A Temporal CSP is a network of time-points related by binary
constraints expressing maximal and minimal durations between
them. This formalism has proven to be useful in many different
research areas, but it has also proven to be limited in the sense that
it does not take into account the <em>contingent</em> nature of some
constraints: in many real applications, we cannot decide some
effective durations that are not under our control but will be
provided by the external world. This paper aims at proposing an
extension of TCSP that enables the expression of such constraints.
Here indeed the classical consistency property must be redefined in
terms of <em>controllability</em> of the network : in outline, a
network is controllable iff it is consistent in the classical sense in
any <em>situation</em> that could arise in the external world, i.e.
whatever the actual durations of the contingent intervals are. A more
in-depth analysis leads to the identification of three different
levels of controllability, the <em>Strong</em>, the <em>Weak</em> and the
<em>Dynamic</em> one. The paper will focus on the representation and
concept issues, with some hints on their relevance in different
domains. The reasoning aspects (complexity and algorithms) will only
be sketched in this preliminary report.