publicly accessible document, describing what and how will be collected and used/accessed. We could openly discuss and improve that document so that all interested people/parties will see the expected result.
I already answered to you on that… but if you wish, copying my previous post here once again:
Regarding your quote of F-Droid privacy policy. “Anonymous installs tracking” as I described above is not a user or their device tracking, because no data, identifying a user or a device is stored. This is anonymous application installs tracking.
So words “We don’t track you, or your device. We don’t track what you install…” are still true.
The only phrase that needs amendment is “and it sends no additional identifying data when talking to our web server other than its version number”
to e.g. “and it sends no additional identifying data when talking to our web server other than its version number and identifiers of applications, installed from the repository”.
In my opinion, there is not much value for both users and developers from adding any sort of statistics to F-Droid. As a F-Droid user, I don’t care about ratings/downloads/etc when I install apps. It’s enough to read the description usually. As a developer, I don’t pay a lot of attention to downloads, because these numbers by usefulness are close to zero. Well, most part of developers don’t sell additional features here, so they don’t have to use marketing tools and methods, where stats come in handy. If developers want to use these stats, they may add open source analytics library in their apps with the ability to opt out from tracking, but they shouldn’t demand F-Droid to add tracking, because it is completely against the goals of the community.
@HenriDellal As you can see even from responses in this thread: people are different. Some of them see a value in F-Droid statistics, some don’t.
To what “community” are you referring to? Are people, who have different view, not a part of it?..
