Open Source: Your Expectations are Broken

In Open Source: The Model Is Broken, Stuart Cohen complains that the business model based on selling support and services around open-source software is “not meeting the expectations of investors”. In related news, the sky is failing to meet my expectations by not raining tasty soup.

Mr. Cohen notices that “Some have succeeded. Many others have failed or will falter, and their ranks may swell as the economy worsens.” Congratulations, Captain Obvious! Because, of course, there is no other economic sector in which the vast majority of business ventures fail.

I have to wonder what planet Mr. Cohen has been living on, and how he managed to miss it eight years ago when I warned everyone (in The Magic Cauldron) that the day of the high-investor-multiple software startup was over. Service businesses simply don’t have multiples as high as capital-intensive manufacturing businesses; not only was the decline in expected returns from startups completely predictable once they stopped being able to charge secrecy rent, I actually did predict it. Repeatedly and loudly.

Mr. Cohen also brings us the news that running a telephone support desk is a thin, low-margin business; and that successful companies like Red Hat find other ways to add value to their open-source codebases. Zounds! Captain Obvious strikes again! How does he do it?

Mr. Cohen is actually missing the biggest news, which is that “failing to meet investor expectations” is a good thing. Investors thrive on arbitrage — capital arbitrage, location arbitrage, time arbitrage, IP arbitrage. As markets and production systems get more efficient, they compete inefficiencies out of the system and both risk and opportunities for arbitrage decrease. This is why mature industries have the steadiest but lowest returns on investment.

Investor expectations about software-company multiples were formed in the industry’s squalling infancy and lusty adolescence. The “failure” to meet these is nothing more than a sign that, with open source, the software industry is finally growing up.

60 thoughts on “Open Source: Your Expectations are Broken

  1. Neal Stephenson (among others, certainly including yourself) noted that Bill Gates’ singular invention was the meme of software as a manufactured product. But if that’s wrong…

    The other interesting thing is that capital-intensive manfucturing businesses are structured differently from knowledge/skill intensive industries. I would argue that that’s actually the correct model for software, and that those businesses would naturally be structured similarly to other businesses where practicioner skill is paramount – law, medicine, maybe investment banking. Most of those seem to run as employee-owned partnerships (often with several tiers of equity ownership and some less-skilled salaried staff). I’m not sure I know enough about their arrangements to say why that’s so (to retain intellectual capital?) or speculate much further, though…

  2. >The other interesting thing is that capital-intensive manfucturing businesses are structured differently from knowledge/skill intensive industries. I would argue that that’s actually the correct model for software, and that those businesses would naturally be structured similarly to other businesses where practicioner skill is paramount – law, medicine, maybe investment banking.

    Correct, or at least exactly coincident with what I’ve been telling investors for ten years.

  3. “the biggest news, which is that “failing to meet investor expectations” is a good thing”

    Very smart remark. Thanks for it.

  4. >>The other interesting thing is that capital-intensive manfucturing businesses are structured differently from knowledge/skill intensive industries. I would argue that that’s actually the correct model for software, and that those businesses would naturally be structured similarly to other businesses where practicioner skill is paramount – law, medicine, maybe investment banking.

    >Correct, or at least exactly coincident with what I’ve been telling investors for ten years.

    What is missing from software development that you find in those other fields? Individual accountability for the outcome of the individual’s work and the impact of error/failure, usually enforced through government licensing of the practitioner and regulation of the practice. Even in those businesses where the firm (partnership, etc.) indemnifies its members, the firm does so by absorbing the risks of misconduct or error and enforcing individual accountability and corrective measures internally. Cases where this system is not properly implemented and/or executed are newsworthy because they are rare compared to the number of cases where this system is implemented and executed as well as humanly possible.

    Until software developers are held to equivalent levels of individual accountability, don’t expect to see them considered and treated as the professional (and economic) equivalent of practitioners of law, medicine, investment banking, physical engineering (civil, mechanical, aeronautic), etc.

