An endless amount of future work can be suggested, but to name a few:
The environment used in this thesis has been very simple and could easily
be broadened or generalized, e.g., multiple servers, peer-to-peer
communications, etc. There are many open roads.
I think that it will be possible to have greater (better) utilization of
TACO's QoS facilities. Maybe it would be a good idea to let applications
specify both time bound and quality of service parameters. Better integration
and TACO is a certainly worth looking at.
Other consistency measures than time might be considered; the number of
updates to a file, the difference (in terms of contents) between a replica
and the primary copy, .... Notions of fidelity as in the
Odyssey  system might also be included in the model.
An implementation of the suggested transactional facility, i.e., AMIGOS
An investigation of whether the ideas from this thesis make sense in an
object-oriented environment, i.e., AMIGOS phase four and five.