A well-planned Web Service interoperability environment begins by clearly defining who your Web Service consumers are now and in the future. There was a time not so long ago when you could count on a fairly homogenous consumer population. This was about the same time that you were happy just to be able to get a Web Service running in the first place and finding a consumer who could actually interact with your Web Service was cause for celebration. Those days have changed however and Web Services interoperability, once a "fancy" addition to your SOA design, is now a key and indispensable requirement in most SOA scenarios.

Today, SOA architects must contend with complex scenarios that assume a variety of Web Service consumers or in many cases are asked to create Web Services that are generic enough to be interoperable with just about any known consumer. If you're serious about implementing SOA in your worldwide enterprise you'll have to forget about the luxury of being able to dictate the configuration of all of your consumers and instead build fully interoperable Web Services. Most robust Internet Web Service APIs, familiar to all of us such as Amazon.com, eBay, or Salesforce.com, have learned this lesson already with very successful results. Those companies know that Web Service interoperability lies in the architecture approach and not in the implementation of new middleware.

What Makes Interoperability So Challenging?

Standards Proliferation and Complexity
The interaction between Web Services and consumers is rooted on a set of standards developed by standards organization committees at OASIS, W3C, and WS-I. These standards such as XML Schema, SOAP, WSDL, and WS-* protocols are intended to provide a technology-agnostic level of abstraction over the service implementation which should theoretically guarantee interoperability. However, it's not the standards themselves, but their implementation by individual Web Service technology vendors that must be interoperable. In a perfect world, every vendor would implement Web Services standards in exactly the same way, which would guarantee interoperability out-of-the-box. As you must know, that's just too good to be true. In the real world, every vendor implements these standards sometimes with just slight variations and other times entirely different versions. In many other cases vendors choose which aspects of a standard to implement, or maybe even choose not to implement a given standard at all. This proliferation and complexity is most obvious when we look at the WS-* protocols.

When this article was written, up to four different WS-Addressing versions were in use. Three versions of the specification are named by their release date: the March 2003 version, the March 2004 version, and the August 2004 version, developed before the specification moved to W3C. The fourth version, 1.0, was completed in May 2006 and developed after the specification went under the W3C umbrella. And if that's not confusing enough, after moving to W3C, the specification split into multiple parts: a core specification and two other specifications that describe bindings for SOAP and WSDL.

There are also different WS-* protocols that address very similar scenarios such as WS-Eventing and WS-Notifications, WS-MessageDelivery and WS-Addressing or WS-Reliability and WS-ReliableMessaging. Committees from OASIS and W3C are working to unify those overlapped protocols into a single set of standards. Vendors can and often implement different versions of the same WS-* protocol or implement one of multiple similar standards such as WS-ReliableMessaging instead of WS-Reliability.

Although the combination of SOAP and WS-* protocols provide a solution for some of the most interesting challenges in distributed computing, the complexity of this approach makes it impractical and nearly impossible to implement in real-world interoperability scenarios. Just a small subset of the over 100 WS-* standards specifications available today have been implemented by vendors. One last, but certainly not least, challenge involved in using SOAP and WS-* protocols is that they can also limit the service availability to clients such as script applications or Web browsers that don't support that generation of SOAP messages.

Best Practice: Use WS-I Profiles
Some WS-* protocols have a WS-I profile available that contains some of the principles and a subset of features of the protocols that should be implemented to guarantee interoperability. Making the services compatible with a WS-I profile (is this one exists) is a standard and globally accepted approach to interoperability. However for some protocols, a WS-I profile isn't available or doesn't address the interoperability requirements.

Best Practice: Implement Multi-Binding Services
To support multiple versions of the same WS-* protocol it's recommended to design the services with multiple bindings per specific versions of the protocol. This approach segments the different types of interactions at the binding level improving aspects like versioning, management, etc. The code in Listings 1 and 2 illustrates a sample Microsoft Windows Communication Foundation (WCF) Service configured to support multiple types of interactions using bindings. Specifically, this service supports secure and reliable interactions as well as basic SOAP interactions using two different bindings.

Best Practice: Exposes SOAP and REST Interfaces
Resource State Transfer (REST) provides a simpler alternative to the use of SOAP and the WS-* protocols for some scenarios. The fact that REST is based on XML messages over HTTP makes it accessible to most of the client technologies on the market including browsers and script languages. Some services can expose both SOAP and REST interfaces and offer a broader set of options to consumers. The example in Listing 3 illustrates that approach using Oracle Application Server.

Aligning Code and Contract
When designing a service you must first consider where to start. Do you create your code first or your contract first? Or do you give them each equal importance by creating them in parallel. However, even if you were extremely diligent in developing perfect synergy between your code and your contract you would still find that the limitations of the basic standards themselves, such as XSD and WSDL, could easily do you in. For example, the XML Schema model presents severe limitations in terms of composability compared with most programming languages data structures. This is often reflected in non-optimal translations between XML Schema structures and programming data structures. Similarly, WSDL 1.0 and 2.0 are both too abstract for describing services that can be interpreted by Web Services frameworks in a consistent way.

There's a lot of debate in the Web Service community around whether to use a contract-first versus a code-first approach to develop Web Services. One of the common arguments in favor of a contract-first approach is that it facilitates interoperability. Whether that's arguably true, the reality is that just a few mortals know WSDL and XSD well enough to design solid service contracts. Given the complexities of both standards developers often end up designing non-optimal WSDLs and XSDs that are translated into poor service implementations. On the other hand, a code-first approach is more familiar to developers but can produce contracts that aren't interoperable.

Best Practice Recommendation
Some of the most successful Web Service implementations have been designed using a hybrid approach that combines the agility of a code-first approach with the flexibility of a XSD/WSDL-first approach. Following these technique developers can leverage their existing skills on a particular development platform to guarantee an optimal service implementation while the WSDL-XSD experts verify that the contract is suited to meeting the interoperability needs. Figure 1 illustrates this approach.

Multi-Transport Services
SOAP, WSDL, and the different WS-* protocols are transport-agnostic specifications. Theoretically it's possible to host the same service using multiple transports such as HTTP, TCP, JMS, etc. Although this feature is supported by some of the Web Services technology frameworks on the market such as Windows Communication Foundation (WCF) or Apache Axis, only HTTP has been adopted widely enough to be considered for interoperability scenarios. Another factor to consider when implementing multi-transport services in real-world scenarios is that certain transports require a specific behavior of the service. For instance, a Web Service that uses JMS as a transport probably implements one-way, long-running, and asynchronous operations. That behavior is fundamentally different from a Web Service that uses HTTP as a transport on which it makes sense to implement atomic operations using different multiple exchange patterns. Hosting the same Web Services using JMS and HTTP makes little or no sense in most of the scenarios.