An oldie but goodie
GATEWAY | FRONT PAGE | CITY | SPORTS | BUSINESS & TECHNOLOGY | NATIONAL | WORLD | OPINION
ENTERTAINMENT | STOCKS/MUTUALS | INTERNET | ABOUT US | CLASSIFIED | SEARCH/ARCHIVES
Business/Technology
Local Business
High Tech Report
High Tech Semi-Annual
National & International
Stocks/Mutual Funds
Your portfolio
Ad Information
Get Our Media Kit
Sunday, September 5, 1999
LOCAL BUSINESS
Monday 9 November 1998
Random Excess
Common sense got lost in rush to embrace Java
The Ottawa Citizen
Rod MacIvor, for The Ottawa Citizen / Eid Eid on Corel's plan to develop applications for Java: 'Why spend the effort to be number two when we might be number one?'
AP Photo / Scott McNealy, of Sun Microsystems, put Corel and Michael Cowpland on a pedestal.
The following excerpt, dealing with the birth of Corel's Java strategy in late 1995, highlights both Mr. Cowpland's ambition to be a global software titan, and his propensity for taking huge -- often ill-advised -- risks in pursuing this ambition.
It wasn't long before Corel joined the squad of Java cheerleaders. At the outset, however, Cowpland's engineers saw Sun's new programming language mainly as a tool for adding animation capabilities and enhanced formatting features to the company's planned stable of Web publishing applications. "A number of engineers were playing around with it," Chris Biber says. "It was eventually decided that we could put together a fairly interesting addition to our products by doing Java animations. They started to work on it, and in about three months they had a demonstratable version, which we called Barista." In simple terms, Barista was a small program that could be used to translate text and graphics into Java applets so they could be displayed on any Web browser that incorporated a JVM. Combined with Ventura, it allowed publishers to put documents up on the Web with much richer formatting than was possible with HTML, in essence by displaying a snapshot of the original document based entirely on Java. (Many Web site developers were already doing something similar with Adobe Acrobat, but the drawback was that files written for Acrobat could be read only with a special program known as a plugin that had to be downloaded from the Internet and installed on the client computer's hard drive. With Java, the plug-in -- or JVM, was already built right in to the browser.)
By the time Corel's engineers had come up with a working version of Barista, word of the project had reached Sun Microsystems. Sun was anxious to prove to the rest of the industry that Java was more than just a toy, so it sent two representatives of its Javasoft subsidiary to Ottawa to find out what Corel was up to and see what, if anything, Sun could do to help out. The two visitors were Lou Tucker, Javasoft's director of corporate relations, and Nazila Alasti, an independent software consultant who had been hired by Sun as a Java evangelist. In effect, her job was to convince software developers to support Java by writing applications for it -- the more the better.
In late March 1996, Tucker and Alasti had a fateful, two-hour meeting at Corel's head office with Cowpland, Eid Eid, Carey Stanton, Chris Biber and Vincent Lin, the engineer who had first taken an interest in Java and had led the work on Barista. Tucker and Alasti started by congratulating Cowpland and his staff on their work, which, in only a few short months, had established Corel as a leader in the rapidly growing community of Java developers. Then they got down to business. Now that Corel had acquired some experience with the programming language, Tucker asked, why not put it to good use by creating a full-blown business application in Java? Indeed, since Corel had only just purchased WordPerfect, the obvious next step would be to write a complete Java office suite, a software package that could be downloaded from the Internet or distributed over a local area network and run on virtually any kind of computer. If it worked, it would not only place Corel at the forefront of the hottest trend in computing, it would give Cowpland a clear shot at overtaking Microsoft in the market for productivity software. The likely reward would be billions of dollars in revenue, not to mention the distinction of being the first man to humble Bill Gates. "The message from Sun was that it would really be in our best interests to move this way, to counter Microsoft," Carey Stanton recalls. "They told us, 'What would really make Java go would be if you guys had an application for it, and if you do that we'll commit to giving you the spotlight.'"
Cowpland was intrigued by the possibilities, to say the least. Despite his public bravado about overtaking Microsoft in traditional Windows office suites, he knew as well as anyone that the chances of that happening were between nil and nonexistent. Microsoft's position in the office-suite market was similar to Corel's in the Windows graphics market: having established a strong user base, it was practically unassailable. Corel's only real hope was to redefine the battle, to open a new front, as it were. In the same way that Microsoft had trounced WordPerfect when the PC world migrated from DOS to Windows, Corel would have a chance to beat Microsoft if and when Java became accepted as the new computing standard. "At the time we thought 'Why should we be trying to fight Microsoft where we do not have a level playing field?'" Eid Eid says. "Here was an opportunity with a whole new platform, and we could be first in developing apps for that market. Why spend the effort to be number two when we might be number one?"