  5. Actually, I’d say investor expectations about multiples are based on the lofty margins that proprietary software companies like MicroSoft, Oracle, and Google are able to get today that open source still struggles to achieve (I was surprised to find out a couple years ago that Red Hat has comparable margins but on much smaller revenues). I know that you’d like for the low to negligible returns of open-source business models to signal that the software industry has matured into the lower-return strata but the aforementioned proprietary software companies’ obscene margins, and many more smaller proprietary software companies’ high margins too, put the lie to that notion. The real issue is that open-source does not have a viable business model, aside from the relatively small niche of consulting-ware. Companies like IBM, Sun, and Red Hat can customize open source software and sell such consulting-ware into the enterprise market, but the notion that open source can ever succeed for product software like the desktop OS or other desktop software has long been shown to be unworkable (hence desktop linux has never taken off). The reasons are simple. No end user wants support for such software products, they want a product that works well for some longer period of time, 6 months to a couple years, and then is replaced. This product market is unworkable for open source because making any money off of it depends on scarcity. When there are a couple hundred linux distros and anybody can take any GPL code from any other distro and add it to their own, any possible scarcity is destroyed. The problem for the open source zealots, one of whom I increasingly suspect esr has become, is that this software product market is much more profitable than the consulting market because it scales much better. It is much easier to make a software product and copy it many times over than it is to hire new consultants when you want to scale your consulting business, as Joel Spolsky of the famous http://joelonsoftware.com has pointed out. I have commented on one of esr’s earlier posts about what I think the winning business model will be, a blend of the open and closed source approaches.

    As for the linked interview about software not generating returns from capital, this strikes me as esr rationalizing the failure of the open-source business model, not any real analysis of business models. The fact is that pumping money into capital-intensive businesses used to be the main way to get returns in a manufacturing age. In the arriving information age, you pump money into labor, information workers’ labor. This is how successful companies like Google, MicroSoft, or Oracle produce outsized returns, their traditional “capital” equipment is a minimal cost, even for companies like Google or Microsoft that are building giant server farms, because the real “capital” is now the information worker. The truth is that returns come from what is scarce or has value and it used to be capital equipment-intensive manufacturing; now it’s labor-intensive information work. Of course, as esr points out, most investors will fail at these new ventures as it takes intelligence and study to succeed at something so risky and most investors simply want to throw money in, sit around being lazy, and hope that returns pop out of the black box company that they invested in. It is interesting that the VC model has largely failed in the last couple decades for precisely these reasons: VC firms were built around investing large sums in large capital-equipment intensive businesses like semiconductor fabs or PC factories but have not made the transition to a proper investment model for labor-intensive software. Regarding Mike Earl’s point about the optimal organizational structure for software/information worker businesses, it’s interesting that I reached the same conclusion some time back without making the analogy to law firms or other partnerships but simply by thinking about what the optimal structure would be for information work: a flat structure where information can come from everywhere and where profit-sharing is the main incentive rather than stock options.

  6. Assigning some value to an occupation based on the perceived level of individual accountability doesn’t happen very often. For example, The professional (and economic) equivalent of practitioners of law, medicine, investment banking, physical engineering (civil, mechanical, aeronautic), etc is often exceeded by athletes, entertainers, and other similarly blameless individuals.

  7. I’d actually agree with Randy at least so far as to admit that we don’t currently do a very good job of distinguishing between nigh-useless, marginally adequate, talented, and superstar programmers. Such differences exist, but when they aren’t generally recognized it’s clear that the management of programmers will be necessarily deranged.

    And I note that in the professions Craig mentions we *do* recognize and pay huge premiums for the top talent…

  8. > No end user wants support for such software products, they want a product that works well for some longer period of time, 6 months to a couple years, and then is replaced.

    This could be seen as an opportunity for open source. A lot of the time building the disposable gadget is cheaper and faster if you can use open source software to do it rather than license proprietary software, let alone develop your own. Nokia has stated that this was the case for the N800 (and whatever the predecessor model was). And at least initially they did not release software updates for it and the community complained rather loudly. That should probably not be taken as an indication that there is a market for updates, though.

