Communication environments involving wireless networks challenge the
system software supporting applications that end-users are used to, but
which are primarily developed for wireline networks. The characteristics
of wireless networks are very different from those of wireline networks;
long latencies, highly variable delays and sudden disconnections (or
extremely long latencies) create problems that are not met in the
wireline networks. It is widely recognized that these problems must be
tackled on all levels of communications. Moreover, the characteristics
of available Quality-of-Service for nomadic users vary both in time and
in location. Therefore, it is insufficient to react to the current
situation. In the research project Monads we have addressed short term
predictability of available QoS so that applications can adjust their
behavior to the forthcoming QoS. The results presented in this paper
clearly indicate that intelligent agents can learn the key
characteristics and their temp-spatial variation of available QoS quite
quickly, so that reasonable predictions of available QoS in the near
feature are possible.
