A couple of folks have written in about a comment made in my del.icio.us links for last week, in which I expressed mild disagreement with Mike Olson’s post on Oracle’s Linux offering. Having been necessarily brief to comply with del.icio.us’ regrettable text field limitations, I thought I’d expand on the nature of my objection just a bit.
Before I continue, the obligatory disclaimer: I’ve known Mike for several years now, worked with him and the rest of his team at Sleepycat, and have a great deal of respect for him personally and professionally. As for existing client relationships, while his Sleepycat folks are still considered RedMonk customers, even subsequent to their acquisition by Oracle, his new employer is not at this time a RedMonk customer. We have been briefed, however, by the Oracle Linux folks on more than one occasion.
With that out of the way, let’s review the basics. For those of you that haven’t been following along at home, Oracle entered the Linux market – by proxy, at least – back in October, as I discussed here. Rather than create their own unique distribution, or tab an existing volume distribution that lacks enterprise heft (think Ubuntu), they chose to fork (my word, not theirs) Red Hat’s Enterprise Linux (RHEL) offering. Fast forward to the end of March, when they announced a relatively impressive customer list for their fledgling Linux offering, and we’re more or less up-to-date.
The question that many had subsequent to the original announcement, was why fork RHEL? Why not partner? Or create their own distribution? The answer, to me, is very straightforward: the value in many enterprise contexts is not in the technology, but the ecosystem that surrounds it. Case in point is NetBeans.
As I tell the Sun folks from time to time, the gains NetBeans has made technically in the past few years are startling. The product has achieved technical leadership in certain areas of the IDE space, which is quite an accomplishment for a project that was barely usable several years back. But despite the progress, I’m no more bullish on the prospects for NetBeans than I was two or three years ago, and my reasoning is simple: the battle is in the ecosystem, not the technology – at least on a volume basis, individual developers may disagree. And Eclipse’s ecosystem is far, far larger than NetBeans, and very few of the ISVs I speak with feel there’s sufficient incentive to invest in supporting the NetBeans platform alongside of Eclipse. Biggest Community Wins, in other words.
Unsurprisingly, the Linux community is no exception to this particular rule. Oracle starting a distribution from scratch, then, was probably a non-starter all along, because the effort required to push the rock up the hill in no way justified the probable returns. Likewise, Ubuntu represented an only slightly more attractive offering – and that’s speaking as someone who believes that the Debian/Ubuntu/* communities collectively are significantly larger than Red Hat’s. The problem with Debian/Ubuntu is not community size, but rather community type. Based on the package repository numbers alone, it’s difficult to argue with the idea that the extended Debian ecosystem – which includes Ubuntu – far outstrips that of any of the higher profile commercial supplier. With one important exception: commercial support. The commercial community that writes to Red Hat dwarfs the Debian community, although I still hold out hope that the LSB has the potential to change those dynamics in time.
As it happens, the commercial community is the most interesting for a vendor like Oracle, as they tend to like to sell software rather than give it away. So with Oracle’s interests in the largest addressable commercial Linux community, and Red Hat’s leadership position in the same community, well, you’re smart folks. It’s not hard to see where the interests lie. Based on the above, you might logically conclude that Oracle would simply acquire Red Hat, but instead Larry Ellison decided to take advantage of what he perceived to be – and many would agree – an inherent weakness in the open source model. The fact that the source is, you know, open. Thus it was that Oracle entered into the Linux market at the cost of some bit scraping (for trademark purposes), hiring (qualified kernel engineers), and the usual operational costs (marketing, sales, etc).
While it’s been argued by many – myself included – that community is one of open source’s most effective barriers to entry, Oracle chose to – to use Matt’s word – co-opt that community. To effectively leverage Red Hat’s own community against itself, in a bit of corporate jujitsu. The barrier to entry in this case, then, was approximately as effective as the Maginot Line. Makes you wonder what might happen if NetBeans were to bite the bullet and introduce Eclipse plugin compatibility to its product.
Speculation abounds as far as Oracle’s longer term motives with this move, but suffice it to say that they are well within their legal rights to do precisely what they did.
But the larger question of whether this is an astute business decision on their part remains. At their current pricing levels, it seems unlikely to me that this will be a financially rewarding standalone business in the near to mid-term (assuming, of course, that Red Hat continues to be a viable competitor). But perhaps that’s not the strategy, perhaps its intended to merely be an additive component to other larger, enterprise sales. The less money that customers shell out for the operating system to Red Hat, one might assume, the more dollars available for the higher margin database products. Or perhaps this is merely the prelude to a longer term end – a Red Hat acquisition. If Oracle can materially impact Red Hat’s business, it may be in a position to acquire the company at a significant discount; I won’t even try and do the math comparing the launch costs for Unbreakable Linux to potential Red Hat valuation discounts, but suffice it to say that’s the assumption some people are working under.
What is the nature of my objection, then? Mainly that I don’t see an overwhelming positive here. For anybody. Take Oracle. At best, the operating system business is yet another major technology area that they’ll be required to invest in and support longer term – for what I’d expect to be non-trivial but non-excessive returns in the financial sense, and potentially negative returns in the PR sense. At worst, it’s a significant distraction from other businesses under pressure. For Red Hat, the impact should be fairly negligible over the short to medium terms, as strategic operating system decisions tend to have long buy cycles, but I cannot help but think that Oracle’s entry into their market will have longer term negative repercussions – in margin, if nothing else. As for the most important part of this equation, customers, I’m not sure they’re better off. They’ll be transitioning from a relatively mature and well understood Linux provider market to one in which they may have to make difficult choices: pay the premium for the devil you know, or go cheap for the devil you don’t.
The net of my objection is this: the money Oracle will promise to save customers has to come from somewhere. This isn’t a magic trick. Clearly, Oracle intends for the majority of that money to come from Red Hat’s sales pipeline, and that’s certainly their prerogative. If they wish to subsidize their efforts to poach Red Hat customers with the margins they make on their database business, that’s their affair. But get this: while Oracle’s apt to damage or at least inconvenience Red Hat, they are also dependent is simultaneously dependent on the longtime commercial Linux leader – a paradoxical and rather unenviable position that is likely to have implications at some point down the line, for the vendors and customers alike. Not good ones.
Over the long term, it’s possible that the Oracle/Red Hat dynamic could ultimately prove less divisive than an LSB unified Linux. But in the interim, Oracle’s likely to be lowering the oxygen content in the largest commercial Linux market in return for what I believe are likely to be modest gains. Do the risks justify the returns in this case? Not from where I’m sitting, but I promise to keep an open mind.