Soon after the meeting with Tucker and Alasti, Cowpland convened another brainstorming session with his senior engineers. Vincent Lin, it turned out, had already done some preliminary work on new Java tools for Corel applications. One was a small charting program; another was a Java-based interface for a spreadsheet program that would run over a network. After being briefed on those projects, Cowpland asked Lin if the idea of a Java office suite really was feasible. According to Cowpland, Lin went away and thought about it for a while, then came back and said yes. The resulting software package would clearly not have all of the power and features of a regular Windows suite, for the simple reason that a Java rewrite of WordPerfect would be too big and bulky to operate efficiently over a network. Nevertheless, Lin said, the basic concept was perfectly sound.
That was all the encouragement Cowpland needed. Before the end of the day, he gave Lin the go-ahead to design a working prototype of what would be called Corel Office JV, and later Office for Java. Cowpland wanted the prototype ready to show at Sun's inaugural Java developers conference, JavaOne, which was scheduled to take place in San Francisco's Moscone Convention Center from May 29 to 31. Lin would have a team of about a dozen developers at his disposal, and all the pizza and Cokes they could consume. With less than two months to go before the unveiling, there were going to be plenty of all-nighters.
"Basically," Cowpland recalls, "I said, let's put some resources into this and see how far it could go. And the more we did it, the better it looked."
Down the road, when it became obvious that Office for Java had become a colossal waste of money and manpower, almost everyone connected with the project would try to shift blame to someone else. The real problem was that, in the rush to embrace a promising new concept, nobody ever stood back and asked three simple questions. First, what exactly was Corel trying to achieve? Second, if it worked, would there be a market for it? And third, was the underlying technology at a stage where it could support such an ambitious undertaking?
The answer to the first question might have seemed obvious, but it wasn't. There were, in fact, several different agendas at play. Even at that early stage, the objective in Cowpland's mind was to create a marketable application, and get it out there as soon as possible before anyone else had the same idea. The guys in engineering were equally enthusiastic, but for another reason: like most software developers, they got their kicks out of working with new technology, and Sun's Java language was far and away the hottest thing in the business. To some of them, at least, Corel Office JV was really more of an experiment, a "proof of concept" rather than the first step toward a viable product. Whether it actually ever brought in any revenue was beside the point. "I never had the intention of developing a full suite," Lin insisted later. "The whole department was clear on this."
According to Chris Biber, the aura of excitement surrounding Java clouded everybody's judgment. "At that point it wasn't hashed out what it would really mean. The idea was to see what an office suite in Java would look like, and the verdict from on high was to see if we could have something done by May of that year. It crossed everyone's mind that it was a very ambitious undertaking, but at the same time there was general euphoria -- and not just at Corel -- that Java was indeed going places and was going to make a huge impact in the market. What was undertaken was seen as an engineering challenge more than anything else, to put together a working prototype in an incredible short period of time."
To Lin's credit, he and his skunk works team pulled it off. On May 30, 1996, a day after the launch of WordPerfect Suite 7 in Salt Lake City, Mike and Marlen Cowpland flew into San Francisco, accompanied by Paul Skillen, Eid Eid, Carey Stanton and Chris Biber. They carried with them a three-and-a-half-inch floppy-disk on which was stored a copy of Lin's Office JV prototype, a barebones (it occupied less than a megabyte) demo of what Corel's PR staff were touting as a fully functional Java office suite. The demo version couldn't actually be used to do anything, but it was good enough to convince most developers that the idea held merit.
If Cowpland had harboured any remaining doubts about the promise of Office for Java, they would have been erased by his experiences at JavaOne. Sun's executives were so impressed by Corel's little demo application that they treated Cowpland and his group as honoured guests, seating them at the front of the convention hall. At a news conference arranged to discuss Corel's Java strategy, Scott McNealy himself dropped by and introduced Cowpland to the press. Cowpland basked in all the attention and was amazed by the turnout for the conference. The organizers had expected about 2,000 developers to show up, but instead there was a capacity crowd of more than 7,000 -- incontrovertible proof that Java had caught the industry's imagination. Analysts and Silicon Valley hotshots who had never even thought of Corel before were milling around the company's display like groupies at a rock concert.
"To me JavaOne was the real turning point in terms of our strategic focus," Stanton says. "Before that trip, Mike was interested in seeing what Java could do, but there wasn't really a full-scale commitment. All of a sudden we were being escorted to the front row at the Moscone Center and it's like, 'Holy shit, look at all these people -- this is bigger than a Windows developers conference!' And we were treated like stars. The Corel logo was flashing up there on stage on the big video screen, we were being written up in PC Week É What it came down to was that we were getting more mileage out of this little Java project than we got from the launch of WordPerfect 7. So right then and there, you could see there was a dramatic shift in our thinking. In Mike's mind it was, 'OK, boys, we're going with Java.'"
Biber, too, got swept up in the excitement of JavaOne: "Without overstating it, Corel was literally the sensation of the show. Up to then Java had been used mainly for things like stock-tickers and other types of animation on the Web --cutesy little things. And here was a fairly well-recognized company saying, 'We can do this,' and people went, 'Wow!' They saw a running spreadsheet and a running word processor. They saw a running business graphics app, which definitely caused a huge stir."
As soon as he returned to Ottawa, Cowpland gave the authorization to turn Office JV into a commercial product, the world's first integrated Java-based application package. Almost overnight, Lin's team quadrupled from about a dozen developers to 50. An article in the June 10 issue of the Seybold Report on Desktop Publishing said that the first component of the suite would be WordPerfect 7, followed by Java-powered versions of the Quattro Pro spreadsheet and CorelCHART. "Essentially, Corel has rewritten these applications in the Java language," the newsletter said, "thereby picking up both platform independence and network access. A beta version of Corel Office JV will be ready by the end of 1996, by which time we believe that many of Java's performance problems will have been solved." The same article noted that Java, in its initial release, suffered from prolonged downloading times, problems with cross-platform compatibility and occasional security glitches, but that all of those flaws were in the process of being corrected. "All in all, by the time the snow flies, we can expect Java to be much faster, cleaner and safer."
It all sounded quite straightforward, but the view from the trenches was nowhere near as encouraging. The ugly truth was that Java, despite all the hype, was still an exceedingly rudimentary medium, somewhat like the early versions of Windows, only worse. The difference was that Microsoft had, by the fall of 1987, assembled a wellstocked, if poorly documented, library of authoring tools -- sections of code that help developers and designers build other software programs by performing routine tasks, such as opening a dialog box on the screen or activating the program's search engine. In the case of Java, most of those tools had not even been designed yet. Rather than relying on Sun for most of the primary authoring tools, Corel's engineers would have to write most of the basic code themselves, practically doubling their workload.
The developer who was given responsibility for creating those tools was Claude Montpetit, a talented senior programmer who had been reassigned to Lin's group after his own Windows-based project had been cancelled in the wake of the WordPerfect acquisition. "It was right after the JavaOne announcement," Montpetit explained to Electrical Engineering Times. "I was just back from a week's vacation and we had to come up with a whole design and develop everything from scratch. It was a huge task." Added Montpetit: "No one had a 'mind' for Java. We all came from a background in using Microsoft Foundation Classes (a set of developers' tools for Windows). We thought there was a need to develop a similar framework for Java so we could share all the Office components. In Java, everything was very basic." With the help of several other engineers, Montpetit created what became known as the Corel Application Framework, a basic library of routines on which any application could depend. As soon as CAF was ready, other groups began writing applications on top of it.
By the fall, the project had progressed far enough that Corel decided to post a preview version of Office for Java -- the company described it as a "pre-beta" release -- on its Web site. That way, Java devotees and the merely curious could examine and play with it for themselves. The early feedback was by and large favourable, which wasn't a surprise considering that only confirmed technology enthusiasts were likely to take the trouble to download the product and run it on their desktops. Consultant Geoffrey Moore defines such people as "the gatekeepers for any new technology," adding that because they appreciate technology for its own sake they represent an ideal testing ground for new products. "They are the ones who will spend hours trying to get products to work that, in all conscience, never should have been shipped in the first place," Moore wrote in his 1991 book Crossing the Chasm. "They will forgive ghastly documentation, horrendously slow performance, ludicrous omissions in functionality, and bizarrely obtuse methods of invoking some needed function all in the name of moving technology forward."
The big problem with technology enthusiasts, as Moore himself is quick to point out, is that their love of innovation frequently blinds them to serious flaws in design and execution. To some degree, that's what happened to Office for Java. The techies were so bowled over by the concept that they were slow to appreciate the product's major shortcomings.
