Cloud-based Java offers risks and rewards

VMforce shows the revenue potential of Java as a service, but will the cloud model save the Java community or split it apart?

1 2 Page 2
Page 2 of 2

Don't be surprised if other platform vendors increasingly move toward similar platform-as-a-service (PaaS) models. If it can work for VMforce, it can work for Oracle. In fact, PaaS may well turn out to be the strategy Oracle uses to underwrite future Java development, instead of traditional licensing.

The pitfalls of platform-as-a-service
But can it really work? Although the concept is appealing, PaaS as a business model remains largely unproven. Microsoft plans to move its Windows Azure PaaS offering to a commercial production stage this year, but developers will be wary if they judge by Microsoft's own performance. Microsoft execs claim the cloud model will earn Redmond more revenue even as its customers pay less. Yet despite this paradoxical assertion, in truth Microsoft's online services division is hemorrhaging cash at an alarming rate, posting a loss of $713 million for the most recent quarter alone.

Past attempts to enable Java in the cloud should also give developers pause. For example, Google added Java support to its App Engine cloud platform last year, but Sun's Simon Phipps has criticized Google's implementation for supporting only a subset of Java classes, a "forbidden" practice that Phipps deems "wanton and irresponsible."

According to Phipps, when a large vendor uses its resources to back a specific configuration of the Java platform, developers inevitably begin coding to that vendor's implementation, rather than the accepted standard. That flies in the face of one of the foundational principles of Java, which is that an application that runs on one implementation of the platform should run on others without modification. If that principle breaks down, the Java community risks market fragmentation and manipulation, as when Microsoft tried to ship a nonstandard version of Java in the 1990s.

In a market driven by PaaS, the risk of such fragmentation is great. Already Google offers its subset of Java while VMforce encourages the use of Spring, and each platform is based on a unique cloud storage back end. When Oracle enters the market it will likely favor Java EE and a more traditional database. As a result, apps written for one platform will not be readily portable to the others.

To forestall this fragmentation, Phipps believes Java needs a specific "cloud profile" that defines a standard set of APIs and functions that are available to Java apps running in a cloud environment. That profile should furthermore be developed as a community-based effort, he says, adding, "If we allow each cloud provider in turn to define their own subset, we will be left in the same ugly position we have with Java on mobile phones."

There's just one problem with that: Oracle, now the driving force behind the entire Java ecosystem, arguably has a stronger interest in offering Java as a service than any other vendor. In such an arrangement, other players may be hard-pressed to make their voices heard, making fragmentation all but inevitable. Faced with that alternative, a licensing model may still be Java's best bet for the future.

This article, "Cloud-based Java offers risks and rewards," originally appeared at InfoWorld.com. Read more of Neil McAllister's Fatal Exception blog and follow the latest news in software development at InfoWorld.com.

Related:

Copyright © 2010 IDG Communications, Inc.

1 2 Page 2
Page 2 of 2