Information Technology Reference
In-Depth Information
datastore that references a set of LUIDs cannot be translated (or identified as refer-
ring to the data on the server by a different name) by the server. SyncML [40] ,for
example, uses a Map operation to send the LUID of newly created data to the server.
This allows the server to update its mapping table with the new LUID, GUID asso-
ciation. However, this is not sufficient to address all possible problems. If the LUID
gets incorporated directly into application data, the synchronization process may not
know how to modify the application data, or even detect that it should be modified
during synchronization.
6.
Summary and Conclusion
In the past, device resource constraints such as CPU and memory precluded even
considering whether to execute business applications on disconnected devices. Now
that such resource constraints are disappearing, we are forced to determine whether
the algorithmic issues related to client synchronization of disconnected work pre-
clude disconnected business applications. Similarly, the feasibility of programming
models that facilitate the development of such applications becomes increasingly
important.
We discussed these two areas of concern in this chapter. We explained why busi-
ness applications are particularly hard to “project” to disconnected devices. We then
introduced two general approaches used to perform synchronization: data replication
and method replay .
We also discussed two programming models as applied to disconnected business
applications. The first, Enterprise JavaBeans, was originally designed for connected
environments. Due to its component-based design, we showed that it can be pro-
jected to disconnected environments as well. The second, Service DataObjects, is a
new programming model which provides unified data access to heterogeneous data
sources, unified support for both static and dynamic data APIs, and support for tools
and frameworks. Although SDOs are targeted at “support for disconnected program-
ming models,” we show that the SDO definition of “disconnected” is significantly
different from the definition that we consider in this chapter. The SDO programming
model is designed to eliminate lock-holding during user think time, and is shown not
to be suitable for environments where communications between client and server are
interrupted, in its current incarnation.
Finally, we analyzed two different projections of the EJB programming model
to disconnected environments. We compared data replication to method replay, dis-
cussed alternative disconnected programming models, and showed that a connected
programming model can be usefully projected to a disconnected environment.
Search WWH ::




Custom Search