A case in point was an article posted on the Internet on Nov. 4, 1996, by Michael Cullison, a columnist for an Ohio-based online magazine. "Imagine running WordPerfect or Quattro Pro on any platform that has a Java-enabled browser," Cullison wrote after downloading Corel's new office suite. "Check this one out. It requires a pretty high-speed connection and some patience, but its worth the wait just to say you were one of the first to try a spreadsheet or word processor written in Java."
Around the same time, PC Week ran some tests on the software and published an approving if not entirely positive review. "Corel Corp's Corel Office for Java, the first suite of office applications written entirely in Java, proves that it can be done and paves the way to platform-independent computing É Although containing only a few of the planned features at this point, the product we examined had a working version of the WordPerfect word processor and Quattro Pro spreadsheet and gave a good indication of the suite's potential." On the downside, the magazine noted that Office for Java was sometimes "painfully slow" when running in a standard Web browser, and that many of the program's functions, including the ability to save and print documents, were not working.
Clearly, an enormous amount of work remained to be done. Yet Cowpland's public pronouncements were unfailingly upbeat. He repeatedly promised that Office for Java would be ready to ship in the first quarter of 1997, backed by a "multimillion-dollar" ad campaign. And at Comdex that fall, Cowpland declared Corel's intention to enter the hardware business by launching a network computer with enhanced video capabilities called the Corel Video NC. In effect, Cowpland had decided to rip the guts out of Corel's money-losing video-conferencing unit and reposition it as a stripped-down computer suitable for corporate networks. Realistically, Cowpland knew he could never expect to make much money from hardware, but he hoped that the availability of such devices -- similar machines, known in the industry as "thin clients," were beginning to appear from companies such as IBM, Oracle and Sun -- would benefit Corel by stimulating demand for Java-enabled software.
All this time, however, Corel still hadn't stopped to consider whether its goal of an integrated office suite was the right one for the dawning era of network-centric computing. If office workers throughout North America were going to start accessing their key business apps from some central server, those programs would have to be as light and nimble as possible to avoid lengthy downloads and sluggish performance. Yet the further along Corel got with Office for Java, the larger the application became, an monolithic Windows office suite, with all of the standard bells and whistles. From inside and outside the company, there was pressure to add all of the functions of the existing WordPerfect suite, as well as file compatibility with a wide range of popular PC programs so that users could import documents and work on them in Java.
"Collectively, I think we all got carried away with the fact that we had already shown something working," Biber says. "Especially because the press and crowd reaction was so incredible. Rather than stepping back, reviewing our experiences and starting again from ground zero, we just charged ahead with the same fundamental architecture. And what made it worse was that we had a number of pilot users and early adopters saying to us, 'This is great -- now, can you add this feature and this feature? Can you do a spell-checker, and how about something like revision control?' Everybody in the organization got swept away by that. So over the longer term, what was beginning to emerge was exactly what we had been trying all along to avoid, namely a rewrite of WordPerfect."
Corel's developers had to contend not only with the increasing scale and complexity of the project but also with changing technology. In 1997, Sun introduced a new Java development kit, JDK 1.1, that was better than the original version but also very different in the way its components interacted with one another. Corel was among the first major software companies to begin using JDK 1.1, but instead of making things easier, the switch necessitated an enormous amount of recoding for little perceived gain. Worse, the early releases of 1.1 were maddeningly unstable. Looking back, Biber believes it was a mistake for Corel to have jumped on JDK 1.1 when it did. It would have been better and far cheaper, he says, to have waited several more months while others discovered all the bugs.
Having been widely hailed as one of the pioneers of Java, however, the last thing Mike Cowpland wanted was to slow down and allow his competitors time to catch up. Like everyone else in the Java universe, he knew that IBM'S Lotus Development division was hard at work on its own Java suite, code-named Kona, which was due out in September 1997. (In contrast to Corel's all-in-one strategy, Lotus's efforts were aimed at creating several smaller Java apps to handle basic functions such as e-mail and database connectivity.) In addition, there were persistent rumours that Microsoft itself was working on a collection of Java business applications. If those reports were true -- Gates and his officials would neither confirm nor deny the stories -- Corel had not a moment to lose. It was crucial to continue moving forward at full speed, both to avoid being overtaken and to maintain Corel's image as a leading-edge Java developer. "Being first to market with one hundred per cent Java software has, if nothing else, gained the company incredible mindshare," reporter John Spooner observed in Marketing Computers magazine. "Cowpland has the attention of the press, analysts, Java developers, even companies such as Sun Microsystems that never even thought of Corel before. É What Larry Ellison (the chairman of Oracle and a vocal proponent of network computers) has mostly talked about, Mike Cowpland seems on the verge of delivering."
