Ask TPM: The Economics of Open Source Software
January 22, 2007 Hey, TPM
You are incredibly insightful about the IT world, at least insofar as I can understand your arguments. That’s a compliment. Sometimes I come away puzzled by your analysis. You have one article on the projected lack of IT talent per PwC. Then there is an article projecting collapse for expensive proprietary packages because the open source community is going to provide free applications software. A free ERP–wow!
[Editor’s note: The reader is referring to PwC Consultants Predict an IT Talent Shortage, from the December 4 issue of this newsletter, and The X Factor: You Can’t Steal What’s Free, But You Can Pay a Lot for Something That Isn’t Worth It, from the same issue.]
Here’s the puzzling part. The open source goose seems to be laying golden eggs and giving them away. I really don’t know where all these folks find time to earn a living and write tons of code. Further, if there really is a looming IT talent shortage, is the goose liable to wake up some day and have second thoughts about free eggs?
In other words, can the open source community continue to provide free stuff when the value of their labor keeps increasing? It reminds me of the concept that the Internet should be free, which always seemed absurd to me. As it has turned out, the “free” part is supported by advertising, which is supported by consumers. So we pay less than the true cost of Internet access through free enterprise. Is there a parallel to this in the open source community?
Can you write another article explaining how all this free stuff works economically?
I think you could–or someone could, and maybe even me–write a book about this. If it was entirely clear how it was all going to turn out, that is! While I am a strong believer in open source software, and use it as well as closed-source products in running Guild Companies, the firm that owns IT Jungle and some day other sites, I do not believe that the open source business model is an easy one. I do, however, believe that it is an inevitable one.
In the closed source model, a company hires a bunch of people to create a software product, and then spends a huge sum of money at a sustained rate to tell people about the product and to convince them to buy it. Then, these companies charge even more money–usually 15 to 20 percent of the initial licensing fee for the software–for maintenance on the code. But you and I both know that this revenue stream, which can represent a very large portion of the annual sales at a company, is not dedicated to maintenance. (Vendors usually lump implementation and customization services into the same category as software maintenance, so you cannot pull it apart so easily.) Then, they charge upgrade fees for major version changes.
If software is like hardware in the IT business, about half of the money that is brought in for a product–including licensing fees, services, and maintenance–ends up being spent on sales and marketing. A sliver–maybe 7 to 10 percent of revenue, sometimes as high as 15 percent–goes into research and development of future products, which is probably more accurately called software maintenance except where it is explicitly related to a future, really different product–but don’t get me started. (Too late, Ed.) The rest goes to pay actual workers and to cover the overhead of the top brass, who like to buy jets and sports teams if they are huge companies, but who like to take out a lot of cash if they are smaller because the membership to the golf club and the marina is not exactly cheap, you know. The burn up a lot of money, and most of it is not spent on making software, but on making business happen and enjoying that process.
The reason software companies run this way is simple: software doesn’t fail the same way hardware will eventually fail, because software is not subject to the laws of entropy like hardware is. (I know, it may seem that way, considering the cruft that seems to build up in software systems, but it is all in your imagination. . . . and mine, too.) You have to make changes to take in new accounting rules, new government regulations and such, but that would be easily covered by maintenance fees, which means in the end, once a software package is done and working well, costs should radically go down. But they don’t, because software companies have to explain to you again and again why you need to upgrade, rather than fixing problems in existing software. (How many times in the past decade has poor security been used by Microsoft to convince us to leave a product? Vista is but the latest example.)
Now, jump over to the hybrid open source/commercial model, which has had lots of success with Linux in terms of the effect on the market. But, for all of the noise about Red Hat and its ridiculous market capitalization during the dot-come boom–remember when Red Hat had a market capitalization of over $20 billion and it had sales of $11 million and had lost money for several years?–there is not a lot of money here. Ditto for MySQL for databases and JBoss (now part of Red Hat) for Web application servers. (You can throw Solaris 10, the Unix platform from Sun Microsystems, into this mix, too.) Red Hat is a very profitable company–or at least was until the JBoss acquisition, and it will be again–but in 2007, Red Hat may do $500 million in sales in calendar 2007 compared to Microsoft’s $50 billion or so in the same period.
Here is what open source software companies do: They create a services organization to provide very inexpensive patch and technical support for an open source product–a Web server, an operating system, a database, an ERP suite, what have you. While a project may have started out as something more akin to a hobby, once some services money comes in (because enterprises only use supported products unless the IT managers don’t know what the techies are doing), they can actually afford to plow the money into paying for specific development. Now, the commercial supplier of support for an open source project, which was owned by no one or everyone, depending on your outlook, is now helping steer the development of that project, which is now a product.
As the open source program gets established, IT vendors who provide systems and services want to get into the racket, and they have some of their nerds help out on the code–thus giving them street cred in that community of developers and letting them steer development. Other vendors kick in. With Linux, Linus Torvalds gets paid by Open Source Development Labs, which gets its money from vendors; the IT vendors collectively have several thousand on their payrolls who now represent the bulk of development that goes on in the Linux community. This is not Linus leading a bunch of teenagers in their basements anymore. And so it goes with open source projects that make it to this commercial product status.
