Conclusion for Shared Repositories

Most popular approach for maintaining shared state

Support a limited number of simultaneous users - on the order of 50 or 100

Each client's performance is closely linked to the performance of the other hosts

Approach is great for small-scale systems over LANs or engineering systems requiring highest levels of state consistency

Let's keep exploring the Local Cache idea, because, eventually...

Problem of the centralized repository

Too high overhead in communications and processor

Many applications don't require the "absolute consistency"

Example : Flight Simulator

In case of that error is limited and temporary

eScience