    Somewhat off topic: ESR has written that Microsoft is dependent on the level of profits allowed by its monopoly. Have the folks here read the Wired piece on Ray Ozzie (http://www.wired.com/techbiz/people/magazine/16-12/ff_ozzie)? I do not follow Microsoft news at all and I had not realized how much of an outsider he was and how long he has been there. Do you think he will be able to save MS from itself? (I realize MS may not currently look like it is in need of saving.) He seems to be a brilliant man and to be actually doing some things differently from what has been the Microsoft way. At the same time it looks to me like their ‘Live’ thingies are unlikely to be in a monopoly position and they are the sort of low-margin businesses that ESR writes about.

  9. It seems that some industry watchers just can’t shake their software-in-a-box fetish. Some of my peers still think I’m an idealistic hippie for advocating open source.

  10. Until software developers are held to equivalent levels of individual accountability, don’t expect to see them considered and treated as the professional (and economic) equivalent of practitioners of law, medicine, investment banking, physical engineering (civil, mechanical, aeronautic), etc. — Randy

    ————————————————————————————————–

    Somewhere along the same lines as barbers, beauticians and interior decorators? Even funnier is that you included investment bankers. Which is odd, it didn’t seem to help Lehman Bros any! Professional associations end up being quasi unions and gate keepers.

    Test for you. I presume you have a doctor you see regularly. Have you inquired if he has ever killed someone in his line of work? You should but the doctor is not obligated to divulge that data. Fact the AMA in most states end up placing any inquiries as to medical malpractice into sealed records. It is impossible to know the competency of your doctor. But I can query a BBB database and see if my prospective mechanic has any complaints against him.

    Knowledge, application and integrity are not earned by holding some certificate. The integrity you seek started long before that him or her earned the first degree. The knowledge you seek maybe assisted by training but is generally learned by doing. The application you seek is earned by the dint of hard work.

  11. No end user wants support for such software products, they want a product that works well for some longer period of time, 6 months to a couple years, and then is replaced. This product market is unworkable for open source because making any money off of it depends on scarcity. — Ajay

    —————————————————————————————————

    At the bottom end of the scale you would be correct. If I walk in to BestBuy and purchase a OEM license on a HP Presario your model holds. But in that case the OEM has taken on the support role to the customer. But on the enterprise end of the scale I can assure you maintenance contracts for either Red Hat or Microsoft are live, well and booming. I spent a third of my time negotiating them or pricing them for a Fortune 10. Companies WANT the support to mitigate risk both for themselves and their downstream customer base.
    ————————————————————————————————

    As to capitalization comparison of Mfging vs IT. 10 years ago doing such comparisons were a worthwhile endeavor for the MBA types. But with services like S3/EC2 and GAE it is common to see 2-3 person teams starting up services with no capital equipment at all. They just need a runway of cash to carry the labor till the income stream matches. Which to the accounting types is all expense.
    ————————————————————————————————–

    As to Open Source, funny I never viewed it as a business model. That some have looked at it that way is probably misdirected. That Red Hat can make a go of it is kudos more to their business acumen than what they are selling. When you step back and look at applications like Bind, Samba, Python and others, their creation was to solve a technical need not to map to a balance sheet. They have stayed true to that mission statement, are successful, and in many cases represent the best in programming as art.

  12. “Companies like IBM, Sun, and Red Hat can customize open source software and sell such consulting-ware into the enterprise market, but the notion that open source can ever succeed for product software like the desktop OS or other desktop software has long been shown to be unworkable (hence desktop linux has never taken off). The reasons are simple. No end user wants support for such software products, they want a product that works well for some longer period of time, 6 months to a couple years, and then is replaced. This product market is unworkable for open source because making any money off of it depends on scarcity. ”

    JohnMC, if you’re going to quote something to make a point about how it’s only true if blah, blah, blah… please ensure that they didn’t make your point in their discussion. It can sometimes seem like you didn’t read the post, but you probably just read through everything else and forgot the rest of his post. Ajay’s comment that you quoted was exactly for the desktop user, not Enterprise.

  13. BTW,Eric! You were supposed to publish a paper on c++ as you anounced at one of your earlier posts. How’s that going?I googled that but I couldn’t find anything? Is it published?

  14. >BTW,Eric! You were supposed to publish a paper on c++ as you anounced at one of your earlier posts. How’s that going?I googled that but I couldn’t find anything? Is it published?

