One conclusion they came to is that Web services are simply the new middleware. And middleware is something the OMG knows about. CORBA Orbs sit at the heart of all Application Servers and most integration done to date has been reliant on CORBA to translate between systems. The OMG and its 800 members have learned a lot of lessons about middleware over the last 12 years and there was complete agreement on the enterprise qualities Web services need.

To begin with, Web services is only a part of the picture, it is still essential to consider existing integration platforms and the goal is a seamless integration of all technologies being used. All agreed that the continuing role of Java for tightly coupled and controlled systems will be extended, allowing Java applications themselves to integrate via Web services [Mike Connor, IBM].

While we are quite focused today on the low level of Web services (SOAP, UDDI, WSDL), it is critical that we raise the level of discussion to the enterprise level because the hard part of enterprise integration requires methods of managing the federated systems that are created almost as a side effect of implementing Web services [Peter Herzum, Herzum Software].

Enterprises need insulation from rapidly changing standards built into their modeling methods and development. Model based architectures of collaborating service components should be designed using the OMG’s UML profile for Enterprise Distributed Object Computing (EDOC). This enables organizations to create a lasting development asset that can be automatically mapped to the technology of the day using Model Driven Architecture (MDA) [Cory Casanave, Data Access Technologies].

Trust is key for large-scale use – and goes far beyond encryption and signatures. We need to know that the service we contract with will perform as expected, the identities involved, and that we can rely on them. We need non-repudiation so the sender can’t deny that what was delivered was in fact delivered [Bill Smith, Sun Microsystems].

When crossing enterprise boundaries, legal issues need to be built in so that you have enough evidence if you need to go to court.  Legal requirements demand clear statements of the nature of relationship and boundaries - defining who owns what and what they have rights to; as well as the degree of trust, expectations, liability risks - in short – who’s accountable for what. These are essential for quality of service agreements, for individual privacy, and for protection of shared intellectual property [Matt Hettinger, Mathet Consulting].

For this kind of comprehensive security we need legally binding contracts, registries that can understand contracts, and implementations between contracts rather than between end points [Mark Potts, Talking Blocks].

Current efforts at workflow management need to be expanded to enable the orchestration of transactions within and across enterprises so that controlled transactions can run for moments or months [Matt Hettinger, Mathet Consulting].

The holy grail of Web service to Web service management requires two services talking to each other and achieving transactions through document exchange with no humans involved. To enable such cross-enterprise management of web services we need to provide visibility and control with better dynamic selection and automatic negotiation of Web services [Akhil Sahai, HP Labs].

All agreed that in order for the Semantic Web to provide the ability for machines to reason about and select appropriate Web services, semantic interoperability is the major challenge.  Semantics refers to the meaning of terms and semantic drift tackles the fact that terms continually change meaning. In the end the goal is to have useful information you can process automatically with a program – a machine readable Web [Hugo Haas, W3C].

No matter how good XML tags are, how clean interfaces are, different terms will be used to describe the same things. Ontologies will provide a way to define language and rules based on agreements among those sharing common services. Ontologies can be mapped to one another and they facilitate conversations between agents on our behalf [Elisa Kendall, Sandpiper Software].

Configuration management across federated systems and time will also require strong semantics and ontologies that can map between vocabularies and understand and manage emergent meaning within its context [Matt Hettinger, Mathet Consulting].

Flagged as needed but not addressed were the heavy duty problems of dealing with asynchronous version changes of services within and between companies and the need to selectively expose services according to role and authentication based on business rules [Gerald Edgar, Boeing Corporation].

The only controversy centered around emerging standards, particularly those that the newly formed Web Services Interoperability organization (WS-I) is interested in adopting or ignoring. …

A recurring theme was that there is a huge danger that Web services will become unmanageable. The more complex these requirements became, the more apparent it was that a new trusted 3^rd party middleman might emerge. Web Services management platforms that can manage versions, contracts, transactions, security, and semantics will be essential [Mark Perreira, Talking Blocks].

Users were warned against trying to build their own Web services based on today’s standards and encouraged to find tools that output Web Services from higher-level development and analysis.

Copyright 2002 by Infomaniacs