December 2, 1996 10:00 AM ET
Distributed objects of desire
The Internet and new product classes have made distributed objects viable--even desirable
By Rusty Weston

  When Charles Nettles hits the return key on one of the five workstations in his high-rise office on Dec. 31, he will launch the riskiest application of his 20-year-plus IT career.

Nettles will start the new year in the vanguard of the distributed objects technology movement--best known for its dizzying array of ORBs (object request brokers), development tools and Internet protocols. That kind of leading-edge strategy wasn't mandated by his employer, McKesson Corp., a $21 billion pharmaceutical distributor based in San Francisco. But Nettles, director of technology in the IT division, believes he guessed right in laying an object foundation to create Supplynet, a dial-up network featuring a variety of cost management decision-support tools for McKesson customers.

The crossroads

Object-oriented programming may not have always lived up to its compelling hype in the past 25 years, but now it is gaining the upper hand among ISVs and IT developers. What's changed this time? The easy answer is the Internet, which has ushered in a world of standardized transport mechanisms and browsers and distributed applications.

But a closer look reveals that the intersection of three product trends has made distributed objects not only viable but desirable as well.

First, there's the remarkably rapid acceptance of so-called light objects produced by Sun Microsystems Inc.'s Java and Microsoft Corp.'s ActiveX. These objects are simpler to produce than "pure" C++ or Smalltalk; Java is cross-platform now, and ActiveX is headed that way.

There is also the rise of component-style, object-based architectures in enterprise applications. The basic concept of plugging in additional functionality is being applied to applications ranging from Netscape Communications Corp.'s Navigator to Oracle Corp.'s WebServer cartridges to SAP AG's R/3. IT shops are also starting to think in similar ways about building functional components for in-house applications.

And there's the availability of ORB clients. Netscape will put Visigenic Software Inc.'s run-time ORB that supports IIOP (Internet Inter-ORB Protocol) into the next version of Navigator, code-named Galileo, late in the first quarter of 1997. Given Navigator's vast installed base, by mid-1997 there could be upward of 20 million CORBA (Common Object Request Broker Architecture) clients. This is a breakthrough for distributed objects because CORBA typically is limited to server-to-server applications running on Unix.

"Distributed objects are prime time," asserts John Strand, director of technology planning and integration at Sprint Corp., in Overland Park, Kan. "If you're not thinking about distributed objects, you're behind the times."

Suddenly, it's a brave new world for distributed objects. From the Galileo version of Navigator, a user will be able to issue a request to an IIOP-compliant ORB for an object, no matter where it is located. This not only explodes some of the limitations of two-tier, client/server networks, but it raises the expectation of seamless connectivity to multiple hosts, particularly on "encapsulated" legacy systems.

"Most people have supported these ideas for years but have been unable to build the networks because networks were not connected and client/server applications were built for specific platforms," says Eric Schmidt, chief technical officer at Sun, in Mountain View, Calif. "This has obviously changed with the Net, Java and CORBA."

By any measure, the prospects for distributed object technology are improving. International Data Corp., in Framingham, Mass., estimates the world market for object middleware is a modest $122 million this year, but it will almost double every year until it hits $1.6 billion worldwide in 2000.

Building for 2001

At one point this past year, Nettles had 70 object programmers working on his multimillion-dollar Supplynet project. His development team mostly worked in ParcPlace-Digitalk Inc.'s Smalltalk; Planning Sciences Inc.'s Gentia, an object-oriented online analytical processing development environment; and Oracle7, primarily on Unix. One of his biggest headaches concerned the integration of two CORBA-compliant ORBs: Visigenic's VisiBroker on the client and DNS Technologies Inc.'s Smalltalk Broker on the server.

And Nettles still has his work cut out for him. By the end of 1997, he expects to move all the customer applications over to Java--both on the client and server--and to make Supplynet available on the World Wide Web.

Nettles is not only building a flexible foundation for adding new application functionality, he's also seeking a strategic business advantage--mass customization for McKesson's clientele. And that demands distributed objects, insists Nettles. "The alternative is to code everything individually for customers, which would be exorbitant," says Nettles. "Or we would take an approach that one size fits all," he adds, and customers might reject the available services.

Similarly, Sprint is looking at distributed objects to build an object-based management system for its Sprint ATM and Sonet products. Strand's development team does most of its work in C++. Like Nettles, he also faced serious ORB obstacles, investing a significant amount of time working with Expersoft to tune its CORBA-based ORB, called PowerBroker, to meet his rigorous performance standards. The customer advantage of Sprint's object-based system will be the option of adding new services or network elements in hours rather than in days or weeks or months, Strand explains.

