Many real-time, distributed applications maintain a shared “scene” or world state: acollection of entities and their attributes that must be updated and observed across multipleprocesses or services. Common examples include digital twins, simulation dashboards,monitoring visualizations, and online control systems. Na¨ıve designs frequently rely on ad-hocmessage formats and full-state broadcasts, which couple producers and consumers tightly,waste bandwidth, and make it difficult to reason about correctness when updates are delayed,duplicated, or reordered. This paper proposes Distributed Scene Update (DSU), a middlewarepattern and data model for partial, real-time state synchronization. DSU treats all applicationchanges as a stream of declarative scene updates rather than imperative commands. Eachupdate is a small, self-contained description of what should change in the global scene,together with identifiers and versioning metadata that enable merging, idempotence, andconflict resolution. We formalize the DSU model, provide merge semantics for applyingupdates to a canonical scene, and describe a transport-agnostic middleware architecture(e.g., compatible with publish/subscribe messaging or RPC). We then analyze correctnessproperties such as convergence under at-least-once delivery and discuss performance trade-offsin update granularity, batching, and indexing. Finally, we outline a reference implementationand an evaluation plan based on a real-time multi-service simulation dashboard. We illustratethe model using a running example involving a simulated vehicle entity. DSU aims to providea reusable abstraction: developers focus on what needs to change in the scene, while themiddleware ensures consistent, efficient propagation of those changes across a distributedsystem.
Robert Guy (Fri,) studied this question.