    No. My coauthor had to fly home before we could finish it. We may get to work on it more next months, he’s considering flying up for Christmas.

  15. The article Jim Thompson links to doesn’t make a bit of sense.

    Open source is a development model/copyright model. Claiming a development model can’t do things unrelated to development is outright silly.

    “An MSDN subscription can’t roast a Christmas goose.”

    It is clear that people hosting services can and do use open source software. The article even mentions a few!

  16. Mike said
    There’s quite a difference between a business model and open source

    esr said
    Mr. Cohen also brings us the news that running a telephone support desk is a thin, low-margin business; and that successful companies like Red Hat find other ways to add value to their open-source codebases.

    I was thinking… It is possible, in some situations, to charge a lot of money for telephone support. Those situations do not include the end-user desktop/laptop support desk Mr. Cohen seems to have in mind.

    The real money to be made in support service comes from supporting more important machines than desktops and can include on site support for hardware as well.

    Of course several companies provide support with incredible response time and get paid very well for it.

  17. Hmm. Telephone/webcam support for non-computer related issues pays well. See all the internet porn sites.

    Wonder if that means that http://www.nerdrotica.com could actually be more than a parody some day…

  18. >There’s quite a difference between a business model and open source

    To be honest, I think this is an evasive and unhelpful style of response. It is clear what Cohen means, and easy enough to refute him on the merits. We don’t have to resort to ducking the criticism.

  19. ESR,

    I have an idea for a blog post for you: compare open vs. closed source to the differences in technological innovation between different economic systems. Up to about the sixties, the technological advantage of capitalism over socialism lied not in invention but innovation: driving down the cost of production and making the new products affordable for the masses. In invention, socialism could match the achievements of capitalism: think Nazi rocket science or Soviet successes in the Space Race. Capitalism only gained a distinct advantage in invention with the advent of electronics and computers. And this is exactly the area where the gap widened more and more. The failure of Soviets to get a man on the Moon wasn’t in classical sciences like propulsion, but generally in electronics and computers. The MIG-29 is better in the classical sense – aerodynamical properties – than anything America came up with, they lost the arms race not in these things but rather in the electronics and computers onboard the plane and in the missiles. Electronics and computers required a new, different kind of thinking about invention, and that sort of thinking just doesn’t work well in a command economy, it requires freedom much more than classical sciences. Perhaps this parallel can say something about the differences between closed and open source as well.

  20. >In invention, socialism could match the achievements of capitalism: think Nazi rocket science or Soviet successes in the Space Race.

    I’m nowhere near convinced that’s true. One of the dirty little secrets of the Soviet strategic-rocket program was how heavily it relied on American tech acquired through grey-market channels – I don’t know that this was also true of their space program, but it seems like a plausible guess. And as for the Nazis…it’s generally forgotten now (because of course the Nazis never put it in their propaganda newsreels), but the WWII Wehrmacht relied considerably on horse-drawn transport. Yes, *horse-drawn*. Overall the tech level of the Allies remained considerably higher throughout the war; it wasn’t the Nazis who deployed radar or the atom bomb. They never even managed to fully mechanize their infantry.

    The only “advantage” the socialists had was the ability of dictators to throw huge amounts of resources at the mad-genius engineers behind boondoggles like the V1 or booster guns – resources that would have been better spent on conventional weapons. Both the SU and Nazi Germany, as a result, had an odd combination of overall lower tech level than their adversaries with a handful of eccentric prestige projects that made their systems look more inventive than they actually were, overall.

  21. > >There’s quite a difference between a business model and open source

    > To be honest, I think this is an evasive and unhelpful style of response. It is clear what Cohen means

    Let’s see what Cohen says:

    > The open-source business model is broken. Companies have long hoped to make money from this freely available software by charging customers for support and add-on features.

    It seems to me that Cohen means that (1) there is only one open source business model (“the open source business model”) which is demonstrably false (compare the open source strategies of IBM, Apple, Sun, Sun’s MySQL division, Redhat, Novell, etc.) and (2) while Redhat executes that business model well, Sun does not.

