Web Services @ Work

Web services is often spoken of as a future technology, yet it's important to understand that Web services is already proving to be a key component of the products and projects of today.

In this article I examine how Web services has become an enabling piece of Epic Data's Connectware for Baan toolkit, which is designed to easily connect wireless devices and other clients to the Baan warehousing and manufacturing ERP modules. I then study the requirements and realities that led Epic Data to select Web services over competing technologies. Additionally, I examine the reasons why Epic selected The Mind Electric's GLUE as their Web services foundation. In summary, I hope to show that even though their role will grow in the future, Web services is at work for developers today.

Before I start, it's also worth noting that this is not meant to be a detailed technical study of a Web service architecture. There are no code samples, and concepts are discussed at a fairly high level. This seems appropriate as the article is intended to be an example of a Web service success and not an instructive how-to for building technology. Developers might use this and similar articles to convince technical managers to adopt or champion Web services within their organization. More technical questions can be forwarded directly to me.

Introduction to Epic Data
Epic Data is a products and consulting company that delivers solutions in the supply chain, wireless, warehousing, fulfillment, and ERP markets. Many of their solutions focus on wireless connectivity and data capture for large ERP systems such as Baan, SAP, and Oracle. These solutions have been used in over 2,000 installations worldwide and support many Fortune 1,000 businesses.

Introduction to Connectware for Baan
Connectware for Baan (CFB) is a well-established, industry-leading product that connects wireless devices (primarily bar-code readers) to the Baan ERP system. CFB is Epic's leading connectivity product and critical to the company's success. CFB's primary responsibilities are:

Warehouse management
Production reporting
Label generation

An example and common use of CFB is the acceleration of receiving goods into a warehouse by the automatic assignment of serial numbers to incoming goods, the generation of appropriate labels, and the suggestion of efficient storage locations. The product acts as the middleware that ties together the barcode reader, label printer, and underlying ERP system. This process saves valuable time and reduces human error (see Figure 1).

Until recently, CFB was implemented as a set of Perl and C programs that managed all of the interactions between the client devices and Baan. This branch of the CFB solution has a broad install base of customers and a strong history of success. It's critical that any future versions of the CFB product line offer a smooth upgrade path.

In early 2000 Epic Data began work on a new version of the CFB product line designed around the following goals:

Cross-platform services: The middleware solution would need to be easily ported/supported across the various flavors of Windows and Unix popular with Epic's customers.
Cross-language clients: Because Epic data deploys these solutions in many heterogeneous environments, they're often charged with integrating the product into other systems written in various development languages. Epic wanted to support integration with popular development environments in a cost-effective and reliable manner.
Cross-Baan support: Versions 4.X and 5.X of the Baan ERP suite are widely used and supported. This makes it necessary for Epic to support both versions of the product. Epic wanted a product which both insulated the client developer from the differences found in these two versions but also provided access to version-specific features at a lower level.
Standards-based: Epic Data decided to make a firm commitment to supporting open standards to reduce development, support, and integration costs and increase customer and partner prospects.
Backwards compatibility: Because of Epic's significant customer/install base, it was essential that any new solution could seamlessly support the existing clients without disruption.

Eric Fritz, Epic Data's director of Baan product development, describes the need for greater product flexibility, saying:

"Our customers typically don't care what technology we use, but they do care about how much our solution may restrict them from making other business decisions. Yesterday if a customer told us they were going to move their servers from Unix to an AS400 we would have to undertake a major project to migrate their solution, and as we all know that means money out of their bottom line. Today, we can say 'No problem.' The same holds true for the version of Baan and the tools that they want us to use. It's all about the freedom to choose.

While the list of goals was much more extensive than what we have discussed, the above requirements are the ones that most impacted the product architecture.

The Selected Architecture
Epic Data made the decision that the core of the next generation Connectware for Baan product would be written in Java and that this functionality would be exposed as Web services. An architectural overview is shown in Figure 2.

Java was selected because it easily satisfied the cross-platform requirements and it also represented a significant commitment to open standards. Epic Data also leveraged Java's interface features and design by contract to expose cross-version features in a generic manner. The selection of Java, however, did not meet all of the above criteria.

To achieve integration with clients across multiple languages and facilitate backwards compatibility with legacy Perl and C code, Epic Data turned to Web services defined as WSDL, SOAP, and XML. CORBA was also considered as a possible solution but was determined to be overly complex and expensive to implement. SOAP was selected over other XML-based Web services because of the incredibly strong industry support that the standards are receiving in popular development languages like Visual Basic, Perl, C, C++, and Java.