Creating a distributed-object-based application wasn't even possible until recently, he argues, because "objects were never defined in a singular way that everyone agreed to and used." While ORB interoperability is becoming a reality thanks to IIOP, one old complaint about object technology still rings true: Reuse is hard to achieve and is often downplayed in business justifications.

On the front lines

Given that and other difficulties, it's no big mystery why traditional object technology has failed to catch on in many corporate circles. "This elitist technology hasn't come close to matching up with mainstream developers' pedestrian skill sets," states Forrester Research Inc.'s Donald DePalma, an analyst based in Cambridge, Mass. He also claims that object technology has failed to mature because it is in a constant state of flux, as witnessed by the shift in interest from reusable C++ code to creating "interchangeable components."

With distributed objects, there are also problems that need to be overcome, not the least of which are ORBs that don't interoperate. Also, in a distributed environment, you can usually count on a small percentage of computers to be offline and carrying outdated objects. "There has got to be one current version of an object that is used as a standard," explains McKesson's Nettles.

But when it comes to objects, nearly every vendor has his or her own standard. And one of the first questions everyone asks is, will Microsoft Corp.'s DCOM (Distributed Component Object Model) obliterate the Object Management Group's CORBA or will the two work hand in glove?

"In the short term, users will very much have to choose sides," contends Forrester Research's DePalma. "As the technology evolves, it will get easier to have both technologies supported by a variety of system servers." DePalma asserts that choosing between Microsoft's and Netscape's market strategies is dependent upon whether your shop is primarily concerned with integrating Windows clients with the Internet (Microsoft) or creating Internet applications "from scratch" that connect to legacy services (Netscape).

According to interviews conducted by Forrester Research, most IT shops haven't picked an object model. Among those aligned shops, however, Microsoft DCOM enjoys a 24 percent to 14 percent edge over CORBA/IIOP.

Oracle, which also is aligned with CORBA, has been talking up its fledgling NCA (Network Computing Architecture), an HTTP server technology intended to enable ActiveX, OLE, Java, CORBA, IIOP and other software components to work together. "Microsoft was the only company that had an object model in large deployments," contends Peter Relan, Oracle's vice president of Internet server technology in Redwood Shores, Calif. "But they completely missed distributed objects."

Microsoft, which is shipping DCOM in Windows NT 4.0, begs to differ. "These CORBA guys had their shot," asserts James Utzschneider, Microsoft group product manager of server development in Redmond, Wash. "They were never able to deliver their component integration story in a cost-effective manner."

The reality is both sides are a tad premature. Oracle's NCA object middleware won't ship until the middle of 1997; Microsoft also is filling out its vision for distributed objects--a DCOM transaction server known as Viper is due early next year.

Still, all the bickering is spreading confusion in IT shops. Rising above the object wars isn't easy, according to Dennis Jones, CIO of Federal Express Corp., in Memphis, Tenn. "I don't think it's possible to ignore the market presence and politics of the industry," says Jones. "FedEx is a diverse environment with multiple groups pursuing parallel efforts. To date, we have not attempted to pick a single strategy."

Annrai O'Toole, chief technology officer at Iona Technologies, in Dublin, Ireland, believes the object war has been overblown in the media. He is confident that Iona's Orbix ORB will bridge the disparate technologies--and there are other ORBs that make a similar claim. "You can make COM and CORBA live together," O'Toole says. "Our motto is incompatibility is business--it's a huge opportunity for us."

Where it's going

Clearly, distributed object technology has a long way to go to become commonplace in most corporations. That's partly because it's hard to grasp. "You can't have a conversation about distributed object management deployment with people outside of IT," contends Boston Consulting Group's Gary Curtis, a partner in the Worldwide IT Practice, in Chicago. "Not being able to explain the business value is a showstopper in a lot of places."

For many companies considering distributed objects, there will be questions regarding performance and transaction management, according to Marilyn Martin, a partner and object-oriented programming specialist at Computer Sciences Corp., in Waltham, Mass. "In the past, we've seen clients who try to do it with very little knowledge of these issues and have bit off more than they can chew."

Unquestionably, distributed object computing is going to pose a daunting hurdle to most IT organizations, especially because of the difficulty of finding the right development staff and choosing the best tools. But distributed object technology will continue to thrive, says CSC's Martin, because "if you do it right, it will provide you with an environment that will take you into the millennium." *

Senior Editor Rusty Weston can be reached at rusty_weston@zd.com.

Copyright(c) 1996 Ziff-Davis Publishing Company. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of Ziff-Davis Publishing Company is prohibited. PC Week and the PC Week logo are trademarks of Ziff-Davis Publishing Company. PC Week Online and the PC Week Online logo are trademarks of Ziff-Davis Publishing Company.

Send mail to PC Week