    Newsflash (for Cohen): Sun isn’t looking for the same business model as Redhat. Sun’s involvement in open source — whether that’s Java, GNOME, MySQL, or some other project — is largely to sell more Sun hardware. This, of course, was explained in The Magic Cauldron.

    So, with all due respect, I believe the original comment was accurate when applied to Cohen’s piece. More importantly, the original comment appears to be aimed at a blog post that Cohen didn’t write ( http://www.xobni.com/asmith/archives/80 ). And that blog post includes:

    > What Open Source Can’t Do … Host software services. …

    > I wonder what percent of blogs today are ran in the old fashioned way: someone went to SourceForge.net and installed it on their shared server, versus using a company set up to host the blog for them. I’d bet in favor of the hosted service.

    > It’s too bad, really. There’s no doubt we’d have better software today if open source projects could get servers and operational resources for free. We’d probably all be using an IM client built around Jabber.

    And, frankly, it isn’t easy for me to figure out what the blogger means, other than perhaps “free-wheeling hippies don’t understand business.” He even points to Fog Creek hosting a service built on open source software!

  22. Some have noted that open source is a development model and not a business model but obviously it can be both, to the extent a company tries to use this development model to pay their bills. The question is how do we advance this development model the most, to the extent that we believe it is the best development model? And obviously that means finding a great business model to fund development, without a business model you’re just gearheads spinning your beanies for fun. My proposal is to accept the fact that scarcity can only come from closed source and work closed source into the development model where necessary. I have in mind the outline for a software license that I think would be best for this type of mixed-source development and I can share it here, or over email with esr if he’d prefer (you have my email address). The only reasons I haven’t published this idea for a license so far is that I need to research one last issue regarding reverse-engineering, which you could perhaps help me with, and I’ve felt that publishing it on a blog someplace wouldn’t lead to much attention or uptake (I looked into contacting OSI about it some time back but forget if I ever sent out an email). However, I’m sure esr could get a lot of people listening if his bullhorn were employed towards this end, assuming he finds the idea worthwhile. As one of the few economically literate people I’ve found in tech (esr’s post on the telco disgrace was spot on), I hope esr will consider what I have in mind or at least poke holes in it if there are any. Nobody can be certain about a model that’s never been tried but I’m pretty sure, based on fundamental principles, that this mixed-source approach will end up being the winning development/business model for software.

  23. The only “advantage” the socialists had was the ability of dictators to throw huge amounts of resources at the mad-genius engineers behind boondoggles like the V1 or booster guns – resources that would have been better spent on conventional weapons. Both the SU and Nazi Germany, as a result, had an odd combination of overall lower tech level than their adversaries with a handful of eccentric prestige projects that made their systems look more inventive than they actually were, overall.

    I’ve been looking for years for a way to show that sort of incentivization in a set of game mechanics.

  24. Ken: I can’t speak for the mainline Civilization series, but in Alpha Centauri it is very easy to end up going down one of the main four tracks too far and, say, being very advanced in your city enhancements but only have access to the local equivalent of pointy sticks to arm your units with. You can even be advanced defensively while lagging offensively, to some degree.

  25. I seem to recall that Masters of Orion had six completely independent technology tracks. But I was in elementary school when I last played it, so I could be mistaken.

  26. >>BTW,Eric! You were supposed to publish a paper on c++ as you anounced at one of your earlier posts. How’s that going?I googled that but I couldn’t find anything? Is it published?

    >No. My coauthor had to fly home before we could finish it. We may get to work on it more next months, he’s >considering flying up for Christmas.

    Hmmm, it might be fun to do it as a Socratic dialogue.

  27. >The only reasons I haven’t published this idea for a license so far is that I need to research one last issue regarding reverse-engineering, which you could perhaps help me with,

    Send me email.

  28. Jeremy – the trick I’ve got is that for the games I build, the “government type” needs to alter your strategy about as much as, say, choice of a homeworld does in Race for the Galaxy. This is for boardgames, so the grind it out model of Civ isn’t optimum.

