A goal can be categorized by the type of requirements
it derives from the the agents involved. For each
such goal category, specific obstacle categories may be
defined, e.g., starvation obstacles for satisfaction goals,
hazard obstacles for safety goals, misinformation and forgetting
obstacles for informing goals, threat obstacles for
security goals, etc. Knowledge of the category of a goal
may drive a search for obstacles in the corresponding obstacle
category, Thus, vLL too have observed the similarity
of methods for dealing with safety and security, and
with other problems as well.
The key idea of their paper is that obstacle
identification needs to be done as early as possible in the
system lifecycle, and as early as possible during requirements
analysis, that is, at the time that goals are being
identified. The earlier such identification is done, the
more freedom is obtained in dealing with the obstacles, at
best by changing goals so that the obstacle does not even
happen. Clearly, goals here correspond to Leveson et al’s
intents, obstacles correspond to hazards, and derived
objects and operations correspond to the blackbox model.
Indeed, vLL confirm this correspondence when they suggest
that goals might provide a precise entry point for analysis, e.g., constructing the fault tree starting from
negated goals. Clearly, the steps to formally derive obstacles
from goals are a formal realization of blackbox
modeling and hazard analysis, which are normally done
manually, albeit systematically, and with tools when
applicable. In any cases, there is no escaping deriving
more details than are normally considered requirements