What open source companies do not do–and I can assure you if this from personal experience–is spend a lot of money on sales and marketing. They rely on the very large community of people who use their free wares–and help them do the beta testing, by the way–to do their marketing by word of mouth. Another way of saying this is that they rely on the self-reliance of a very large number of people to do their marketing by establishing them as a source of reliable, usable software as a means of establishing a potential base of customers and then they rely on a small percentage of these customers–perhaps on a few percent of the user base–to actually pony up cash for enterprise-class support contract.
The basic assumption in the commercialized open source software model is that most people will not initially pay for support, but some (if not many) will eventually get some kind of paid for support contract. Some will never, of course, pay for support, but that is just the way it goes in the real world.
This truth is why, for instance, Sun was absolutely thrilled to give away 6 million free copies of Solaris 10 in the past two years. It is why Red Hat was ecstatic about having 1 million downloads of Fedora Core 6 in 75 days.
Do the numbers. Sun is making more than $1 billion in Solaris licensing and support contracts–mostly on midrange and enterprise Sparc-based servers. The only way Sun could preserve this revenue stream is to crank up the volume with freebie Solaris, re-port Solaris to the X64 architecture, and hope for the best that hundreds of thousands of customers would pay hundreds to thousands of dollars per year per server for support on a vastly larger base. The other alternative was to announce Solaris 10, see a few hundred thousand installs, hear lots of customers grumbling about the future of the Solaris operating system and the Sparc platform, and watch Solaris revenues drop from more than $1 billion to $800 million, then $600 million, and so on as companies abandoned its Sparc-Solaris base for Linux on X64 platforms.
Does any of this sound familiar? IBM, have you figured out the systems and platform business in the 21st century yet?
The misconception that many people have these days is that commercial-grade open source products are dependent on the kindness of strangers for their development, like the hundreds of thousands of open source projects definitely are. The major, established open source projects use a community development model, but the coding done in these projects are by and large dominated by IT vendors who pay the bill and the cost is not in all cases directly reckoned in the balance sheet.
The Internet never was free. Our tax dollars paid for the development through the U.S. military’s Defense Advanced Research Projects Agency, and then AT&T in the late 1960s created operating systems and compilers to make a new platform for that network–we call them Unix and C, respectively. And these tools were implemented by the military for their own purposes, which we paid for, and then by telecom companies, which we also paid for through our phone bills. The academia stepped in the 1980s and helped these Internet technologies mature, and in the 1990s, they went commercial and replaced proprietary networks. Government funding and college tuition paid for this advancement, too. And then when the Internet went commercial, ISPs charged us for entry and telecom companies charged them for high-speed network services; Cisco Systems and others got stinking rich on router technology. Along the way, a lot of people contributed a lot of free time to the effort, and in many cases they were also paid for their work.
Open source software is not so different, and the future and impending open source ERP systems and other software that businesses need–such as code to control business processes at a higher level and outline the workflow of a company beyond its transaction systems in the data center–will not be so different. All of this technology will be derived from a large body of prior work.
People simply do not want to pay as much as they do for software they have, and the recent history of the open source movement suggests that once a product can be shown to deliver good functionality and have adequate support, people will adopt it where they can because it is simply less expensive. Legacy application support is what keeps mainframes, Unix boxes, i5/OS and OS/400 servers, and Windows boxes alive. Lucky for the vendors, there is a very, very large base of these machines and a huge installed base of applications running on these legacy platforms. But eventually, sooner or later, all vendors get pulled into the mud. Make no mistake.
There is an expectation that IT systems get cheaper over time, and hardware has certainly done this thanks to Moore’s Law on most of the electronic components. The only way to get on a Moore’s Law curve for software is to stop spending so much money on sales and marketing as proprietary software companies do (or don’t because they have legacy lock-in) and to get as much community involvement as possible in software development and beta testing of code.
Ultimately, I think this style of software development will end up being an epic battle where only the largest communities and the best code–and the two will be related, rest assured–will survive. But it will take decades and a lot of effort to accomplish this, and the smart proprietary software companies will be those who see this handwriting on the wall and go volume and go community and get in the game for the long haul.
These forces are not just at work in the software industry, by the way. Let me illustrate with a personal example that shows these forces are at work in all manner of industries. When The Four Hundred was first published in 1989, it was an eight-page paper newsletter that came out monthly; it covered what had already happened in the prior month, and was mailed out to hundreds and then a few thousand customers who had the $175, then $200, then $250 to shell out a year for a subscription. As the price of the newsletter was raised in the 1990s, we added pages to make it more appealing, but the advent of competition from other publishers and them the Internet as a printing press put pressure on the original publisher of my beloved newsletter. We could not spend great wads of money on sales and marketing efforts to build subscriptions, and the competition moved out to the Internet and started building up vast subscriber bases for competitive newsletters that were available for subscriptions, but which had a teaser, shorter version that was available for free.
