Focusing on open APIs for enterprise applications

Open Web Magazine

Subscribe to Open Web Magazine: eMailAlertsEmail Alerts newslettersWeekly Newsletters
Get Open Web Magazine: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn

Open Web Authors: Jayaram Krishnaswamy, Bob Gourley, Kevin Benedict, Pat Romanski, Newswire

Related Topics: OSGi, Java Developer Magazine, Open Web Magazine

OSGi: Article

Adopting OSGi in Java Application Frameworks: A Case Study

Defining reusable software system components

Lessons learned
In our work on applying the adaptation pattern to the existing application platform we have been faced with noticeable problems and results not obvious at first sight.

1.  Java as a language makes it easy for developers to start and produce results in a relatively fast manner. Unfortunately this doesn't imply that the code created fulfills certain quality criteria called the best practices (see, for instance, [12], [13] and [14]). On top of that, component based development in general, and OSGi in particular, assumes that application developers at least follow the best practices. This creates the necessity to rearrange and, sometimes, modify parts of the legacy Java code or provide a wrapper to compensate for the shortcomings introduced by that code. Our experience shows that most of the legacy code requires at least minor changes and can't be 'bundle-ized' as it is.
2.  Even if the application framework is based on logical components and claims to be component oriented, its component model may be flawed. Unless the componentization is enforced with a proper code separation and the isolation level guaranteed by class loaders, like in OSGi, the logical components still do not adhere to the real definition (see ?[19]). Very common examples of component model flaws are circular dependencies on the package or even JAR level.
3.  Although OSGi introduces a very potent componentization platform, the focus is on structuring Java classes. Resource centric frameworks, like UIMA, need to create their own infrastructure to compensate for shortcomings created by their resource oriented view. Many shortcomings are not originated by OSGi, but are a result of the Java limitations. For example, the JVM limits the number of custom URL handlers that can be specified for a single protocol to merely one, a similar problem exists with the java.sql. package, which assumes a global implementation instead of a federation of different implementations and versions, another problem is that the protection of the core framework from malicious code introduced by third party components goes beyond the capabilities of Java runtime and the SecurityManager, just to name a few.
4.  Based on the nature of OSGi, every global entity such as a distinct class shared among your application bundles might become a real problem, unless these entities are not treated as first class citizens. Examples of such entities are JCas cover classes in UIMA, which are automatically generated by the frameworks and usually packaged with each analytic component. If such classes are not externalized and referenced as first class citizens, multiple identical classes with different class loaders may appear in the class path. In plain Java, this is no problem, because the application usually has only one class loader. In OSGi, however, this will cause a ClassCastException if one component needs to cast a class, already instantiated by a class loader not compatible with this component's class loader. To avoid problems in OSGi container, these classes have to be isolated in separate bundles and made available as entities of collaboration for all component developers.

Looking from a bird eye view at our approach, we can conclude that we achieved our goal by adding a thin layer of abstraction to hide all the complexity of the OSGi adaptation from the application and component developers. This layer allows both kinds of developers to switch to a component based environment with minimal change of the current code and almost no change in the programming model. However, if one wants to benefit from all the features introduced by this new componentization approach, especially the dynamism and updatability, a shift in the application programming paradigm towards an asynchronous event based application model is essential. Furthermore, component reusability is not guaranteed by simply dividing and packaging an application in several components. We are facing problems here, which can not be easily solved by following coding best practices. A deep understanding of the problem and the ability to abstract from a concrete problem to a general solution is crucial in defining reusable software system components. Problems like the configurability and customization of resources outside of the application framework API, as well as the version management constraints are also left open in this article and need to be addressed separately.

[1] OSGi Technology,
[2] JSR 277: Java Module System,
[3] JSR 291: Dynamic Component Support for Java SE,
[4] OSGi Projects @ JavaOne 2006,
[5] About the OSGi Service Platform,
[6] UIMA @ IBM Research,
[7] UIMA @ Wikipedia,
[8] Ferrucci, D., Lally, A., "UIMA: an architectural approach to unstructured information processing in the corporate research environment", Natural Language Engineering, Volume 10 , Issue 3-4 (Sep 2004), pp. 327-348, Cambridge University Press, New York, NY, USA, 2004, ISBN: 1351-3249
[9] Goetz, T., Suhre, O., "Design and Implementation of the UIMA Common Analysis System", IBM Systems Journal, v.43, # 3, 2004, pp.476-489,
[10] Ferrucci, D., Lally, A., "Building an example application with the Unstructured Information Management Architecture", IBM Systems Journal, v.43, # 3, 2004, pp.455-475,
[11] Schor, M., "An Effective, Java-Friendly Interface to the CAS", Research Report RC23176, IBM T.J. Watson Research Center, Yorktown Heights, NY, 2004, c85256b360066f0d4/d55875841121943e85256e78004bd826?OpenDocument
[12] Gregory Brail, John Wells: "The Good, Bad, and Ugly of OSGi: What we learned building the mSA Backplane",
EclipseCon 2007,
[13] BJ Hargrave, Peter Kriens: "OSGi Best Practices", JavaOne 2007,
[14] Jahn, Mirko: OSGi applied, Teil 3: OSGi best practices. In: JavaMagazin (2007), 9, P. 97-102
[15] Nash, M., "The Java Framework Landscape",
[16] Application Framework,
[17] Kozakov, L., Jahn, M., Doganata, Y., "OSGi Enablement in UIMA Framework", IBM Research Report RC24308, July 2007, 7c85256b360066f0d4/c77885d8151bed8d8525731c0067d980?OpenDocument
[18] Dynamic Module Enabler for Unstructured Information Management Architecture Analytics, IBM alphaWorks, June 2007,
[19] Szyperski, C. et al., Component Software - Beyond Object-Oriented Programming, 2003, ISBN: 0-201-74572-0
[20] Jahn, M., "OSGi applied, Teil 1: Komponentisierung in Java", In: JavaMagazin (2007), 8, P. 54-59
[21] Fowler, M., "Inversion of Control Containers and the Dependency Injection pattern",
[22] IBM WebSphere Application Server V6.1: Componentization Overview, index.jsp?topic=/
[23] Spring-OSGi,

More Stories By Yurdaer Doganata

Dr. Doganata is the manager of the Information Management Solutions group at the Watson Research Center in Hawthorne, New York. He received B.S. and M.S. degrees from the Middle East Technical University, Ankara, Turkey, and a Ph.D. degree from the California Institute of Technology, Pasadena, California, all in electrical engineering. He joined the Watson Research Center as a research staff member in 1989 and worked on projects in many diverse areas, including high-speed switching systems, multimedia servers, intelligent transportation systems, multimedia collaborative applications, eservices, and information search and retrieval systems for technical support. His current work involves designing and prototyping innovative solutions, applications, tools, and utilities in the area of unstructured information management. Dr. Doganata hold several patents and research awards and is the author of numerous papers.

More Stories By Lev Kozakov

Dr. Lev Kozakov is a research staff member at IBM T.J. Watson Research Center and is a member of the dBlue project architecture and research teams. He has worked in many areas, including dynamic systems, applied statistics, information management systems, man-machine interface, medical software, computer telephony, and design patterns. Lev holds a number of patents and is the author of several publications.

More Stories By Mirko Jahn

Mirko Jahn worked for several years as an independent software consultant and tutor before joining the Unstructured Information Management Architecture (UIMA) group at IBM Research. At present, his major research areas cover component-based development strategies, component repositories, and migration approaches toward modular applications, along with service-oriented architecture. Mr. Jahn is heavily involved in making use of OSGi component-based developement
standards in UIMA.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.