NeXTWORLD October 1993
CORBA Ė Common Object Request Broker Architecture
A standard that allows objects to find other objects on the network and send them messages. Using CORBA, entire application programs can appear as single objects or services. CORBA was developed by the OMG (Object Management Group), an industry coalition founded in October 1989 to develop standards for network-based object-oriented environments.
DCE Ė Distributed Computing Environment
A standard for network-based computing developed by the Open Software Foundation. DCE consists of two parts: a developer environment for creating applications and a user environment for running them. The developer environment includes a secure remote-procedure-call (RPC) system, Kerberos-based authentication, a distributed directory service, support for replication, network time service, and support for multithreaded programming.
The DCE user environment includes a distributed file system, diskless workstation support, and "personal computer integration services," which allows file sharing and printing from PCs to DCE.
DME Ė Distributed Management Environment
A network-management system, based upon DCE, that is currently under development by the Open Software Foundation.
DO Ė Distributed Objects
NeXTís system for sending messages between objects that reside in different application programs or on different computers.
An instruction that one object sends to another object. Each message consists of a name and, optionally, one or more pieces of data. A simple message might tell an object to display a window on the computerís screen; a complicated message might compute whether or not to approve a home-mortgage application.
A small, self-contained, and reusable building block used for piecing together complicated programs. Objects communicate with other objects by sending messages.
PDO Ė Portable Distributed Objects
A version of NeXTís Distributed Object system that can run on versions of UNIX other than NEXTSTEP.
NeXTís Portable Distributed Object System
By Simson L. Garfinkel
When NeXT introduced its distributed-objects facility with NEXTSTEP 3.0, many developers hailed it as a breakthrough. Before 3.0, building an application program that operated over the network was a complicated proposition: programmers had to build by hand a set of remote procedure calls (RPCs) for sending messages and exchanging information between the client program and the server. With 3.0ís Distributed Objects, NeXT took all of that complexity and hid it behind a set of powerful yet simple-to-use Objective-C objects.
All NEXTSTEP programs are built out of objects. Using 3.0ís new features, developers could take an object or a group of objects out of an application program, compile them separately, and place them on another computer. The only change necessary to the application program was adding a single line of code telling the program on which computer the object now resided and the name with which it was being offered to the network.
One of the simple uses that programmers found for distributed objects was peer-to-peer communication between different applications running on a single workstation, says Rick Jackson, NeXTís director of developer environment product marketing. Using the distributed objects, a program can easily update information in another programís windows as easily as in its own.
In large, enterprisewide applications, Distributed Objects can play a far more important role. "You can break your application up into task-oriented processes," explains Terry Lindsey, vice-president of technology development at WilTel. For example, says Lindsey, a large application program might have a module that requires a large amount of computer power, memory, or access to a specialized database to perform a series of sophisticated calculations. Using distributed objects, that module can be taken out of the main application Ė which runs on relatively low-cost workstations Ė and placed on a powerful, centralized server.
When customers started breaking apart their programs and putting different objects on different systems, another advantage of distributed objects became apparent: the same object could be used simultaneously by different applications at the same time. This does more than save development time, explains Jackson: it dramatically saves the time necessary to redeploy applications when the object is changed or improved.
Take the example of Swiss Bank, where programmers have developed a series of objects that encapsulate different financial models. The objects are kept in a central object repository used by a variety of different applications running on the firmís trader workstations. Whenever a programmer improves the model object, every application instantly makes use of the new version.
But for all of its power, NeXTís Distributed Objects had a major shortcoming: NEXTSTEP objects could communicate only with objects running on other NEXTSTEP computers. NeXTís Portable Distributed object (PDO) system frees NeXTís Distributed Objects system from NEXTSTEP.
The PDO system will consist of two parts: a development environment and a runtime system. The development environment is what programmers will use to create distributed object based applications. It includes a version of the GNU Objective-C compiler that will produce object code that runs on the HP server, a version of the GDB debugger, and all of the include files and libraries for NEXTSTEPís foundation classes Ė the classes, such as Object, List, and HashTable, which are not involved in displaying graphics on the NEXTSTEP screen.
In addition, there is a portable BuildServer that runs on top of PDO, and talks to ProjectBuilder on the NEXTSTEP client side. This will enable developers to use ProjectBuilder to build their PDO objects on the target server.
For the PDO runtime environment, NeXT is developing a special program that will emulate Mach interprocess communication with the sockets that are available under System V UNIX. This program will also include a portable nameserver that will respond to requests from over the network for named objects and find them in the HP-UX environment.
Using PDO, sending a message to an object that resides on an HP 9000-series server running HP-UX will be no different than sending a message to an object on a computer running NEXTSTEP Ė except the response will come back much faster. Thatís good news for customers who want to use distributed objects for solving problems that would even tax Pentium-based systems.
