It’s perhaps only fitting that I begin a defense of the Debian Linux distribution with remarks from one of the people behind one of Debian’s more significant competitors. It was efforts like Debian, after all, that are likely to have inspired the original comment. The assertion in question was made several months ago by none other than Sun’s Jonathan Schwartz. The context, of course, was Linux. Linux, the open source project that had done more grievous damage to Sun than its supposed arch-enemy, Microsoft, ever did.
Schwartz said in discussing Linux that as a business it was “tough to compete against a social movement.” He was even more blunt on the same subject in a later piece, one not so Linux specific, in which he postulated that “you can compete against a product, but it’s close to impossible to compete against a community.” I include that because I think it’s important to emphasize the link between a social movement, and the manifestation of that movement – community. In any event, the implication of these statements are clear: Sun would choose to focus its competitive energies not on the expansive Linux community – which would in all likelihood be more futile that fighting the tide, such is a community’s ability to dissipate and diffuse focused attacks from a single competitor (e.g. Microsoft) – but on the businesses built on top of that community. Businesses such as Red Hat.
Few in the industry question the fact that for enterprises worldwide, Linux often means Red Hat. This is less true abroad than it is here in the United States, and certainly there are areas where Red Hat has made less headway – like the desktop or embedded spaces. But the simple truth is that Red Hat is top of mind when it comes to Linux server deployments. And Sun, for its part, is only too happy to fuel this assumption. Why? Because they stand a much better chance of competing with a commercial entity such as Red Hat – which has attributes such as pricing that can be attacked – than they do with the Linux community as a whole.
It’s also interesting, if not precisely germane to this particular discussion, to note just how Sun has decided to compete on a long term basis: by founding their own communities. It’s imperative, it would seem, for vendors these days to be able to answer the “you and what army?” question with “us and our community.”
Returning to the subject at hand, however, let’s look at the question of community. It’s a problematic term, because it’s difficult to answer a question such as what, precisely, does that term actually mean in the context of open source? The answer is that it depends. The nuances of community and the way that software is consumed frustrate efforts to succinctly answer that question. Berkeley DB or MySQL, for example, reveal as flawed the idea that a projects importance should be gauged by the number of commiters on a given project. The potential importance of the Asterisk project, for another, belies the notion that significance can be measured solely by users. And so on.
But within the context of operating systems generally, and Linux distributions specifically, community means several things. It means support from application developers targeting the platform. It means support from developers maintaining the platform. It means support from users willing to use and potentially pay for services and support of the platform. Unfortunately, this definition almost entirely precludes accurate measurement. Where vendors once could demonstrate success in the marketplace via sales figures, open source fundamentally undermines that so as to make it nearly meaningless. Eclipse and JBoss, for example, are highly common within enterprise Java shops, but only a fraction of the implementations translate into measurable sales.
Some projects will point to downloads, but that’s only slightly more reliable. RedMonk, for example, runs about 6 instances of MySQL. Zero, as far as I’m aware, have counted towards MySQL’s official download figures because the software was instead download from a variety of Blastwave, Debian and Gentoo package libraries.
If neither sales figures nor downloads are accurate metrics for determining traction, what is? There are a variety of other approaches; DistroWatch.com, for example, tracks page visits. Ubuntu is leading that list, and according to Alex is ahead on Google Trends as well. Other services like Umbria can provide figures relating to the volume of participants within a particular user forum, and trends therein. But these again are of only limited utility. The short answer, as nearly as I can determine, is that determining traction and mindshare within the Linux distribution space is an almost purely qualitative exercise.
So almost entirely qualitatively, I’ve come to the conclusion that Debian – and its various distributions – represent the largest Linux community ecosystem. For some time (here’s me advocating IBM support for Debian in ’04), the datapoints that I track, which include many of the above metrics along with all of the other information at my disposal, from blogs, calls, and conferences, have indicated to me that in terms of traction, the Debian community is far and away the most popular of all of the Linux distributions. In many respects, it’s not close. I say that not as a real member of that community – while I’ve run it for some time, I’m far more active with Gentoo than I am with Debian – but as an observer. An analyst, if you will.
