The paper describes the experience gained in the development and maintenanc
e of the data acquisition system for the RFX nuclear fusion experiment. The
design of the system core started in 1987 and the system was ready at the
beginning of the machine operation in late 1991. Since the early design sta
ge it was clear that a careful definition of the system architecture was cr
ucial for allowing the system to cope with the expected improvement in the
hardware performance. Not all the initial design choices proved successful
for this purpose. In the paper the (positive or negative) consequences of t
he main design choices of the system are reported. (C) 1999 Elsevier Scienc
e S.A. All rights reserved.