The emergence and growing popularity of electronic commerce in general and
Internet auctions in particular, has raised the challenge to explore scalab
le global electronic markets, involving both human and automated traders. W
hile centralized auctions are believed to provide better economic results i
n conventional markets, a single auction server over the public Internet is
not scalable and not reliable, due to the potential for very high load and
because of the unpredictable and generally low network bandwidth, and it m
ay not be feasible due to local autonomy constraints, which demand some deg
ree of control over local trading policies. GEM is a Java based generic fra
mework for decentralized Internet-based markets. It strives to balance the
single-market abstraction, which is important for economic efficiency, with
a geographically distributed physical market, which is essential for syste
m efficiency, scalability and autonomy. This is done through the introducti
on of wide-area-oriented yet consistent distributed trading methods, and a
scalable distributed architecture. The second main contribution of GEM is i
n the design of an individual (local) auction market. GEM provides a flexib
le architecture that delineates the abstract components of a generic market
and specifies control and data-flow constraints between them, but allows v
ariations in the concrete implementation of components with minimal or no i
mpact on the implementation of other components. III addition, GEM provides
dynamic (re)configuration and pluggability of policy components at runtime
, which enables to adapt market policies to variations in trading patterns
or system behavior, adding another level of flexibility to market designers
and administrators. (C) 1999 Published by Elsevier Science Ltd. All rights
reserved.