Many will scoff at that assumption, contending that it is in fact Red Hat and SuSE that are the real Linux standard bearers, because they enjoy the commercial ISV support that Debian and most of the other community distributions lack. On that point, the critics would be correct: setting aside the fact that a great many ISVs provide “unofficial” support for Debian, the commercial ISV support for Red Hat dwarfs that available for Debian. Out of the IBM Software Group portfolio, for example, the only application that I’m aware of certified for Debian or a Debian-based distro (yes, there are distinctions therein, but I don’t have space for that here) is DB2, which the folks from Ubuntu self-certified themselves for.
Will that always be the case, however? Will Sun’s contention that Linux equals Red Hat continue to be true indefinitely? Personally, I don’t believe so, and that’s where Debian comes in.
I’ve been asking the above question to a variety of people for several years now, and the response has always almost universally consistent: Linux will continue to be a two horse game, Red Hat and SuSE (to keep them honest), for the foreseeable future. I did, in fact, ask the former and current Sun CEO’s just that question in November of ’04. Here’s how I described that:
Jonathan Schwartz answered my question with one of his own – “You mean Red Hat?” And that actually segued nicely into my real question – which is really a larger question for Linux supporters. What if Red Hat != Linux?
What if another distribution was tapped as a target for ISV support – say Debian stable? Scott McNealy scoffed at the idea saying “there’s no there there,” and Jonathan more or less shot it down. I can’t say as I really blame them. I’ll certainly grant that it’s a major reach. Supporting a new OS, as any ISV knows, is not a trivial exercise. It takes a Lot of money to make it happen, and a clear incentive from a cost/benefit standpoint.
What’s changed since then? On the surface, not much. Red Hat and SuSE are still 1-2 in terms of perceived Linux importance. Neither IBM, nor any other major ISV, has stepped up and supported Debian. So Jonathan and Scott were right then and are still right today. What I’m wondering is for how much longer. Consider the following:
Solaris, the operaring system once left for dead by more or less the entire industry apart from Sun, has seemingly righted the ship. IBM, for example, which was loathe to commit to support the latest iteration finally did so. In a variety of enterprises, the wholesale migration away from Solaris to Linux has slowed. In some cases it’s even reversed, thanks to features such as DTrace. And as I’ve discussed previously, there are an increasing number of vendors that are beginning to question the economics of supporting Linux – which essentially requires QA and testing for two operating systems – versus Solaris, which is one. That’s not to imply that any of the vendors are contemplating a departure from the Linux platform, but a few of them are looking for ways to simplify their costs of supporting it. What does all of this mean? Not that Linux is going away, nor that Solaris is somehow – as Mark Andreessen put it – a better Linux than Linux. What it means, simply, is that Solaris has reemerged as a competitive threat.
Several years ago, the level of understanding of what constituted open source traction was low. As a result, projects like JBoss had a difficult time competing effectively with their commercial counterparts, because they had to spend an inordinate amount of time educating their customers and staving off questions of viability. Success in those days was measured still by reports from analysts measuring, typically, sales figures. Fortunately for open source projects, however, developers rarely if ever read these types of reports and instead deployed what often was simple, technically credible, and well supported by its surrounding community. This is a simplified means of explaining why Red Hat bought JBoss for $350 million dollars; JBoss flew under the radar for several years, building a sizable community in the process. How does this pertain to Debian? Well, as it turns out I’ve spoken with a great many Red Hat shops that are running Debian in development, staging and test/QA. The perception that Debian can’t be relevant because it’s not Red Hat or SuSE is no more accurate than the assumption that JBoss couldn’t be relevant in a world with WebLogic and WebSphere.
IBM has, perhaps more than any other single vendor, been identified as betting its company on Linux. While it has chosen not to market and sell the operating system directly, it’s been content to go to market with one of the two leading commercial suppliers of Linux – the aforementioned Red Hat or SuSE. But those relationships are evolving. Most obviously, there’s the JBoss acqusition by Red Hat. This makes IBM and Red Hat competitive in a way they were not before, and while hasty realignments are unlikely, you can bet that IBM is not pleased by the amount of account control Red Hat has at the moment. Speculation then turned to Novell, but the company behind SuSE a.) does not have the mindshare that Red Hat does and b.) already had established fairly tight ties to JBoss. But the real concern for IBM might be: what if Oracle did indeed acquire one of the two leading vendors? That would put the folks from Armonk in something of a bind; unless, of course, they’d previously established tighter ties to the largest Linux distribution community – though a commercial layer such as Canonical, if necessary.
