John Mileham
John Mileham
Variant retirement is a confusing vestige of a time when we had ambition to analyze splits despite mid-test re-weightings of variant probabilities. At this point our best practice would be...
We hope to add support for mobile clients in the field who reference since-finished splits and provide a final disposition for those splits. In order to provide the correct answer...
Depends on #74. An unauthenticated endpoint to ask for the decided variant of any finished split. Because there was not always a decision field on splits, this should probably error...
To reduce handshake latency and reduce incidence of negotiation failures when networks aren’t 100% reliable
Now that authenticated web sessions are Identity driven, we can no longer fail back and have the client deterministically decide its variant for a split from its visitor ID of...
Delayed::Job and Clockwork contexts should manage an offline session for you, allowing session lifecycle management/caching of an arbitrary number of identity-driven TestTrack visitors.
Whatever solution we arrive at here should leave us with the ability to configure `Delayed::Job`'s `priority` and `queue` facilities, the support for which was still controversial in ActiveJob among rails-core...
The JS client is designed for client-side usage. Login doesn't need to be part of the JS client because a proper integration of TestTrack with an authenticated app will involve...
We switched to default snake-case analytics event names, and started sending `feature_gate_experienced` events in lieu of `split_assigned` events for feature gates (splits ending in `_enabled`), because they don't occur just...
(equivalent of https://github.com/Betterment/test_track_rails_client/issues/43) Per this comment on test_track server: https://github.com/Betterment/test_track/pull/69#issuecomment-380995245 Since gates (splits having a name with a suffix of `_enabled`) are not used for analysis, we can save storage...