NeXT will develop and support PDO for HPís PA-RISC, to be delivered in Q4 í93, and Sunís SPARC-based systems in 1994. Data General is licensing PDO source code and will port, market, and support a version for their Aviion servers.
But it wonít stop there. "PDO is designed to be portable," says Jackson, who adds that NeXT is having discussions with most major UNIX vendors Ė including IBM, Digital, and NCR Ė to put a version of PDO onto their systems.
Eye on the Prize
Hewlett-Packard and NeXT target financial-services industry with Object· Enterprise joint venture
By Dan Ruby
Itís a simple equation with a big payoff: NeXTís object technology plus Hewlett-Packardís enterprise systems equals new business and big profits for financial-services firms.
Object· Enterprise, the joint venture announced in May, combines the advantages of object-oriented development and client-server architecture to create ad deploy new financial applications and systems. These will be applied first in live trading systems but will propagate out to customer-service applications. Long-term, Object· Enterprise systems will be used to fundamentally restructure information systems Ė all the way, as HP executives say, from the desktop to the data center.
"The benefits of object orientation have been lauded in the industry but never really deployed on enterprisewide business applications," says Ruann Ernst, HPís director of financial services marketing.
So far, a small handful of early NEXTSTEP adopters have demonstrated substantial results using NeXTís tools to create custom trading instruments. With Object· Enterprise, they will be able to deploy these apps in global information systems based on HP workstations, servers, and mainframe-class systems.
Thatís the Object· Enterprise equation on paper. How it plays in the real world, where HP fights it out against companies like DEC, IBM, and Sun, and NeXT is in a David-and-Goliath battle with Microsoft, is very much an unknown.
Risk and Reward
Between deregulation and globalization, financial markets have changed radically during the past decade, as the distinction between banks, brokerages, and money managers has blurred and national borders have become irrelevant. Every financial-services firm is scrambling to find opportunities.
The goal: get more information, get it faster, and analyze it more completely than your competition, so you can make that calculated decisions to buy or sell before anybody else.
"Maybe thereís a weird bubble in an equation between Swiss warrants and Japanese bonds. If you find it, you win," says Jim McCrory, NeXTís point man on Object· Enterprise. "Trading is a zero-sum game: Somebody wins and somebody loses."
Traders are in the business of risk Ė currency risk, interest-rate risk, political risk. In July, when the European Community let member currencies swing to market value, billions of dollars changed hands in a day. With options on securities, currency, mortgages, and every kind of commodity, institutions have the opportunity to make vast profits for themselves and their customers, but they also need to manage their risk.
To deal with the complexities, securities firms have led the commercial world in the adoption of commercial workstations and client-server systems. Now these same firms have an opportunity to take the lead in object technology.
"Objects give us a new way to look at our business. Our ability to manage risk is much more granular than it was three to five years ago," says Dwight Koop, director of information technology at Swiss Bank Corporation, one of NeXTís largest customers. "When you decompose any problem, the objects that you ought to be carrying around and dealing with arenít necessarily transactional in nature."
To take advantage of the changing market, trading firms need an environment for rapid application development. A financial product may last days or weeks. The first to market with a new trading instrument reaps the reward.
Firms also need to integrate multiple sources of information together on one desktop, as well as distribute the information out to systems throughout the enterprise. The focus is no longer exclusively on the traderís desktop, but the need to integrate it with mainstream production systems and extend it to branch-office customer-service applications.
From front to back
A modern traderís workstation provides systems for real-time feeds, analytical modeling, historical data access, graphical displays, and links to the enterprise - all unified on a single desktop.
But it was not always so integrated. A decade ago, traders worked with video-based data feeds from companies such as Quotron or Telerate. Each system had a proprietary terminal that could display only predefined screens.
Later, information providers digitized their feeds so could pipe data directly into their financial models. Now firms could analyze the data and crunch it with powerful analytical tools that ran scenarios and calculated probabilities. Early analysis workstations were closed, but the trend in recent years has been to offer customers platforms that they can use to build their own, customized systems.
While real-rime feeds and analysis tools are fairly commonplace, firms still lack integration with their own databases of historical and technical data and to mainframe-based production systems for order processing and accounting.
These back-office systems are typically batch-oriented processes that involve delays of a day or more before reflecting current transactions. "Portfolio manager and the research group make the decision to buy or sell, traders execute it, and in the back office, people key it in. Right now, the method of communicating between those groups is purely paper," laments Duncan Wilcox, director of investment technology for Nicholas-Applegate, a San Diego-based money-management firm. "If we electronify the decision to buy or sell, and electronify the execution of it, then it is rather simple to automate the operational side."
