First, this paper has reviewed the ODP/OMG CosTrading service. This review has presented the use of the service as being very technical and complex due to the lack of a structured approach. The various drawbacks brought up by the lack of type-checking at compilation time have been underlined. Then, the lack of formalism to define offer types and search operations has been presented as being one of the reasons of the service complexity.
Then, TORBA has been presented as a framework structuring the ODP/OMG trading service use. The conceptual contribution of this paper relies on the definition of the trading contract concept as a paradigm to structure the trading activity. The benefits of the TDL formalism use and its associated tools have been discussed. Using an example, the benefits of TORBA have been illustrated in terms of type checking, simplicity, productivity, and reliability of applications.
All the elements depicted in this paper have been prototyped and experiments have been performed using IDLscript and Java languages, as well as the ORBacus trading service [20]: TDL compilers (BNF and XML versions), proxy generators (OMG IDL, OMG IDLscript, and Java), runtime environments for IDLscript and Java, the trading contract repository, as well as the TORBA explorer are already operational. The next step is to finalize the TORBA environment in order to release it, and to obtain experiment/use feedback from end-users.
From now on, we have lots of work in view around TORBA: (1) support of C++ applications, (2) experiments over other CosTrading implementations, (3) measure of the overhead implied by TORBA proxies, (4) experiments of iterators, dynamic properties, and lookup strategies, (5) extension towards asynchronous trading (notification to applications of newly exported offers), and (6) use of the TORBA approach in the context of Jini, trading serialized objects and not only references.
In the meantime, TORBA is part of our actual research work. We intend to use TORBA in order to experiment the concept of Component Oriented Trading (COT) [27]. In that, TORBA would become the basis of TOSCA (Trading Oriented System for Component-based Applications), whose goal is to provide an environment to deploy and to administrate distributed component based applications [12].
Finally, in a more ambitious vision, we intend to consider the
benefits of a language to perform queries and to act upon distributed
objects. The goal would be to unify search operations on trading
services, object-oriented databases, and object environments à
la JavaSpaces [25].
This language could be named TORBA Query Language, relying upon
the following equation:
