So What Else is New?

In the latest fallout from the Sun/Microsoft Java divorce case, Microsoft has declined to bundle Java Virtual Machines (JVMs) with the upcoming Windows XP operating system. Instead, users accessing web pages with Java in them will be greeted with blanks in parts of the screen, and security warnings if any Java program tries to execute. If you need a JVM, you’ll either have to download it from Microsoft or hopefully get it bundled by your PC manufacturer.

Although the move appears a body blow to Java, in reality, nothing has changed. Java lost the battle for the client years ago, and not just because of Microsoft’s FUD campaign. Users rejected Java on the client because downloading applets and running them added too much computing overhead. Give us Pure HTML. Please.

OK, maybe there might be some inconvenience to users with JavaScript, which Also uses JVMs. Otherwise, the JVM spat is much ado about nothing. That’s because of one thing that Microsoft and Sun actually agree on: In the future, the server will be where the action is. Microsoft has conceded as much by promoting its emerging .NET software-by-Internet-subscription licenses.

Instead, the real battle for the hearts and minds of application development groups and technology vendors will come when the .NET technology framework emerges (probably mid Q4), providing a choice between Java’s write once, run (almost) anywhere vs. Microsoft’s write-in-any-language, run-on-Windows model. At that point, Microsoft will offer a technology answer whose versatility might give Java real competition at last.

The Internet is Dead, Long Live the Internet

Given the carnage in e-tailing lately, the passage of WebVan looks like just another footnote to the dot-bomb revolution. But forget for a moment that the venture burned through more than $800 million of other people’s money in barely a couple short years. If that’s possible, you might find the episode almost laughable in its disregard for Business Economics 101.

Here’s the story: the head of one of the world’s largest management consultancies leaves to head a startup that is essentially a glorified trucking service. And then the venture draws a market valuation north of $10 billion just because of its dot com moniker.

Forget the dot com part, WebVan was as brick and mortar as you could get. Maybe they didn’t open stores, but they bought or built just about everything else. And, owing to the perishable nature of their merchandise, the company had to replicate all this infrastructure in every local market that they entered. So much for the economies of disintermediation.

In pursuit of The Next Big Thing, WebVan and many of its web rivals strove for market penetration at all costs. WebVan didn’t have the chance to get it right in one test market before rolling out nationally. Others made even dumber mistakes, like Webhouse’s (PriceLine’s ill-fated grocery venture) failure to design pricing systems that could differentiate by local markets. So New York shoppers could buy Doritos at Milwaukee prices.

OK, it’s a cheap shot to conclude that The Internet Didn’t Change Everything. But it’s safe to venture that, while e-tailing has its advantages, immunity from the laws of business economics isn’t one of them.

No wonder that “multi-channel integration,” making the shopping experience consistent regardless of whether you’re in the store, phoning a call center, or surfing the web, has become The Next Big Thing for retail. You get the advantages of e-tail without having to build infrastructure-or brand recognition-from scratch. As for me, the customer, click on LLBean.com, but first rush me the print catalog so I can see what that sweater REALLY looks like.

When Giants Tremble

The reversal of the Microsoft breakup ruling surprised next to nobody. Wall Street shrugged, with the stock inching up a few percentage points. The die was cast when Judge Jackson’s critiques of Microsoft after his ruling compromised the appearance of objectivity. Of course there was the DC Appeals Court’s known hostility to antitrust cases, not to mention questions of whether the Bush administration would seriously push the case if the rulings went the other way.

But no court ruling can alter the facts of life: the technology market has consolidated because the market has voted for the Jack Welch philosophy: don’t take a company seriously if it isn’t first or second in its market. Remember, buying hardware or software is like buying equity–the last thing you want is for your vendor to go belly up, and your product orphaned.

And even if Microsoft bullied Netscape, most of the victim’s wounds were self-inflicted. Yes, Microsoft undermined Netscape Navigator, but who could have made a real market in browsers alone? Maybe build a real enterprise business, or bundle in consumer goodies like a music player and legal Napster-like subscription service, and then you might have had something that people would actually pay for. Instead, Netscape let companies like BEA and Real Networks seize the initiative.

Microsoft will likely get its hands slapped during the penalty phase. So what can we learn here? Big remains OK with the courts and the market as well. But big companies are not always safer investments. (Remember DEC?) CA, whose product lines are far less coherent than Microsoft’s, has become the poster child for bloat. Yes, it will probably win its upcoming proxy skirmish, but don’t be surprised if CA has to start shedding weight. Soon.

Keeping the Lid On

Going to any Linux user group meeting is like engaging in time travel. The utopian ideals about placing software in the public domain have clear sixties rings, and on the surface, appear to be pretty flaky ways to develop software. Yet the success of Apache and the recent move by “adults” like IBM to form a group for testing approved Linux distributions has demonstrated that the open source “bazaar” is not so bizarre.

More to the point, open source has placed traditional “good guys” of the open systems world on the defensive. If your software is so open, why not release the source code? The issue has dogged Sun because Java was supposed to be the politically correct answer to the Microsoft juggernaut. Yet the best that Sun would venture was opening Java to a “community process” where it remained first among equals.

Microsoft has also been knocked for a loop by open source. As part of its .NET rollouts, Microsoft recently responded with the “Shared Development Process” (SDP), throwing in Hailstorm services as the first SDP guinea pig. There will actually be three levels of SDP, from review to peer development and arms-length industry working groups. Microsoft is clearly stepping gingerly, in that it has not committed to submitting other technologies (beyond mentioning the possibility of BizTalk industry frameworks) through the new process. But SDP has clear boundaries. Crown jewels like the Windows.NET platform won’t go into anything resembling community development.

Clearly, the success of open source has required the industry to make politically correct noises. But economics, rather than sound bytes, are really driving the quest for community-like alternatives. The seeming chaos of open source has actually proven a rather efficient mechanism for attracting third, fourth, or even fifth-party developers who can transform foundation technologies, such as OSs or application frameworks, into de facto standards.

The bottom line however is that open or community source involves, not just forgoing some level of control, but revenue opportunities as well. In an economic climate where technology vendor margins are strained, the question whether to hold on to the rights to every last piece of intellectual property is no longer so clear.

Boys Will Be Boys

There’s something about the IT industry that causes people-mostly guys-to regress. Go to any Java conference, and you’ll first have to wade through air hockey tables before you can find the exhibit hall or tech sessions. Log onto any IT news group, and it’s easy to mistake yourself inside some locker room.

Or go listen to the grownups of the industry speak.

Late last week, Oracle rolled out 9i, with Larry Ellison spending most of his remarks trash talking IBM. Nobody has ever done clustering because it’s so hard, but 9i’s improved shared-disk architecture performs better than DB2 and is far more fault-tolerant. Showing an SAP ATO (assemble-to-order) benchmark, Oracle claimed 89% scalability on two nodes vs. 34% for DB2 in, and so on. But Oracle concedes that its benchmarks have not yet been officially audited by SAP.

Ellison even tried turning Oracle’s exorbitant pricing into another chance to bash IBM. Yes, Oracle was dumping its hated power unit pricing for CPU, but at $40k/processor for the enterprise version, is still 2x that of DB2 Enterprise Edition. No matter, said Ellison, Oracle throws in things like OLAP, data mining, messaging, security and other features that you have to pay separately for with DB2. Naturally, IBM refutes all this.

But facts aren’t important here. Demos are demos, and any vendor can always skew pricing comparisons. Yes, it’s important that Oracle has improved its clustering. Maybe somebody will finally try this at home. But the real issue is who was Ellison really talking to. DBAs, who trained on a given database, won’t throw out their life’s work lightly. CIOs considering switching must then deal with the mess of replacing their DB2 DBAs with scarce Oracle folks, or vice versa. All this explains why even Sybase has kept its toehold on Wall St.

Database vendors have to try hard to lose their markets. The most important takeaway from last week is that Oracle realized that that was exactly what it was doing with pricing.

Java One 2001: Waiting for the Other Shoe to Drop

Java is becoming a victim of its own success. For now, J2EE has become the only game in town for scalable web applications. Even SAP, which until now was wedded to Microsoft’s web architecture, has bitten the bullet with its own J2EE-based remake of MySAP.com. SAP couldn’t wait for .NET, or for upstarts like BroadVision to steal its thunder B2B-ERP’s next frontier.

But this year’s Java One was more noticeable for what wasn’t there: Real XML integration and real use for products coming from Java’s next frontier, the embedded device. Not surprisingly, most of the questions at Sun’s opening press conference were about its sagging quarterly numbers, not Java.

With .NET knocking on J2EE’s door, Sun is now stealing a page from Microsoft. In the next few months, it will release “service packs” offering goodies such as Java XML binding, rather than wait for the release of J2EE 1.4, now at least 6 – 8 months away.

At least at the enterprise end of Java, there are real stakes to fight about. At the other end of the spectrum, embedded Java, there is little if any agreement on whether J2ME (Java 2 Micro Edition, which was originally aimed at this sector) is adequate, or whether you need the security features of J2SE (Standard Edition, designed for PC clients). Or, whether Sun’s KVM is the right Java Virtual Machine, or whether “clean room” versions from players like HP or Kada are better sized for the job. And finally, can the world do with just the MIDP profile (which specifies all the Java classes), or whether PDAs merit their own “Palm Profile?”

That there is debate isn’t notable. Instead, the operable question is whether there is a market. The proprietary OSs of Palm, RIM, and Pocket PC devices have proven little obstacles to today’s killer hand-held apps: calendar and address book. The benefits of Java on phones won’t matter much until advanced 2.5 or 3G networks proliferate the landscape. The same goes for putting Java or whatever on advanced set top boxes, which await broadband rollout. Even AT&T Broadband, which had signed an almost exclusive agreement with Microsoft, is now hedging its bets.

The message from Java One 2001? Keep deploying J2EE e-business servers. For anything else, wait ’till next year.

Making Web Services Inevitable

IBM has further positioned itself at the epicenter of the XML universe, by being the first player to announce a set of pieces covering everything from the database and application integration through security policy management and, of course, the web application server nerve center.

There is little question that IBM has gotten XML religion. XML, along with Java before it, provide the glue that accomplish using industry standards what Big Blue tried unilaterally a decade ago with SAA. At today’s briefing, IBM’s web services evangelist showed how a page of CICS transaction code could be transformed into a web service, complete with standard XML description language, without the programmers (or business analyst) knowing anything about legacy systems.

If it sounds too easy, it probably is.

Web services have become the latest catchwords in a market that previously had trouble learning how to spell EAI. In the latter case, it was the complexity of trying to package systems integration—which for many enterprises became a code word for bottomless pit. When integrating—or interfacing—an ERP system, developers often got hammered dealing with the unique ways that every application had for structuring data. Integrators learned the hard way that, just because two applications stored their data in Oracle, that didn’t mean that both applications could understand each other’s data, not to mention sharing overlapping processes with one another.

Web services are also being unleashed in a market that learned to spell B2B, but failed to embrace it. The notion that exchanges could improve business efficiency by providing a dating or matchmaking service, pairing any buyer and seller willing to agree on the right price, might have sounded good for buying priceless art or paper clips. However, it clashed against best practices developed over the past two decades, where trading partners actually shrank their lists of trading partners in favor of maintaining fewer, longer-term strategic relationships.

At first glance, web services sound like another step towards allowing anyone to do business with anyone at any time without having to prepare laborious handshakes up front. However, recent B2B experience has proven that most organizations do not want to conduct business like two ships passing through the night. Instead, taking advantage of the ability of web services to dynamically execute on the fly will actually require even more handshaking up front, for combination of business and technical reasons.

Recent B2B experience has taught e-commerce gurus that companies do not want to conduct business like ships passing through the night. Significantly, the web services examples cited by IBM involved making it easier for parties to conduct business from standard menus of transactions in situations where the identity of both parties is known to at least the initiator. This was anything but commerce at random.

There are good reasons for this.

Yes, XML standards appear easy and intuitive at first glance. They use the same ASCII characters as HTML, are easily parsed, and the standards have gathered virtually unanimous vendor acceptance. But there remain formidable barriers before XML-based service requests can magically connect with service providers to make e commerce truly frictionless.

Let’s start with technology. XML data, which is much fatter than conventional structured data or binary code, could add significant overhead to database processing and messaging communications unless some standard techniques are developed to compress or abbreviate it. Today, IBM, which now stores XML data as another DB2 Extender data type, offers the solution of federated databases that are deployed at any point along the network where the processing can be conducted most efficiently. However, when IBM rolled out an example of one of its customers—the Galileo travel aggregator, they talked about an enterprise that handles over 100 billion transactions annually. Just imagine if all those transactions all were conducted in XML.

Then look at the business side of the equation. Having a service description language does not necessarily mean that everyone in the same industry shares the same syntax. In the electronics industry, RosettaNet has begun the process of standardizing terminology and transactions, updating a process that many verticals conducted with EDI back in the 1980s. So more RosettaNets are needed in other sectors to get everybody talking the same language.

More RosettaNets must emerge in sectors outside semiconductors, for starters.

IBM, along with Microsoft and Ariba, have played a leading role in lining up web services as today’s number one e commerce IT agenda item. Are all the standards in place? Maybe, maybe not. According to IBM Internet technology director Rod Smith, maybe we need for yet a new set of APIs that define how to characterize the “endpoint,” where an XML web service transaction ends and a legacy system process begins.

But the emergence off standards is only the first step. To “do” web services, major verticals must spend the effort to define just what exactly it is that they do. As ERP experience taught many organizations, that could take some time.

IBM Buys Informix

Make no mistake about it, IBM’s $1 billion cash purchase of Informix was strictly a play about market share and nothing else. Citing Dataquest and IDC numbers, IBM claimed that the acquisition of Informix would double IBM’s share in what analysts termed the “distributed database” market, which is another way for saying UNIX and NT

Throughout their conference calls with press and analysts, IBM’s Janet Perna, who heads IBM’s Data Management Solutions group, repeatedly made references to the database “war” with Oracle. “We are making great strides,” she noted, citing numbers such as IBM’s 36% growth in this market space, vs. Oracle’s 6% last quarter. As if there were any doubts whom this was aimed at, Perna dismissed Microsoft as a “one-trick pony” in the market since SQL Server is limited to Windows platforms

Prior to the deal, the database market was already consolidating down to three main players: IBM, Oracle, and Microsoft. Once-mighty Sybase and Informix, along with CA-Ingres, the earliest also-ran, have effectively become peripheral players. In effect, the IBM/Informix deal changes few dynamics in the marketplace. Where there were previously three and two-quarters players, we’re now down to three and a quarter

Instead the deal looks like a cash-rich salvage operation. IBM, whose latest quarterlies have shined in contrast to most other tech players, could easily afford a billion dollars of pocket change to buy installed base. IBM has pledged to continue supporting Informix databases, but firmly stated that new development would continue to center on DB2. Yes, IBM will cherry pick some Informix technologies, including selected DataBlades, but their new home will be DB2

The message to Informix customers? Should you consider migrating to a more “mainstream” database that has the most current technology, the obvious path will be DB2, not Oracle. Given that Informix has lacked the market share and resources to keep up with the Joneses—its Arrowhead effort to converge relational, object, and analytic database technologies was years behind IBM’s—the Informix installed base has been ripe for the taking. And, given that database migrations are never trivial—if you replace the database, you may as well replace (or heavily retrain) the DBAs—IBM is going out of its way to pre-empt inroads from Oracle

As for what’s left of Informix, Ascential, it bears striking resemblance to the old Ardent. Except that it is now $1 billion richer, and the proud bearer of a juicy deal for IBM to codevelop and resell its products. Guess who’s smiling now?

IBM WebSphere Finally Supports Mainframe

It’s little surprise that a key ingredient of IBM WebSphere’s success has been its support of existing IBM technologies—especially DB2 databases, plus the availability of bodies from IBM Global Services which has built an e-business practice around WebSphere. If you’re a confirmed IBM shop, WebSphere has long been your logical choice for web-enabling legacy applications.

But until now, running WebSphere on the mainframe was essentially like running a port of an open systems tool in a legacy environment. WebSphere 4.0 for z/OS and OS/390, released this week, adds native support for core mainframe services such as CICS, the transaction monitor that IBM claims runs 40 billion transactions every day. Specifically, WebSphere 4.0 allows Java application developers to plug into these transaction services without having to write custom code. WebSphere 4.0 does the same thing with other established mainframe building blocks such as Parallel Sysplex; IBM’s RACF security and access control programs; and WLM, IBM’s workload manager used for load balancing.

Version 4.0 is also the first release of WebSphere that is officially J2EE (Java 2 Enterprise Edition)-compliant. Not only that, but it’s the “deepest and broadest” implementation of J2EE, claimed Scott Hebner, middleware marketing director for IBM Software. He said that IBM passed more J2EE compliance tests than any other appserver vendor, including all the mandatory tests and 70% of the optional ones run by Sun. For instance, IBM claims that its implementation of JMS (Java Messaging Service, one of the J2EE standards) is more mature than BEA’s, its primary rival in the appserver space. For instance, IBM claims that BEA WebLogic lacks the ability to import transactions originating as MQSeries messages (IBM’s market-leading messaging middleware). BEA was not available for comment at press time.

To place matters into perspective, IBM’s J2EE fervor is rather recent. They admit to being slower in becoming J2EE-compliant than the other Java appserver rivals. “We took a balanced approach to supporting standards since we originally did not believe that J2EE was the only [important] platform,” said Hebner. He claimed that other standards, such as XML-based UDDI and SOAP, were equally important (see adjacent story). Hebner added that WebSphere supported all the J2EE essentials anyway. “We supported EJB 1.1 ‘minus’,” he said. For instance, while IBM supported essential EJB features such as session beans, until now it didn’t support peripheral ones such as XML descriptors. “Our clients weren’t using them [XML descriptors] anyway,” he maintained.

