Getting down to business at LinuxWorld
(IDG) -- The LinuxWorld Conference and Expo (LWCE) was the first major Linux event I'd ever attended. My expectations of the event were largely shaped by reports of the first LinuxWorld Expo and other such events. Articles on Slashdot, Linux Today, and other sites painted a picture of enormous grassroots energy and the reign of the unorthodox.
Even though I didn't have any previous experiences on which to base a first-hand comparison, it was clear to me that this LWCE had a decidedly more commercial bent than the last one.
The free spirit of Linux was certainly in evidence all through the floor, particularly at the .org pavilion, the bright playroom of the show where folks from Slashdot, Freshmeat, FSF, Debian, User Friendly, and all the rest mixed with conference attendees on Crayola-colored inflatable chairs. And such old-school Linux companies as Red Hat, VA Linux Systems, SuSE, and LinuxMall kept the community-feel in evidence at their booths.
But there were many newcomers, and the smell of mainstream business was in the air.
The state of business application development on Linux
Of course, this isn't necessarily a bad thing, especially for those of us who have long been using Linux in mainstream business, and for those of us who have been trying to evangelize open source philosophy in these mainstream business communities. But I must admit that I do look forward to visiting a more grassroots celebration such as the Atlanta Linux Showcase.
Many of the business-oriented newcomers at the Expo were vendors of tools for enterprise application development for Linux. These ranged from old companies now porting to Linux to startups with Linux sprinkled throughout their business plans like so much venture-capitalist sugar.
ObjectShare was selling its CORBA environment, newly ported to Linux. The product has strong ties to Smalltalk and features a set of graphical frameworks. This environment uses XML as a model-exchange language, but apparently cancels out this strength by getting in the way of CORBA's interoperability standards. Unfortunately, this is pretty common practice in the commercial ORB market. Open source CORBA packages do a much better job of pursuing interoperability between languages, platforms, and even rival ORBs.
The creators of Enhydra, Lutris Inc., were also there. Enhydra had a brief whirl in the IT press headlines a while ago as an open source standards-based XML application server.
At LWCE, David Young of Lutris explained to me how Lutris decided as long as two and a half years ago to develop a Java-based framework for its consulting. It eventually released it under the FreeBSD license in order to spread adoption (it claims 510 contributing developers) and boost its core consulting business, using it as an experimental calling card.
This is similar to the more broadly-reported strategy of Digital Creations regarding its Python-based Zope application server. Interestingly enough, Young mentions that other consulting firms as far afield as Australia and Germany have come to base their consulting practice on Lutris's. Lutris was announcing ports of its Java servlet debugger and XML compiler to Linux at the show.
It will be interesting to see how these consultant-fed application servers live in tandem with such grassroots projects as Cocoon and Apache/Jserv/JSP -- or in opposition to them, as the case may be. Commenting on the matter, Young said, "[Enhydra] is designed to be very practical, [and not as if it is] coming out of the lab, because it is based on customer requirements."
Pick Systems Inc. is a venerable name in commercial database management systems. Apparently, it is also an old hand in the Linux market. A representative mentioned that the company first ported its multidimensional DBMS to Linux in 1994. He told me that Pick was inspired to make the port because of the spirit of the burgeoning Linux community, but that there was no customer demand until the recent Linux media hype. Now it has several large customers, including the city of Garden Grove in Orange County, CA, which uses the Pick DBMS on Linux for municipal services.
At the show, Pick Systems was touting its D3 DBMS engine and FlashCONNECT middleware on Linux. Despite its long-standing presence in the world of Linux, my source told me that Pick had no plans to become more closely involved in the open source community.
Unify produces a series of middleware products with the eWave Enterprise JavaBeans server at its heart. It has ported its entire product line to Linux, and it reports that many of its clients are making the move to Linux, including those who have been using Unify on SCO or some other commercial Unix since 1988 or so. It was pressed to make the port by client demands, yet Unify too has no plans for involvement in the open source community in any broader manner.
POET Software, one of the pioneers of object databases, is meanwhile expecting to release the Linux port of its object DBMS and most of its tools in October. POET's representative said that the company had released a Linux port a couple of years ago, but didn't maintain it. Strangely enough, POET's people talked about how strategic the coming port will be in the same booth where they featured a brochure, dated August 1998, touting the company's "focus on Windows NT," and how it was "mothballing" its OS/2, Mac, and Unix versions.
It's amazing how quickly times change, and with them business strategies. In 1998, it probably would have been considered basic business sense to put all eggs in the NT basket. A more recent POET brochure, dated July 1999, heralds POET Object Server Suite 6.0, with no sign of Unix in the mothball drawer.
POET's ODBMS supports portions of the Object Database Management Group's (ODMG) standard, especially the Java Binding and OQL (Object Query Language) queries. Other aspects of the DBMS are subject to proprietary extensions and substitutions. Playing loose with the standard appears to be as common in the commercial object database market as it is in the CORBA market. In the case of ODMG, there is some justification because of such omissions as nested transactions.
Talarian sells a distributed messaging product along the lines of IBM's MQSeries and TIBCO's products. It has ported its core component, RTServer, to Linux, as well as its APIs. It claims that many of its existing customers are looking to move to Linux, not least because, as Talarian representative Tim Young said, "Linux is the highest-performance platform we have." However, the company doesn't see any intersection between its business model and the open source development model.
