Event-driven programming is a popular paradigm for programming sensor nodes. 
It is based on the specification of actions (also known as event handlers) 
which are triggered by the occurrence of events. While this approach is both 
simple and efficient, it suffers from two important limitations. Firstly, the 
association of events to actions is static - there is no explicit support for 
adopting this association depending on the program state. Secondly, a program 
is split up into many distinct actions without explicit support for sharing 
information among these, except through global variables. These limitations 
often lead to issues with code modularity, complexity, and correctness. To 
tackle these issues we propose OSM, a programming model and language for 
sensor nodes based on finite state machines. By extending the event paradigm 
with states and transitions between them, state-based programming supports a 
dynamic association of events to actions. For removing the second limitation, 
OSM introduces state attributes that allow sharing of information among 
actions. They can be considered local variables of a state with support for 
automatic memory management. OSM specifications can be compiled into 
sequential C code that requires only minimal runtime support, resulting in 
efficient and compact systems.