1908182 - [meta] Improve Preference and Nimbus support in Android

Status: NEW

(Firefox for Android :: Experimentation and Telemetry, task)

Description

[Olivia Hall [:olivia]]

This meta is to improve Gecko preference and Nimbus support in GeckoView and its embedding applications.

Comment 1

[Olivia Hall [:olivia]]

See this document for proposal and connection between these bugs.

Comment 2

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug should belong to the 'Fenix::Experimentation and Telemetry' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 3

[Andrew Creskey [:acreskey]]

Mentioned as possible future work in the proposal, but just seconding that it would be valuable to have an Android equivalent of about:studies for experiment verification, user visibility of our work, etc.

Comment 4

[Andrew Creskey [:acreskey]]

Olivia -- it's exciting to see all of the progress on this important bug.

Would you say that we can now run gecko pref-flip experiments without all of the deep plumbing and geckoview api changes?

Comment 5

[Olivia Hall [:olivia]]

Would you say that we can now run gecko pref-flip experiments without all of the deep plumbing and geckoview api changes?

Essentially, yes, once a couple of things resolve! Everything is setup and we have a stable v1. If you wanted to test an experiment locally using nimbus-cli, it is now possible.

To be 100% production ready, we are cleaning up a couple of loose ends as we speak:

• We're needing to make changes to nimbus.fml.yaml parsing, which has proven difficult because of a lot of circular dependencies between Application Services and Fenix, but :beth is close to having that solved. (bug 2026043, bug 2027633).
• The error in Bug 2025031 is handled correctly when it happens, but I'm not quite happy that I'm seeing it at all. I'm checking up to see if I can improve that scenario.
• The team needs to sync one more time to make sure there isn't any important aspects we are missing. This flow inherently has a lot of edge cases.

Bug 2027008 will close when these dependencies resolve and it is fully considered production ready. I expect within 151.

If you or your team already has something in mind, we'd be glad to work with you! I'm specifically watching for the first experiment using this mechanism, in case we need to make any changes or discover any problems. It is a complicated workflow with a lot of back and forths across many different layers. (Technical documentation in D290481. Usage documentation will end up in Experimenter.)

Comment 6

[Andrew Creskey [:acreskey]]

If you or your team already has something in mind, we'd be glad to work with you! I'm specifically watching for the first experiment using this mechanism, in case we need to make any changes or discover any problems. It is a complicated workflow with a lot of back and forths across many different layers. (Technical documentation in D290481. Usage documentation will end up in Experimenter.)

Thank you Olivia, this sounds great.
And yes, we'd very much like to run a Fenix variant of this desktop experiment. We don't have dates yet, and it's not urgent. But we would very much like to be able to measure the impact when feasible.

Comment 7

[Olivia Hall [:olivia]]

Unlinking some tickets that lost relevance or aren't blocking the project or are more nice-to-haves. Designated follow-ups for v2 are at 2026348.