It was also interesting to note the experiences of consultants moving clients into the Linux space. At many of the sessions, consultants discussed their experiences dealing with Linux's lack of flashy IDEs and combatting the efforts of Microsoft to push COM and Visual Basic with Windows 2000 into the improbable area of scalable business systems.
One small consulting firm with its own booth was Database Specialists Inc. (DSI), a California Oracle consultancy which has seen a lot of client demand for Oracle on Linux. DSI's representatives pointed out that the company had at least two major clients running Oracle and Linux on production systems. In fact, Roger Shrag of DSI led a tutorial on performance-tuning for Oracle/Linux, an opportune topic considering all the recent bother over tuning Linux for high-end server applications.
So it does appear that there is a lot of activity in the middleware market under Linux. To paraphrase one of the LWCE session titles, there is no doubt that all three tiers are available on our favorite platform.
However, I'm losing my (probably vain) hope that Linux's openness would encourage middleware vendors to be more standards-oriented. As a consultant in the area, I can attest to the frustration of dealing with the partial compliance -- or, in some cases, the complete noncompliance -- of commercial products with open standards. As I've mentioned, open community efforts, from Cocoon to WDDX to Orbit, provide excellent examples of how interoperability need not be impractical. Hopefully, commercial middleware vendors will get more from the Linux community than just a lucrative platform.
What are you going to do for us today?
And, speaking of fair exchange, you may have noticed that, besides Lutris, all of the development tools vendors to whom I spoke admitted that they had no plans for involvement in the open source community beyond porting their proprietary apps to Linux. I'll go a bit further and mention that I got some very odd (and in some cases even contemptuous) looks when I asked this question.
Perhaps this attitude was warranted: the various open source business models have yet to be tested over the long term. And with all the media hype, it is hard to know what is valid and what is, well, hype. The business middleware market is cutthroat even by software industry standards, and it's only natural for companies to port to the hottest platform, while casting a dubious eye at any impetus to tamper with their familiar sources of revenue.
But granting all this, it was strange not to see at least a bit more diplomacy at a convention celebrating open source's greatest success to date.
Modeling open development
And what indeed does the open source development model have to contribute to business application development? Larry Augustin of VA Linux Systems led a panel with leaders of the most prominent open source projects, including Samba, Linux, XFree86, Apache, Perl, and FreeBSD.
Now, such a discussion may not, at first, seem to bear any relevance to enterprise software-development methodologies. The open source projects represented were large, highly distributed, noncommercial, and not driven by a uniform business interest. And yet businesses are increasingly multinational or otherwise dispersed, and crucial apps must sometimes be developed in concert by spread-out groups.
Also, the most expensive software has long ceased to be shrink-wrapped. These days, most software dollars go to complex and heavily customized business-support software, such as enterprise resource planning set-ups. This sort of effort may have much more to learn from open source methodology than one might imagine, even if the openness is all contained within a source-proof wall.
The panel focused mostly on the social issues of open source software development: government, communication, and conflict resolution. There were some striking differences amongst the panelists, and to a certain extent these differences mirrored the different tactics of the software projects that the panelists represented.
Brian Behlendorf of Apache discussed the consensus decision-making of Apache, which left me amazed that the project is not mired in gridlock. Jordan Hubbard of FreeBSD explained how that project managed revision control when over 150 people have permission to commit changes to the repository. Such differing, yet successful, approaches are enough to make one think hard about the homogeneous way that software is developed in most businesses.
As a consultant, I have seen a lot of inefficiencies in the typical development process. This process usually features a project manager, who is primarily trained as a business professional, herding a team of programmers and designers of varying competence and interest towards a goal. Large companies might find unexpected efficiency by allowing projects to develop from the bottom up, putting safeguards and milestones in place to ensure alignment with immediate business goals, but leaving the box at least partly open.
The most complex and widest-ranging projects would probably evolve a unique management and technical dynamic. This sounds inefficient, but that may be only because we are so inured to the tremendous inefficiency that afflicts general software development. We all know the anecdotes and the statistics of how many software projects are too late or too costly when they finish -- if they finish at all. Perhaps a radically different approach is in order.
So long, farewell, auf weidersehen, goodbye ... and thanks for all the fish
So sang the keepers of Slashdot in the introductory video to the LWCE keynote. And that wasn't even the weirdest part of the whole tradeshow. There was much other silliness to report, from the farce of presenting the Linus Torvalds/IDG community award to Richard Stallman, to frenzied members of the press trying desperately to find ragtag Red Hat investors on IPO day.
But, from an enterprise development point of view, it is easy to summarize the show: Linux Greeks invaded corporate Troy disguised as reliable file, print, and Web servers, but a good number of them are also busy changing corporate software development. In addition to the products I mention above, there was news from Borland, Cygnus, SGI, BEA, IBM, Oracle, Sybase, and ApplixWare. From IDEs to DBMS, Linux is increasingly well-represented in the enterprise.
LinuxWorld Expos.past and future
RELATED IDG.net STORIES:
LinuxWorld: Can business and Linux work together?
|Back to the top||
© 2001 Cable News Network. All Rights Reserved.|
Terms under which this service is provided to you.
Read our privacy guidelines.