Recent advances in wireless communications and the availability of more 
powerful and comfortable portable computer and communicator devices have 
made mobile computing a reality. The population of mobile users (clients) 
continue to grow as well as the number of multidatabase systems (servers) 
offering services that allow mobile clients access and share data. 
Transaction processing which is an important component of a database system 
has faced new challenges to accommodate the limitations of a mobile 
computing environment. Disconnection and low bandwidth which are 
characteristics of mobile computing have suggested that mobile computers 
should take more responsibility in the management of mobile transactions.
In this talk I present a new protocol for transaction processing in mobile 
environments that aims to provide more local autonomy to mobile clients. In 
the proposed scheme, mobile computers can locally commit read-only 
transactions and early abort transactions that cannot be committed. 
Serializability is the criteria used to ensure data consistency and data 
caching is adopted in order to increase local data availability. The 
protocol combines two approaches for checking serializability of concurrent 
executions: a pessimistic approach applied at the mobile clients and an 
optimistic approach applied at the server. A pessimistic approach is 
applied at the mobile clients to control the execution of local 
transactions each time an object is requested from the server. This process 
relies on the partial view of the serialization graph provided by the 
server with the requested object. The partial view is used to locally 
commit read-only transactions and to early detect some non-serializable 
situations involving updating transactions. Updating transactions that 
reach the end of their execution are sent to the server to be certified. 
Unlike in related work in mobile transactions that use dependency graph 
information to check serializability, the proposed protocol does not 
require each mobile client to build the serialization graph which can 
generate overhead in bandwidth utilization and mobile clients resources 
utilization. Rather, a mobile client maintains information only about 
dependencies between objects cached at the mobile client. This information 
is not delivered by the server as a separate message but is incrementally 
updated at a mobile client each time an object is requested from the 
server. Another feature of the proposed algorithm is that the completeness 
of the dependency information sent to the mobile client is a parameter that 
can be adjusted to take into account differences in workloads.