The WebSphere 4.0 announcements were accompanied by an important upgrade to Visual Age, IBM’s umbrella IDE (integrated development environment) used for developing many of the applications that run on WebSphere. For the first time, IBM is finally offering Visual Age as a suite, bundling all languages including C, C++, COBOL, and Java into the same package. In so doing, IBM is following the lead of other tools vendors, such as Microsoft, Rational, Sybase—and just recently, CA.

CA Throws Hat in Development Suites

Application life cycle tools were an afterthought of Computer Associates’ 1999 Platinum acquisition. Ironically, thanks to its acquisitions of Platinum technologies, Sterling Software, and Nantucket (the latter, over a decade ago), CA has accumulated one of the largest application development product portfolios in the business.

Not that anybody noticed—CA included. With the sole exception of Nantucket, none of these acquisitions were driven by application development tools. Platinum was bought for its DB2 tools, while Sterling was acquired for its storage management offerings, both of which neatly complemented CA’s data center products.

This week, CA raised the profile of its application development business with the release of ERwin Modeling Suite 4.0. Don’t let the version number of the product name fool you, this is actually CA’s first attempt to bundle development tools. It’s just that ERwin and Paradigm Plus, the product families included in the suite, both happen to be on their 4.0 release cycles, according to application life cycle brand manager Gregory Clancy. The result is CA’s first serious challenge to market leader Rational.

ERwin modeling suite contains four products, headlined by the ERwin data modeler, the market leader for database designers; plus BPwin, a business process modeler; Paradigm Plus, a UML modeling tool; and the almost brand-new ERwin Examiner, a tool which checks data models for inconsistencies. The products are also sold separately.

Each of these products has significant enhancements. ERwin sports new capabilities for separating logical and physical models. The logical model organizes data into entities, such as “customer” or “product order,” while the physical model maps them into the table and column data structures of relational databases. By separating logical from physical, data models can be spun off for multiple databases, a feature that is useful if an organization has different databases (e.g., Oracle and DB2) and wants to ensure that overlapping data uses the same data structures. Additionally, ERwin has added new model viewing capabilities that provide hierarchical views of data models, a feature that makes it easier to work with larger models.

Other product enhancements include Paradigm Plus’ support of UML 1.3 (the latest version of the object modeling language), XML-based round-trip engineering, and the ability to extend the tool using well-known VBScript or JavaScript. Meanwhile BPwin has added organization chart views and the ability to import bit map graphics—but does not support direct import of Microsoft Visio charts, which many organizations use for illustrating process flows.

So how do these tools integrate? CA is relying on XML to provide an open interface, and for translating data models into UML models, wizards are provided. This differs from Rational’s proprietary hooks, although Rational’s integration is far more ambitious, spanning object modeling, data modeling, testing, requirements management, and configuration management. However, if CA could broaden the XML integration to other tools inherited from the Platinum stable, such as CCC Harvest, it could claim a useful competitive advantage.

ERwin Modeling Suite 4.0 is an important first step for CA to demonstrate that rumors of the attrition of its inherited tools businesses are greatly exaggerated. CA has headroom for product growth, if it follows up by extending the suite to other tools inherited from the Platinum stable, including configuration management.

Clancy conceded that, in the aftermath of the Platinum acquisition, it was often difficult to get the world’s attention on the company’s development products offerings. With the exception of ERwin, none of these products were market leaders. But Clancy adds, these products have “grown significantly” under CA’s watch, noting that channel sales for modeling tools alone grew 344% over the past 16 months (he did not have figures for sales growth via direct or telesales)..

Obviously, integrating data modeling and object/component modeling has long been a holy grail of application development—but does the idea make sense in practice? Database architects and component designers are each jealous of their domains. In most cases, the database predates the application. “Just ignore the other side, data is king,” is a common refrain, according to one industry consultant who has worked both sides of the modeling fence.

The application life cycle market remains fragmented, with market leader Rational, Compuware, Merant, and Computer Associates each offering varying arrays of development and modeling tools. No single vendor offers a complete solution, and until now, only Rational has offered out-of-the-box integration.

The significance of ERwin 4.0 is that CA is finally demonstrating that it is serious in pursuing the development tools business, and that Rational finally has some competition in the suites sector.

Insights on the world of Information Technology — Views expressed here do not reflect the opinions of Ovum.