To make up the discrepancy between the real-time needs of the front office Ė the traders and analysts Ė and the batch mode of the back office, a new function called the midoffice has emerged to handle the infrastructure software that serves as the link between the trade and the audit trail.
Peripheral to the trading function, financial firms also want to provide many of the same information systems elsewhere in the enterprise, such as retail brokerage and customer service.
With Object· Enterpriseís development system and client-server deployment architecture, firms can seamlessly integrate the trading process from front to back office. Ultimately, they can re-engineer their entire business around objects.
The benefits of object technology are no revelation for HP. The company was one of the early advocates of object activity at the enterprise level, helping to define distributed computing standards such as DCE and DME (see "Object Glossary"). Ernst was personally involved in the formation of the Object Management Group, which specified the CORBA standard for object intermessaging.
One piece that HP did not have, however, was a standout object-oriented operating system on the desktop. "Thatís what NeXT has Ė a world-class object-development environment that is deliverable today," says Ernst.
Together, the companies wanted a packaged solution that made it easy to market and sell the combined technologies. By using the Object· Enterprise name, NeXT hopes to develop a brand identity that is recognizable in the market.
To be precise, Object· Enterprise is a NeXT campaign that is not exclusive to any hardware vendor. Other server manufacturers such as Data General, NCR, and Digital Equipment may also be brought under the umbrella, but the HP relationship is special, according to Ron Weissman, NeXTís director of corporate marketing.
"Our goal is not to exclude anybody, but to have a fabulous relationship with HP," he says.
Nor is NEXTSTEP exclusive in HPís view. The companyís own Motif-based HP-UX will continue to be available at the desktop, as will connections to other client operating systems, including DOS, Windows, and Macintosh. HP is also expected to offer a port of Windows NT to its Precision Architecture (PA) RISC chip.
"We believe in choice," Ernst says. Nevertheless, NEXTSTEP will be HPís preferred object environment in financial services, just as HP is NeXTís enterprise provider of choice. "It is a matter of resources and focus and putting money on the line," Ernst says.
While Object· Enterprise is initially focused on the financial-services industry, it is not limited to that application. NeXTís markets in health care and telecommunications and HPís base in manufacturing and retail industries are also ripe for this technology, but the two companies felt that the alliance would have the most impact if it was tightly targeted on the financial sector, including the securities, banking, and insurance industries.
Object· Enterprise isnít alone in the market, of course. The dominant player in trader workstations is Sun Microsystems, which owns 45 percent of a trading market that exceeded $8 billion in 1992 and is projected to hit $10 billion in 1995, according to Market Intelligence Research Corporation. Together, HP and NeXT controlled about one-third of workstation sales in 1992.
"Sun dominates this part of the financial market," says Peter Vescuso, HPís commercial market development manager, who explains HPís lack of presence by the fact that it didnít offer a RISC workstation until 1991, when it introduced its Series 700. "It wasnít until then that we had a product that could even compete. Weíve had a lot of successes since we introduced our RISC product line."
But despite its high profile, the trading floor represents a very small part of the financial industry. According to Waters Information Services, there are only about 800 trading rooms in North America, totaling about 35,000 trading stations. Worldwide, the figure for traderís and analystís workstations may be 100,000, NeXTís McCrory says.
HPís vision is not just the desktop but the entire enterprise, where HP is already the leading supplier of both RISC systems and UNIX software, according to Ernst. "Sun is not the competitor we run into at the enterprise level," she says.
The big payoff for Object· Enterprise is the opportunity to connect the tools on the traderís desktop to midoffice databases and back-office production systems, as well as make trading information available to branch-office customer-service operations and even to commercial customers. Now instead of tens of thousands of seats, Object· Enterprise is targeting millions of prospective users.
The key, as always with NEXTSTEP, is application development. By tying its development tools and user environment to HPís scalable hardware architecture, Object· Enterprise will permit deployment of NEXTSTEP custom applications throughout a business.
While much of the focus of the HP deal is on the port of NEXTSTEP to PA-RISC workstations, the biggest benefits will derive from NeXTís Portable Distributed Object (PDO) system running on HP servers. Formerly, NEXTSTEP objects had to reside in a single application on a single machine. With the Distributed Object System introduced in NEXTSTEP 3.0, applications could include objects residing on remote NEXTSTEP machines over a network. With PDO, objects can be stored on non-NEXTSTEP servers (see the sidebar, "NeXTís Portable Distributed Object System").
This permits systems to keep many virtual objects alive at all times, even when the application that spawned it goes away. If a new object is put into an app, it can be brought to life without a recompile. Before PDO, Swiss Bank had to cobble together a "black wall" of NeXT machines to act as an object store. "It got the job done, but what we really needed were server-class machines," Koop says.