It was largely to maintain that aura of leadership that Corel posted an updated version of Office for Java on its Web site on April 2, 1997, coinciding with the start of the second annual JavaOne developers' conference in San Francisco. Speaking to reporters at the conference, Cowpland boasted that Office for Java was the first such application to be certified by Sun as "100 per cent Java pure," and he brashly predicted that it and other Java-related products would generate $40 million in revenue for the company that fiscal year. "We're out of the gate," he added. "It does establish us as the front-runner in Java applications." For good measure, he announced that Corel's goal was to become the premier Java applications provider, the "Microsoft of Java," as he put it.
Unfortunately for Cowpland, much of the novelty of Office for Java had worn off by that stage. Out in the real world, testers were beginning to judge the product not by its potential but by its practical utility. They were not impressed. Although Corel described the new release as a beta version and promised that the finished product would be ready to ship by the summer, Office for Java still fell well short of the mark. Critics panned it as being too slow, too big -- the portion of the software designed to run on a network computer had ballooned to almost 10 megabytes -- and rife with bugs. "I think this was a case of shoot first and then aim," said Jeffrey Tarter, editor of Softletter, a computer trends newsletter based in Watertown, Massachusetts. "It's not clear why anyone would want a Java version of Office."
Finally, after a year of hype, the reality of Java development was starting to sink in at Corel. As a programming language, Sun's creation held great promise, but it was still evolving and clearly not up to the challenge set by Cowpland. What's more, Corel had erred badly in trying to use Java to recreate a monolithic office suite loaded with features and built on tons of code. It was undoubtedly true that Java represented a significant advance in computer programming, but to get the most out of it would require a new, more modular approach to software development. In some ways, the learning curve for Java recalled the pioneering days of television in the 1940s and 1950s. Because few people back then understood the full potential of the new medium, most of the early programs were really just radio with pictures. Only later did producers and directors figure out how best to exploit the technology's visual dimension. Similarly, the engineers at Corel -- not to mention their bosses -- were slow to appreciate the unique properties of Java, and the extent to which they demanded a rethinking of traditional software design. "There was too much code," Montpetit complained, referring to the bloated beta version of Office for Java. "lt was too big and it took too long to download."
Inside the company, there were growing tensions over the future of the project. On paper, the 50-member Office for Java development team in Ottawa reported to Vincent Lin, who in turn answered to Paul Skillen at WordPerfect headquarters in Orem, Utah, several thousand kilometres to the southwest. But geography wasn't the only thing separating the two men. Lin complained that the word-processing software for the suite, which was being Java-coded by a separate team of WordPerfect veterans under Skillen's direction in Utah, was poorly written and so slow that it made the rest of the project look worse than it was. Meanwhile, Skillen, who in March had succeeded Eid Eid as vice-president of engineering, was becoming increasingly convinced that Lin's whole approach to Java architecture was wrong. "Paul eventually convinced Mike that we needed to change direction," says Carey Stanton. Rather than designing a full suite, Skillen favoured a new, network-friendly paradigm known as "distributed computing," in which applications are broken up into collections of small components that can be shipped quickly and efficiently across a network and reassembled as needed on the user's desktop. "The analogy I like is that we had been trying to stuff an elephant down a garden hose," says Stanton. "With distributed computing, it's more like a bunch of ants."
The growing animosity between Skillen and Lin finally came to a head in the summer. On June 30, Corel put out a press release announcing that it was centralizing development of its Java office suite in Orem. "We're talking about cranking up a team in Utah to be a Java powerhouse," Cowpland said. A Corel spokeswoman explained that the decision meant that 22 positions would be transferred to Orem. Oddly, however, none of Corel's Ottawa-based Java engineers were being asked to relocate. Instead, the Orem jobs would be filled by existing WordPerfect staff and some new employees.
The newspapers reported it as a straightforward reorganization, but in fact the announcement was the first hint of an abrupt change in direction for Corel's Java strategy. Vincent Lin, who only two months earlier had been profiled in The Ottawa Citizen as the driving force behind that strategy, was being pushed out. And with Skillen now firmly in control of the project, Office for Java was for all intents and purposes dead.
Copyright 1998 Ottawa Citizen
GATEWAY | FRONT PAGE | CITY | SPORTS | BUSINESS & TECHNOLOGY | NATIONAL | WORLD | OPINION
ENTERTAINMENT | STOCKS/MUTUALS | INTERNET | ABOUT US | CLASSIFIED | SEARCH/ARCHIVES |
