I’m perplexed by Ellison’s FT interview re: open source. I only half believe him when he says that Red Hat doesn’t own anything. Ellison is trying to cast a shadow on the Red Hat deal from two weeks ago, and, to me, it speaks volumes that he’s so prepared to answer questions about the JBoss deal. Here’s an illuminating (and perplexing) quote:
“… Why didn’t we buy JBoss? Because we don’t have to - if it ever got good enough we’d just take the intellectual property - just like Apache - embed it in our fusion middleware suite, and we’re done. We always have that option available to us - IBM always has that option available to them.”
Nonsense. Not buying JBoss was a mistake. Red Hat now owns the JBoss brand, owns a lucrative services operation, and employs a set of very talented and loyal developers. Ignoring the licensing differences for the moment, you won’t just embed JBoss “and we’re done”…”just like Apache”, in fact, i’m pretty sure you are going to be cursing the fact that you didn’t pay that premium you so swiftly dismiss in your interview. You see, once you are forced to “take the intellectual property” (which you can’t anyway), your customers are going to start to wondering why they are paying Oracle so much for your “fusion middleware suite” when they could just get an fully integrated solution from Red Hat. And, if Red Hat were to purchase MySQL AB, you’d have a single corporation providing a fully integrated open-source solution for both LAMP and Enterprise Java. Ellison - two big ships on the horizon Red Hat and MySQL (which is growing up faster than you can say “solid enterprise class database”). Sleepycat and InnoDB acquisitions notwithstanding, this is still an interesting game of chess.
This interview puts Ellison’s strategy on the table: he’s making the argument that companies like Red Hat don’t “own anything” and that open source isn’t “good enough” yet. Plus, there’s been some nefarious speculation about Red Hat being the next SCO courtesy of a perfectly timed opinion piece at Eweek masquerading as “news”. Oracle passed on the JBoss deal, and now they want us to think that there are “IP ownership” issues with JBoss code. Don’t believe the hype.
Embedding JBoss “just like Apache”
He sets up a false comparison. He compares JBoss to Apache (I’m assuming he means the web server). Big differences there: licensing, branding, services, and architectural significance. Licensing: Apache HTTPD is under a completely different license than JBoss, and while many will argue that LGPL is similar to a BSD-style license, I’d argue that the LGPL license requires a level of reciprocity not present in the Apache license. While Oracle has shipped a customized “Oracle (Apache) HTTP server” in the past, I’d be surprised if they ever option for something similar with the JBoss codebase. Then there is branding, JBoss/Red Hat own a brand, own the development community, and they are the logical choice for services. Then there is complexity, the Apache HTTP server isn’t an application server, it isn’t a differentiator that sets Oracle apart from the rest of the pack. When people are choosing an enterprise platform, the Apache HTTP server is less of a conscious choice and more of a foregone conclusion. Plus, the web server can be swapped out and replaced with ease, the application server and J2EE stack is a completely different kind of integration. While web server selection has a minimal impact on implementation strategy, application server selection has a broad impact on everthing from training to services to build systems and implementation strategy.
Contrary to what Ellison states, just “embedding” JBoss in his middleware suite presents a few problems. First, JBoss already has a good middleware suite in JEMS, why would an Oracle customer want to pay Oracle for the privilege of running JBoss within a middleware framework not designed with JBoss in mind. Second, because of the license, Oracle will find it more difficult to “embed” the product “just like Apache”.
Then he brings up IBM, which is curious - “IBM always has that option available to them”. IBM already made a move in this space. When IBM acquired Gluecode, they essentially gained a core group of Geronimo developers, and are already offering a viable free product that compliments the WebSphere product line and helps to cut away at Oracle’s application server market share in the meantime. I’m not 100% sure exactly what he means by this reference to IBM. Ellison is assuming that the cost of an acquisition needs to be justified by ownership of intellectual property and/or assets, he keeps on making references to “ownership” in his interview. Whether or not Red Hat owns anything is debateable, they might not own a great deal of proprietary, closed-source software, but they drive a large community and offer subscription-based services and support. They own a “stake” in a “software ecosystem”, one that is going to be hard to challenge no matter how much money Oracle throws at it.
If JBoss “ever got good enough”…that’s a good one…
You can tell Ellison had to rationalize not buying JBoss, catch the phrase “if it ever got good enough”. :-) Here’s some news. JBoss is “good enough”, in fact, JBoss 4 is great. Forget my tirade against corporate LGPL from last week, you might find this surprising, but I *use* JBoss all the time. In fact, I recommended JBoss to a colleague the day after I wrote my previous entry about JBoss. In this statement, I don’t see a triumphant Oracle CEO touting the merits of his “fusion middleware suite”, instead I see a CEO on the defensive trying to poison the well for a competitor that is going to give Oracle’s application server product a run for it’s money.
Oracle’s “Embedding” Strategy
Lastly, the Oracle organization has this theory that they will just absorb successful bits of open source technology into the core Oracle platform, and I use the word “platform” on purpose. Contrary to what Oracle’s marketing department would like to think, the majority of people I work with consider Oracle to be a database vendor. Most of the Oracle specialists I know of call themselves DBAs. At least to this engineer, Oracle is a database, but they make me suffer through the process of installing a bloated platform on my development and production environment.
I cringe everytime someone informs me that a project involves Oracle. “Oh, it’s easy just install Oracle.” When was the last time you installed Oracle. Ever tried to download the full copy of Oracle client. The 10g release 2 download of just the server and the client weigh in at a little over a gigabyte. Now, if you are like me, you are interested in using Oracle…..the database. and, the fact that Oracle is bundled with everything from an httpd server, to an application server, to a JVM, to a set of XML servcies, ad infinitum is more of a nuisance than anything else. Now the CEO is telling us that if jBoss “gets good enough” he’ll just embed it into his middleware product. If you are already downloading a gig, what’s another one on top of it.
Don’t get me wrong, Oracle is a solid product. But, IMO it is as bloated as the budgets it requires. Again, advantage JBoss, i’m assuming that Jboss will be integrated into Red Hat Network, making it even easier to deploy.
Wait? What’s all of this about JBoss being “great”?
Yeah, seriously, didn’t I just write a whole piece about how much I disliked JBoss? No, that piece was more about my distaste for a specific approach to open source licensing and corporate control. JBoss is a great piece of software and they have a great team of developers. I’ve met more than a few and they are brilliant and optimistic - Hibernate is great, and EJB3 (largely chamioned by JBoss) rescues J2EE from the dark ages. When I sit across a table from a JBoss employee, I have much more in common with them than an Oracle or Microsoft employee. I prefer open source, I prefer code availability.
I’m not anti-FSF or anti-GPL. If I said that I’m a hypocrite. I’m perfectly happy to use JBoss, just like I’m happy to use Linux. If I’m developing a web application for a client that doesn’t require distribution, I’ll use GPL and LGPL licensed libraries and software. Where I differ is in distribution and customization, if I’m going to customize, extend, and distribute a product I avoid dependencies covered by GPL or LGPL….especially those maintained by a single corporation. I prefer BSD-style, it just feels freer.
It’s a subtle disagreement, a subtly that is lost when my blog post is picked up by TSS. :-)
So for the record. JBoss is open source. JBoss is great software. And, JBoss employees are not nefarious corporate robber barons looking to squeeze every last dollar from us via software licenses and subscription fees.