At Midrange Computing, Monday Morning AS/400 Update, as the freebie version of AS/400 Informer was called, became very popular. This newsletter was supported through advertising and was intended to feed subscriptions into the paid-for newsletter. After the original publisher of The Four Hundred failed, I took it over in late 1996 and spent two years trying to make the transition from paid-for paper subscriptions to Internet-based and cheaper subscriptions, with the ultimate goal of building up the freebie base and feeding into the paid-for newsletters. After 18 months, I had failed miserably, since I never had enough money to invest in this effort.
So I mothballed The Four Hundred and went to work at MC to be the editor of Monday Morning AS/400 Update. After MC went bust in July 2001, I revived The Four Hundred and built out a stack of five newsletters that provided much more content than the original MC newsletters did. Arguably, we provided as much content as the monthly magazine and the online newsletters did–and still do to this day. And to a much larger base of readers.
And yet, because the time value of information goes down every year, we have to work that much harder every year to stay in the game. The Internet has accelerated this trend by lowering the barrier to entry for publishers, which has made the cost of gathering and presenting information the dominant cost. And there has been some shaking out and specialization as users find quality and advertisers seek focus, to be sure. (I am not saying that all information gets cheaper over time. A 30-year-old baseball card radically increases its value over time, thanks to its scarcity. We treasure old memories–the ones we actually remember, that is.)
Because of these competitive pressures for lots of content in a timely fashion, we spend the vast majority of our funds on content. And like others in charge of news and technical publishing organizations, some days I wonder if any publisher of any kind of publication can live in such an environment for more than a couple of years. But, we are in year six of the business and the bills are all paid, so we must be doing something right.
Here’s my point. In the publishing business, there is a natural curve to things, and a force very much like Moore’s Law–the time value of information goes down every year–is putting immense pressure on publishers, who see the cost of delivering content go up every year because of salary, benefits, and other expenses relating to the delivery of content. Anyway, here’s the basic curve. You start out as a small, subscription-based publication with high subscription rates and usually low numbers of subscribers who tend to be elites in their fields. Over time, you ramp up subscriptions, and your success brings in competition. If you are lucky, you can lower prices and make it up in volume. (I was not so lucky.) Then competition puts on pricing pressure and at the same time a need to add more features, so you expand coverage, which costs a lot more money, and you maybe try to justify a price increase in subscriptions by expanding content. Thus begins a content arms race among suppliers of a particular kind of content. Sooner or later, you move on to advertising support to help bear the costs, and sometimes you just go straight to advertising-only support for the publications in a kind of surge on the market. Modern blogs that are trying to make money, for instance, count on star power to draw in readers and just cut out the whole subscription phase and try to get volume, hoping that enough eyes will eventually mean money.
Software is not much different. Open source software companies spend the vast majority of their money on building support operations, since the content–the software itself–is free. They might have proprietary add-ons they charge for; they may have a useful but dumbed down version of their software that it free, but a more complex version that is only available with a support contract.
Proprietary software can stay in the subscription-only mode, charging relatively high prices as long as there is competition that does things more or less in the same way and charges more or less the same prices. (In effect, the proprietary software development model is in a sort of arrested development.) Or, you can be smart like Microsoft and drive out competition by goosing your features and lowering your price every year–as Microsoft did in the late 1980s through the 1990s–as the market itself grows from tens of millions to hundreds of millions of units. (PC operating systems and then office automation systems, in this case, being the units.) Once you have the monopoly, you can jack prices and reap huge profits, and the vastness of the installed base and the legacy applications means you can pretty much get away with it. But this is the exception, not the rule. For every legacy, there are competitors gunning for it, and sooner or later, they get it right. Either the monopoly reacts–as Microsoft has successfully many times–or it gets taken down–like IBM has many times.
Most software will face the same future as my poor paper issues of The Four Hundred did, unless their vendors wake up to volume and community. I may be an old man before that happens, of course. The mistake many people had in the heady days of the dot-com boom was to think that open source would wipe all other software out, much as the Internet would wipe out brick-and-mortar businesses. The biggest proponent of that idea–a publication called The Industry Standard–died very shortly after the boom went bust.
So, in sum, open source software that goes commercial is not free, but I think it inherently means a lower unit cost for paid-for support than a typical proprietary software license and support contract that has fees associated with both parts. Rather than thinking of this as a goose that lays golden eggs giving its eggs away, it is probably best to think of open source software as a chicken coop that has a lot of hens laying real eggs. These are eggs that your customers can afford, and eggs that they can eat, but you ain’t gonna get rich–not even if you sell a zillion eggs. If anything, the goose that laid the golden egg–the proprietary software company that has customer lock in and that can charge an arm and a leg for software licenses, installation services, and ongoing maintenance–is cooked. But it is cooking very slowly–on the order of decades, perhaps.
I hope this ramble helps. It was fun to write on a Friday night after dinner with the wife and kids. I think I need a drink now, though. Thinking about all this doesn’t exactly make me happy. HA!
If you have a question you want me to answer that is not something that our technical gurus would not be better at answering than I would be, click the Contact button at the top of all of the IT Jungle pages and fire away. If I can answer your question, I will.