We consider linear equality systems with crisp variables and fuzzy number parameters
of symmetric trapezoidal form. It is shown that (i) the fuzzy solution (defined by
R. Bellman and L. A. Zadeh [Management Sci. 17 (1970/71), B141--B164]) of these systems
is stable under small variations of centres of fuzzy parameters, and (ii) any maximizing
solution can be regarded as an approximate solution of a certain generally ill-posed
classical linear equality system.