Summary and conclusions

The ECIMF project made an attempt to address the interoperability problems by providing a single general and holistic view of all major aspects involved in solving concrete integration scenarios between e-commerce partners.

The most important outcome of the project seems to be the 4-aspect model of interoperability:

We have investigated various existing approaches that address each of these areas, and tried to indicate which of them need further research. Based on this, we present the following conclusions and recommendations for future work.

1.1 Interoperability of Business Contexts

REA models (retro-fitted to virtual organizations) help to understand interoperability issues on the value-chain level. This is because they provide a formal framework to describe contractual commitments and their relationship to partners’ collaborations, transactions and processes. They also help to identify differences in local business context.

Recently, REA Enterprise Modeling Framework has been adopted as a central part of business models in ebXML.

The conclusion of ECIMF project is therefore that the application of this or similar framework is required for proper understanding of business-related constraints of integration scenarios. We recommend that further work be expended to formalize this aspect of interoperability, and especially how it influences the interoperability of technical infrastructures.

1.2 Semantic Interoperability

Today there are islands of well-defined semantics for use in e-commerce, such as universal classification schemas (EAN/UCC, UNSPSC …) and standard e-commerce frameworks (RosettaNet, OAGIS, ebXML, xCBL …).

But there is no generally available, overall and unified business semantics across existing standards. Similar business concepts are being expressed differently, using different semantic depth, which results in ambiguous and overlapping concepts when considered in an integration scenario. This in turn leads to drastic increase in complexity and cost of integration. This also prevents ad-hoc collaboration scenarios between partners using different e-commerce frameworks. Well-established older standards will linger, so that this aspect of integration will not go away any time soon.

The ECIMF project group has identified the need for better and more effective methods for semantic mapping. Some of the most promising methods use upper-level shared ontologies – however, there is no such common unified ontology available at the moment. Readers are encouraged to review Annex 3, where this problem is discussed in depth.

Some of the existing projects are working intensively in this area, specifically:

· ISO TC/154 Basic Semantic Register: provides a cross-linked reference to key concepts across several existing e-commerce standards.

· ECIMF Semantic Mapping Tool: provides a prototype tool to facilitate semantic translation process, with use of shared ontology.

· OntoWeb projects: several projects, e.g. on ontology-based integration of content standards (SIG1), and industrial applications of ontologies (SIG4)

and other similar projects. However, there is still much to be done before the average e-commerce user begins to benefit from this work.

The ECIMF project clearly identifies this issue as a fundamental integration problem, and recommends both further basic research into efficient methods of semantic mapping, and a development of upper-level shared e-commerce ontology for the purpose of such mapping.

1.3 Interoperability of Business Processes

The ECIMF project has identified the need to reconcile incompatible definitions of business processes, as specified by different e-commerce frameworks.

Although good and comprehensive models for business process modeling exist (e.g. the one developed by UN/CEFACT ebXML project), there is little or no work being done on process mediation across standards. This is a very complex and non-obvious issue, which involves elements like transaction preservation, observing the timing constraints, compensation for failed transactions, legal consequences of failed transactions, partial fulfillment and others.

The ECIMF project recommends further research in this area. We also suggest that a separate, well-defined module (here referred to as Process Mediator) should be responsible for addressing these issues. Initial requirements and suggestions for possible architectures have been presented in this document.

Currently, the project members are aware of just one research project, which tries to address this integration aspect in a systematic way – the Process Broker project at Swedish Royal Institute of Technology (http://www.dsv.su.se/~pajo/processbroker/index.html), led by prof. Paul Johannesson.

1.4 Syntactic interoperability

The issue of syntax mapping is the most common aspect of interoperability being addressed today by software vendors. There are many existing software suites which concentrate mainly on this aspect, while offering only very limited functionality in all other integration aspects, as identified above.

Unfortunately, as the ECIMF project concludes, interoperability of message formats and transport protocols is also the last issue to be addressed when implementing integration solutions, and probably the most straightforward – that is, as soon as all other constraints (semantic and dynamic) are well understood. This low-level mapping quickly becomes very complex and difficult to maintain, if it is not driven by underlying higher-level models.

Therefore we recommend that vendors of integration software suites should concentrate on development of model-driven tools for system integration, taking into account the high-level e-commerce models being developed by recognized standard bodies and industry forums (such as UN/CEFACT ebXML, RosettaNet, OMG, OAG, UBL and others).

1.5 Software tools

ECIMF Project has delivered a prototype tool that illustrates some of the principles developed by the project. This tool is available under liberal Open Source license (so called Mozilla License), and can be downloaded, together with Java source code, from http://www.ecimf.org .

In the course of working on the tool, one of the obstacles was the lack of machine-readable models of e-commerce frameworks. In some cases, like EDIFACT directories, even though such sources exist they require substantial development effort (or investment) to process the data due to their historically complex formats.

Therefore ECIMF project members recommend that efforts should be spent to prepare (or convert) machine-readable models of existing e-commerce frameworks in popular formats, such as XML, and XML applications like XMI and RDF.

If these existing sources of e-commerce concepts and models become easily available for processing and analysis in contemporary well-documented formats, for which parsers and development tools are freely available, then we should expect both a significant increase in reuse of this rich heritage, and a decrease in cost of software solutions.