What is it?

In working through T327979: Enable persistent fixed width setting for anonymous users, it became apparent that we need a general-purpose solution for storing client-side user preferences that impact page layout and other fundamental pre-render activities. Such a solution could be used for persisting the width persistence itself, but could also be used to store other upcoming layout-related preferences.

This feature generally has the following requirements:

• Happens early enough to avoid page flashes
• Persists between refreshes and ideally sessions
• Works for anonymous users
• Avoids splitting the cache
• Takes into account other concerns of Performance and SRE from the outset
• Is generalizable / can support multiple different kinds of use cases, beyond simple binary storage

End-User Impact

• (+) Ability to persist preferences that impact user between refreshes and sessions
• (+) Ability to persist layout-related preferences for anonymous users
• (+) Prepares us for other potential upcoming use cases for this feature
• (-) May take longer to implement
• (-) Requires coordinating between multiple teams

Engineering Impact

• (+) Ability to store future preferences in this way
• (+) Enables incorporating performance and SRE concerns into implementation

Design Impact

• (+) Enables the persistence of future user-focused, layout-related design concerns
• (-) Depending on the solution, may be constrained to specific kinds of preferences or specific parameters around this storage

Other Impact

• This will serve as a pilot for how these teams can work together to solve cross-functional and cross-team concerns in a way that is sustainable for the longer term.

What does it block?

• We have potential upcoming feature work (READ: yet to be prioritized, not necessarily on our roadmap) that needs this ability, including (potentially but not necessarily) other desktop improvements-related layout change persistence, dark mode, and font size preferences.

What does done look like?

• We have identified an end-state solution that Performance, SRE, and Web all sign off on
• We have iteratively worked toward this solution by shipping incremental progress toward three client-side preferences
• We have made subtasks for the identified use cases, and those tasks have been completed.

Phases of Work and Owners / Timeline

• Phase 1: SRE+Perf+Web meet together to discuss roadmap and timelines, and identify goal end state. This may spin out necessary related spikes.
• Phase 2: Hardening of the width persistence solution (https://phabricator.wikimedia.org/T327979) and/or other Desktop Improvements-related work.
• Phase 3: We support a use case identified to be one month out - iterating toward our ideal solution but not necessarily satisfying all criteria for this
• Phase 4: We support a third use case with a solution that represents agreed-upon best practices.

Cross-Team Dependencies

• Performance - participation in working group and related conversations, input into performance-related considerations
• SRE / Data Persistence - participation in working group and related conversations, input into SRE-related considerations
• SRE / Traffic - participation in working group and related conversations, input into SRE-related considerations
• Architecture - Participation in meta-discussions about how to approach cross-team collaboration and input

Existing Artifacts

• See solution implemented here for how we initially solved this using a client-side cookie.

Open Questions

• What are the limitations and pitfalls of the client-side cookie-based approach?
  • The hope is we can discuss and identify these and arrive at a means of addressing
• What metrics should we consult related to performance, and how can we work to optimize these together?
  • The hope is we can arrive at specific numbers and means of measuring them.
• How can we approach working across teams to balance the concerns of each group without a full Technical Decision Forum process?
  • The hope is we can pilot an abbreviated process for moving toward more effective architecture while consulting the relevant parties
• How can we deliver on necessary Product deliverables while also working toward better longer-term architecture in this space?
  • The hope is that by taking a phased approach we can increasingly improve the quality of this solution in the face of real-world needs.