A typical digital repository attempts to disseminate the most up-to-date information it has. The motion picture, on the other hand, only disseminate the timed snapshots it takes from the fast changing states, and let the clients to reconstruct the dynamics from these static snapshots. Like all data driven dynamic web applications, this model has a scalability bottleneck while this fares much better under heavy loads. We therefore proposed a repository model and data replication algorithm that’s more like the motion picture. These are the key words in this new approach: global time, timed snapshots, web caching, lazy, master-slave style replication control, client-side database, etc… We can prove this model has the desirable property of one-copy serializability, meaning when the system is replicated to web scale, it still maintains a single copy illusion to the clients. Although an obvious requirement, this is surprisingly hard to achieve in today’s inter-linked bigdata environment, and is not promised by the popular NoSQL paradigm.