For as long as it’s been around, the LSB has been a chicken or egg question. ISV’s contended that it would be more relevant if the platforms invested in it more seriously, and the various Linux distributions were likewise unimpressed with lukewarm ISV acceptance. Two things might be changing this: first, the aforementioned economics of supporting two official (many ISVs will tell you that they provide unofficial support for a third, typically Debian) Linux distributions is burdensome on ISVs that are watching their margins drop. Second, according to the folks I’ve spoken with representing the LSB, the standard is aiming to become more aggressive and forward looking than it has been in the past. How is that important? See my comments on Debian and the LSB here.
If one accepts Schwartz’ contention that competing with social movements and communities is tough, the obvious conclusion is that Red Hat and SuSE will be competing with Debian sooner or later. After all, Debian alone has managed to build and sustain a sizable community; SuSE’s effort – OpenSuSE – is relatively nascent, and Red Hat’s Fedora is sizable but not comparable to Debian. Of all of the distributions, Debian enjoys the single largest repository of packages – a good indication that it enjoys significant advantages in terms of developer support. The question will be whether the current disadvantages in terms of support from the commercial ISV community can be mitigated, either by LSB support or simple politics, in which case Debian’s advantages in application support and developer traction will become more compelling.
For the above reasons, among others, I believe that Debian is a tremendously important project – one poised to assume a much more significant role than it has had previously. Through a combination of its own virtues and external developments, I believe it’s possible – likely, even – that Debian becomes a first class citizen in the Linux distribution market. Whether that comes alongside of, or at the expense of, one of the two current frontrunners remains to be seen. There’s no question that change will not come quickly – it rarely does, but should one of the larger ISVs indicate its support, it could happen sooner than people think.
Astute readers will recognize that there is one significant question still remaining: where does Ubuntu fit in all of this? While the entry purports to be about Debian and Ubuntu, the discussion has centered almost entirely on Debian. The reason for that is simple: unlike some in the Debian community, I believe that Debian and Ubuntu are inextricably interlinked, and the success of one fuels the success of the other. It goes beyond the fact that Ubuntu employs Debian core developers, or the fact that they share a package-management infrastructure: Debian and Ubuntu are, to me, an indication of how Linux distributions could and should function alongside one another. Do they compete head to head for deployments? Indeed. But if enterprises were considering deployments of Red Hat, SuSE and Ubuntu, the latter would seem to benefit Debian the most. Ubtuntu is in some respects the Linux distribution equivalent to Ruby on Rails: it values convention over configuration, and makes certain choices in the name of simplicity. In that, its approach differs significantly from its parent distribution, but that’s not a bad thing as far as I’m concerned. The development of distribution families – Red Hat / Fedora / CentOS, SuSE / OpenSuSE, Gentoo / Kororaa, Debian / Ubuntu / Xandros / Knoppix / etc – is certainly not a universal positive, but it’s a.) inevitable and b.) not without its advantages. It’s my hope that the Debian community will ultimately see the benefits of a popular Ubuntu, the most important of which is that the underlying Debian technology has a chance to reach a serious volume audience.
This is particularly important, because from where I sit it seems entirely possible that Ubuntu and its corporate partent Canonical could be tabbed as the corporate interface into the Debian community. It’s difficult to imagine large ISVs such as IBM or Oracle dealing effectively with the Debian community, due to the cultural gap alone. Canonical, however, would seem to be an effective bridge between the two parties, having as they do one foot solidly in the Debian community with the other in businesses models the ISVs would understand.
Whatever happens, however, it will be interesting to see how Red Hat and SuSE fare as they compete with the largest Linux distribution community on the planet. Anybody care to wager? 😉