Services as a Paradigm of Computation - Formal Methods and Hybrid Real-Time Systems - page 523

more expressive than shared variables is the technique of remote procedure calls,
which is a fundamental principle of middleware systems, in particular CORBA.
Misra and Cook in their ORC language recently generalized this principle, re-
placing the call of procedures by the call of services [9].

Interacting and communicating systems : Peter Wegner's contributions of the late
1990ies ([10,11]) boosted the awareness of a greater public, that interaction and
communication was indeed a decisive argument to search for a new paradigm of
computation. Many authors took up his arguments (as, e.g. in the volume [12]).
Some authors extend the classical models, in particular Turing Machines (exam-
ples include [13] and [14]).
We follow this line in the sequel, too. To this end, we discuss some elementary
aspects of the new paradigm in the next section, and pose some fundamental
questions.

1.3 Aspects of the New Paradigm

The above described aspects models and representation techniques for informa-
tion processing systems share a couple of aspects. Here, we focus just two of
them.
Firstly, in the classical setting, non-termination of a computation denotes fail-
ure, as no output is generated in this case. Two different non-terminating com-
putations cannot and need not to be distinguished in any respect. In contrast,
in the new paradigm, a computation is in general not envisaged to terminate.
Infinite computations are of utmost interest. Two different infinite computations
in general very well exhibit different input/output behaviour. Interaction of ser-
vices is usually split into finite slices: An instance of interaction is intended to
terminate in a “reasonable” state. But a service is assumed “always on”, capable
to engage in interaction ad infinitum.
The second consequence of the new paradigm is related to the composition of
systems. The classical setting offers sequential composition A ; B of two systems A
and B as the only choice: A 's output is B 's input. In contrast, the new paradigm
permits composed systems A and B to exchange data at any time during a
computation.
Though we have identified only two aspects of the new paradigm, it is obvious
that fundamentally new problems arise that cannot be identified, let alone be
solved, in the framework of classical system models. Typical problems of systems
that follow the new paradigm include:

- What kind of properties are important for such systems?
- Is there a canonical notion of equivalence for such systems?
- Can any two such systems be composed, at least syntactically, resulting in
a (may be, futile) system?
- What, precisely, is refinement and abstraction, and which properties should
refinement and abstraction preserve?
- How can the effect of such systems be abstractly described (in analogy to a
function f in the classical case)?