Hamina Wireless
Aalto University · Computer Science · 2026
The Limits of Generalized Sync: A Taxonomy of Architectures, Trade-offs, and Decision Factors
My thesis investigates how far client-side data synchronization can be generalized across web applications.
I combined a literature review, 69 practitioner sources, 13 interviews, and a production case study. The result is a taxonomy of 14 sync engine instances across 13 projects and a decision framework for adopting one.
Sync engines form four architectural clusters. The online/offline boundary sets the limit of generalization: read paths generalize well, while offline writes remain tied to authorization, conflict resolution, and application-specific business logic. Real-time B2B tools are the broadest fit. Transactional and data-intensive applications fit poorly.
Main finding
The sync boundary is the offline write path
Always online
- Read path
Generalizes
Reusable subscriptions, IVM, and shape filters.
- Write path
Generalizes
Engine can own replay and reconciliation.
Offline capable
- Read path
Generalizes
Local replica serves reads from disk.
- Write path
Resists
Authorization, conflict resolution, and business logic stay application-specific.