Finally, it was the possibility of meaningful backwards compatibility that sealed the Web services/SOAP selection. Epic Data decided that they would create a Perl Legacy Adaptor to interface existing Connectware for Baan implementations with the new service. This Perl layer "intercepts" requests intended for the legacy back end and routes them to the new service via SOAP (see Figure 3).

GLUE: A Web Services Platform
Upon selecting Web services as a key part of their solution, Epic Data then had to select a Web services platform to enable their code. Ultimately, Epic selected The Mind Electric's GLUE. The Mind Electric describes GLUE :

"GLUE is 100% Java and has a rich set of features including an embedded Web server, servlet engine, SOAP processor, XML parser, graphical console, dynamic WSDL generator, dynamic Java/XML mapping, UDDI client, UDDI server, WAP support, and XML persistent storage system."

A recent review of GLUE in Information Week by Jason Levitt noted that GLUE's purpose was to bring the same simplicity seen in .NET to Java developers. GLUE succeeds by offering a high degree of Web services functionality combined with incredible ease of use. Epic Data selected GLUE for the following reasons:

Relative maturity: At the time of selection, GLUE had a very competent implementation of SOAP 1.1 and great support for WSDL. GLUE also had an expressive implementation of UDDI's Publish and Inquiry functionality.
Rapid advancement: TME's pace of development for GLUE would be rapid for a large company, let alone a typically understaffed startup. Releases are frequent and bugs are quickly reported on the product's mailing list and confirmed/fixed by TME staff.
Ease of use: Both publishing and subscribing to Web services is made easy with GLUE. In both situations GLUE dynamically creates the necessary classes leaving the developer free from excessive hand-coding or messy code generation tools.
Speed: GLUE was much faster than many of its competitors. TME has stated that this is mainly a result of using TME's proprietary XML parser and toolkit ElectricXML.
Low vendor lock-in: The use of GLUE is nearly transparent on both the client and server sides. Epic Data estimates that less than 1% of the project code knows anything about GLUE or SOAP.

Jonathan Barksdale, software engineer at Epic Data, describes his experience with GLUE:

"I initially implemented and tested my would-be server methods locally. Then I added GLUE to the mix in a matter of minutes using only a few lines of additional source code. I can't believe the ease with which my formerly local application methods were working as distributed, standard Web services."

While maturity, rapid product advancement, ease of use, and speed were all strong factors in its selection, the almost non-existent vendor lock-in made the final difference. GLUE, aptly named, binds cross-language ponents together and leaves a minimum of technical residue. For Epic Data to switch from GLUE to another Web services platform, there will be very little code investment to throw away. Essentially, the only thing Epic would need to do is the total amount of work required by the new solution. This all suggests a high degree of confidence by TME that GLUE will continue to be used on its own merits rather than maintained because it's painful and expensive to remove.

Summary
Epic Data, with a significant history of success in the Baan connectivity space, has made a significant commitment to Web services and to open standards. This investment was made not to satisfy a need for market hype but rather because these commitments offered tangible technical and cost benefits over alternative solutions. By using Web services combined with Java on the server side, Epic Data was able to build a solution that was portable across languages, platforms, and versions of Baan. Additionally, we saw Web services empower Epic to select a new development language and environment while preserving backwards compatibility.

Through the selection of GLUE, Epic ensured that Connectware for Baan's Web services were standards-compliant, high-performance, and on the leading edge of development in this new market. GLUE's lack of heavy coding requirements made it especially attractive, as Epic did not feel it was headed toward irreversible vendor lock- in. As CFB evolves, Web services are sure to remain a critical component of the overall architecture.

Web services is at work in a quiet but pervasive way across the computing industry. Epic's focus on these technologies is cutting-edge but certainly not unique. Web service-oriented architectures are being sketched on white boards around the world and have already made their way into a first wave of products and projects. While many of the underlying technologies are new, Web services already provide a better solution, in many cases, for cross-platform and cross-language computing than preceding technologies and standards. Web services is hard at work today and this trend shows every sign of increasing.

Author Bio:
Michael Sick is an independent Java architect helping clients solve complex product definition and design problems. He has more than eight years of experience in the construction of distributed information systems and Internet technology, holding such positions as architect and VP of development. mike_a_sick@yahoo.com

Java and Java-based marks are trademarks or registered trademarks of Sun Microsystems, Inc. in the United States and other countries. SYS-CON Publications, Inc. is independent of Sun Microsystems, Inc.