Disclosure first, I wrote this, and I'll keep it short since the rule here is justified or not at all.

It measures per-app mobile data for your billing cycle rather than the calendar month, splits foreground from background, and blocks apps per network so something can be dead on mobile and fine on Wi-Fi.

Why it's relevant here rather than just another utility. It has no INTERNET permission in the manifest, so it cannot phone home at all, and it has no Play Services dependency. The only library it pulls is androidx.core-ktx and there is no com/google/android/play anywhere in the release dex. It reads NetworkStatsManager and uses VpnService in package-list mode with no packet forwarding, both plain AOSP, so nothing about it needs Google on the device.

The one part that sounds wrong for this community is the protected-apps list, so I'll explain it rather than let someone find it. There's a fixed list of push carriers the app refuses to block, because blocking the thing that carries push silences every notification on the phone and people do that to themselves without realising. It includes Play Services, but also Samsung, Xiaomi, Huawei, Honor, OPPO, vivo and Meizu push, since those matter more in the markets this is aimed at. Matching is by package name and anything not installed is skipped, so on a degoogled phone with no Play Services there is simply nothing there to protect and you lose nothing. The Play Store itself is deliberately not protected, you can block it, which on my test phone was the single biggest consumer at 444 MB a cycle of background auto-updates.

The honest gap: right now it's only distributed through Play. The F-Droid request is open at fdroid/rfp#4520 and has been waiting since 8 October, which from what I can see is normal queue time rather than anything wrong. Until then it builds from a clean clone if you'd rather do that.

GPL-3.0, free, no ads, no purchases, no accounts, no analytics. Android 7 and up.

https://github.com/anubisproductions/data-saver-usage-blocker

you are viewing a single comment's thread
view the rest of the comments
[–] 2 points 15 hours ago*

I'm not sure what the problem is with using LLMs on open source code. Everyone can see what Claude did, check it, and edit if it's bad. It seems a good way to stand up a project quickly, so people can see the idea in action, then help out.

  • source
  • parent