Conclusions and future work

Separation of concerns is also helpful for defining and understanding
programming language primitives. 

In earlier papers we expressed the semantics of a traversal strategy
through a set of paths in the class graph. This required a 
heavier definitional approach. In this paper we separate the issues:
we define the FIRST predicate basically at the
level of object graphs, only referring to paths in 
object graphs and using an existential quantifier over object graphs.
And only later, to compute FIRST, we refer to paths in the class graph.
This separation of basic semantics from efficiently
computing the semantics leads to a concise definition
of the meaning of a traversal specification basically only 
referring to object graphs.
The paper also systematically derives the efficient computation
of FIRST.
The conciseness is also helped by the observation that a relational formalism
is the right approach to express the non-standard path concept needed.

The paper achieves another separation of concerns:
In earlier works we defined the meaning of a 
traversal specification for an object graph to be  traversal history:
an ordered sequence of nodes and edges describing the visiting order.
In this paper, we separate the definition of the traversal history
into two parts: first we define the subgraph of the object graph
selected by the traversal specification (using FIRST and search)
and then we may use any
traversal approach we choose, e.g., depth-first, to define a traversal history.
This separation of traversal graph from traversal history
leads to a more concise and 
understandable definition of the concept.

The paper presents a graph-based crosscutting technology.
We have two graphs: a class graph and a strategy graph that crosscut
each other based on a map function. We have implemented
an efficient algorithm to compute the crosscut of the two graphs
\cite{strategies-tr:LP97} and we will explore further applications
and extensions of this approach to aspect-oriented software development.