Who Owns the State? React, Redux, and Legacy Models in One Application

Who owns the state? React, Redux, and Backbone — a shared data block surrounded by interface cards.

The list shows the new name, reopening the form brings back the old one, and an API response overwrites the text the user has just entered. It is tempting to treat this as a synchronisation bug. In an application combining Backbone, React, and Redux, start with a different question: which part of the system has the authority to decide the current value?

One entity can have more than one meaningful value

A saved product name and the name being entered into a form are different pieces of information. One describes accepted data; the other is the user’s working proposal. They can legitimately differ for a while. Trouble starts when the application cannot distinguish an intentional draft from an outdated copy.

The selected product ID, whether a panel is expanded, and the save status do not necessarily belong in the same place either. Instead of asking where to keep application state, list the information, its consumers, and its lifetime.

Define ownership for one feature

Consider a product editor being migrated to React. During the transition, saved data might remain in a Backbone model. The React form owns the draft field values. The existing navigation mechanism owns the selected product. The layer performing the request manages the save status and exposes it to the view. This is one possible arrangement, not a universal blueprint.

For each piece of information, answer four questions: who changes it, who reads it, when does it reset, and how does it respond to server updates? Recording those answers can be more useful than adding another abstraction over the store.

Let the adapter carry data and intent

A component does not need to know Backbone events or how the legacy model persists changes. An adapter can pass a plain data object and a save callback. In the other direction, the form reports intent: save this set of values. It does not modify the model from several places while rendering fields.

The flow becomes explicit: model → adapter → props → form, followed by save intent → adapter → persistence → confirmed state. The adapter must also detach its subscriptions when the view is removed. Keep the coexistence of technologies inside an identifiable boundary.

Do not reset an edit on every refresh

Suppose the user has changed a name but has not saved it. A background update arrives for the same product. Unconditionally copying new props into the fields discards their work. Ignoring incoming data altogether can hide a conflict. State management alone cannot make this product decision.

Possible policies include refreshing only a pristine form, notifying the user about a newer version, or deliberately merging selected fields. Define switching to another product, a failed save, and a successful save separately. If the API supports versioning, use it to detect conflicts; the frontend alone cannot guarantee that one user will not overwrite another user’s change.

Give Redux a specific job

The Redux documentation allows local component state and recommends choosing a location according to how data is used. Introducing Redux during a migration does not justify moving every field into it. Shared workflow state may need broader ownership; whether a single tooltip is open usually has a much smaller scope.

Likewise, React describes lifting shared state to a common parent. That coordinates components; it does not require putting every piece of data at the top of the application. Choose according to responsibility and the consumers of that information.

Check whether the ownership model holds

Test external data updates, unsaved edits, save failures, switching products, and reopening the view. For each scenario, you should be able to explain which value takes precedence and why. If the answer is “whichever listener runs last”, the boundary still needs work.

Further reading: React: sharing state between components and Redux FAQ: organising state.