This was linked for web-platform-tests from the Intent to Ship on blink-dev. Is this API possible to write meaningful shared tests for, or is the heuristic entirely UA-defined so that no behavior can be ruled out and cause a test failure?
> Is this API possible to write meaningful shared tests for, or is the heuristic entirely UA-defined so that no behavior can be ruled out and cause a test failure?
I think there are some cases we can test, like what is reported when there is no connectivity, plus other edge cases. Otherwise, I do think we want to keep the behavior as UA-defined and allow for flexibility.. Tarun, WDYT?
Re#13: yes, that sounds about right. The estimation algorithm is UA dependent, but we can check for certain high level conditions (e.g., if the device is connected to a network, then the estimates should be in a valid range).
Comment 1 Deleted