Thatís where HPís hardware comes in Ė a complete family of workstations, servers, and mainframe-class systems, all built around the PA-RISC processors (see the sidebar, "HPís PA-RISC Servers and Workstations"). Customers can begin deploying Series 800 servers with run-time PDO in the fourth quarter of 1992. NEXTSTEP as a whole will be available on Series 700 workstations by mid-1993.
The NEXTSTEP and PDO ports will support the Open Software Foundation standards for distributed computing, as well as the Object Management Groupís CORBA standard. "We support the standards that make sense. The message is that Object· Enterprise interoperates with, and adds value to, open systems," McCrory says.
Object· Enterprise may even help set future standards for full object interoperability, a step beyond CORBA. I believe that object standards will be driven by those who really do something, not just put the paper spec in place," says Ernst.
With Object· Enterprise, and unlike some other industry alliances that provide little of substance, HP and NeXT are committed to doing something for customers. So far, the two companies say, the integration is working well. "As our people have started working with NeXT, they are impressed with the quality of the technology and the capabilities that are there. That creates a very solid base of mutual respect," Ernst says.
The extent to which Object· Enterprise becomes a core message for HP remains unknown. "We are a piece of HPís story. It is not yet clear if it is a big or small piece. It is critical for us to integrate into their strategy," NeXTís Weissman says.
With HP as a partner, NeXT has credibility that is lacked alone. Last year, NeXT fought tooth and nail for sales to major financial firms such as J.P. Morgan and too often came up empty. Now potential buyers perceive less risk.
Just as financial firms manage risks in interest or exchange rates, they also manage risk with their technology portfolio. With Object· Enterprise, the risk is modest while the potential gains are immense. Thereís that equation again: NEXTSTEP objects plus HP enterprise equals future profits. And customers like Dwight Koop couldnít be happier: "Every month we deploy we make money. Itís a no-lose situation."
Homer on the Range
By John Perry Barlow
About five years ago, when I left the cattle business and took up such smoke shoveling as you find me engaged in here, I was intrigued by the myth of Steve Jobs, and the drama of his dealings with John Sculley. I thought that their story might tell us something about the future of our collective endeavor.
Of course, I didnít know beans about either one of them, but there was something about their caricatures, the barefoot visionary and the thin-lipped Prince of Sugar Water, that rattled something deep inside me. I pumped them so full of my own imaginative gases that their struggles came to seem Homeric. Steve was Achilles, sulking outside the walls of Apple, his heroic hearth, though shredded by the cold claws of organizational efficiency, still beating strong with the idea that the personal computer might free modern workers from the corporate bondage.
I thought of Apple under Jobs as being somewhere between a cybertribe and a techno-commune, literally dreaming up tools that embodied a sense of mission far more potent than mere productivity solutions. I arrogantly dismissed Sculley as a tool himself, a brittle bottom-liner who would make Apple into another bland corporate engine.
I shudder at these cartoon images now. In the years since I sketched them as mythic characters, Iíve gotten to know Jobs and Sculley as human beings. They are both complex and interesting people, each large in intellect and imperfection. Jobs is far more expedient than I once thought, and Sculley turns out to be a genuine visionary, despite a personal delivery that can seem about as spell binding as golf on television.
Their old epic, once so compelling to me, returned to mind the other day when I heard that Sculley had resigned as Appleís CEO. Into his place diesels Eurodroid Michael Spindler, a man whose unsentimental management style will make Sculley look like Percy Bysshe Shelley in no time. I cannot imagine Spindler dreaming about anything but victory.
This news landed like a flat stone on mud. There was nothing in it to inspire imagination; it was just something that happened. I realized suddenly how little room there is for myth, or drama, or even dreams in the computer business as itís once again come to be.
Apple differs from other large California-based companies only in being a harsher place to work than most. It is less interested in changing the world that remaining in it. There remains something heroic about both NeXT and Jobs, but only in the sense that they are running head-on at mean and mighty Microsoft. Even there I have a hard time getting my heart above a resting pulse. What, outside of survival, is the point?
If NeXT survives, which I now believe it can do, itís hard to imagine that the world will be better or even visibly different. Iím as likely to get emotional about NEXTSTEP as I would be over the future of any number of Smalltalk development environments.
Over the long run, I wonder if either company can survive long without that almost religious sense of attachment their early customers brought to them. Buying a Macintosh or a NeXT was once a statement of devout personal belief against which people were willing to put out some additional cash. Without that poetry, what remains to set Apple apart from Dell, or NeXT apart from the Santa Cruz Operation?
Everything changes, I know. But it seems that the computer business has lost its heart. There arenít many stories worth hearing or telling these days. But maybe thatís the way business is supposed to be.