Optimistic data replication is an important enablingtechnology for high availability and scalability, but thedisadvantage is that it can only guarantee eventual consistency. Inconsequence, apps might observe stale data and write conflicts. Inpractice, optimistic data replication is usually part of databasesystems that implement detection and resolution of conflicts withpurely syntactic techniques. On the downside, this can lead toconflicts that are impossible to detect and resolve correctly byusing the database framework only, ultimately resulting inconcurrency anomalies. In order to avoid concurrency anomalies, developers are forced to use error-prone workarounds, therebyincreasing application complexity significantly. An alternative tousing existing database replication would be to implement acustom solution tailored to the application semantics that couldminimize the occurrence of write conflicts and concurrencyanomalies. However, the realization of custom solutions driven byapplication semantics is very difficult in practice, due to thecomplexity of concurrency control in distributed systems. In thisposition paper we demonstrate practical problems related to stateof-the-practice optimistic data replication and motivate the needfor semantics-driven solutions. Our research aims at simplifyingthe development of custom solutions by delivering a frameworkfor semantics-driven optimistic data replication. This frameworkconsists of a technical part and accompanying concepts for thedesign of extended domain models optimized for optimistic datareplication and conflict minimization.