  29. Historical correction: The Germans used horse drawn transport because they were extremely short of petroleum, not because they were tech laggards. The Germans pretty much matched the Allies in tech: behind in some areas, ahead in others. They had radar – not as good as British/American radar, but good enough that the Allies invented “window” to neutralize it. They wasted resources on mad-genius projects, but then so did the Allies, in effect: the Manhattan Project succeeded too late to affect the war in Europe. One big mistake they made was investing heavily in infrared optics rather than radar, but that was not a system problem.

    Their real failure was in industrial organization: Germany produced much less war material than its raw industrial capacity should have achieved. And of course in grand strategy – taking on the enormous resources of the British Empire, the USSR, and the USA all at once pretty much guaranteed defeat, regardless.

  30. My suggested historical correction: the Germans used horse-drawn transport because they did not mechanize their army before they launched WWII in Europe and were unable to do so afterward because they lacked sufficient industrial capacity (at least, did not exploit such capacity as they had to produce motor transport). The Germans “conscripted” all the civilian motor transport they could (especially from conquered regions) to support their logistic effort, and they still fell far short of their military needs. I have many sources for this, but see Van Creveld, “Supplying War”[1] for technical information, and Siegfried Knappe, “Soldat,” for a fascinating narrative of the WWII German army’s dependence on horse-drawn artillery and supply transport.

    [1] Van Creveld’s work has been criticized from various angles, but so far as I am aware, his description of the logistics of Germany’s WWII Eastern campaigns is well respected.

  31. Marco: if it’s in the NYT, it’s surely part of a Gramscian agitprop campaign hatched by the far-left to discredit capitalism. Or something :-)

  32. >Well, Eric summarily dismissed my ideas, without seeming to consider them much

    No, I considered them pretty carefully, with an eye on the experience of L. Peter Deutsch and others who have tried similar hybrid licensing schemes. All have failed to gain traction. In fact, the trend of the last five years has been away from complicated hybrid licenses (including various derivatives of the Netscape Public License) towards the classic four (GPL, MIT, BSD, and Artistic) plus the Apache Public License.

    One substantial reason for this is that complex hybrid licenses are difficult to reason about. So, even if you can find a developer population that thinks your proposal solves a real problem, it’s unlikely to be adopted because it creates too much uncertainty about both the scope of granted rights and how it mixes with other licenses.

  33. OK, I assumed that you didn’t consider my ideas much because you only emailed me the same reason: that hybrid licensing hasn’t worked in the past so it won’t work today. I would quibble with your characterization that the recent licensing trend has been so broadly towards the classic licenses, as probably the biggest open source effort of the last couple years has been OpenSolaris, whose CDD License is a Netscape Public License descendant. As such, Solaris can be used in a hybrid manner today. You can write binary drivers or rip out their scheduler and swap in your closed-source scheduler into OpenSolaris and it’s compatible with the CDDL. I think the only reason this isn’t being done already is ignorance.

    I agree that hybrid licenses are more complex, but I think they’re necessary to mix the benefits of open and closed source. Better to try and fashion this more favorable melding than to continue using the simple tools of the past, simply because they’re already lying around. I think my formulation articulates a hybrid license that could work well and that is no more complex than it needs to be, so obviously I think the mixing can be done and without too much needed complexity. As for license compatibility, the GPL already is incompatible with most licenses yet people still use it, probably cuz of its viral attributes that cause it to dominate, causing many holdouts to give up and license as GPL also just to go along. This creates a licensing monoculture that isn’t good, plus I predict that all GPL code will have to be ditched, as it won’t be able to compete with hybrid-licensed code and of course its obnoxiously viral license is incompatible with most other licenses.

  34. >I agree that hybrid licenses are more complex, but I think they’re necessary to mix the benefits of open and closed source.

    But I’m not sure I see any significant benefits of closed source. The only one worth even bringing up would be the ability of developers to capture secrecy rent in order to fund larger development projects than they could otherwise, and that’s only a benefit if it looks like software would be underproduced without it. Sorry, but after the last decade that’s a very difficult case to sustain.

    You should understand that as open-source developers and thinkers go, I’m a relatively easy sell for hybrid licenses — I at least acknowledge that secrecy rent might not be under all circumstances a bad thing. Since I am pragmatically unconvinced that you are addressing a real problem, how do you think you’re going to get anywhere with hackers who are much more ideologically fixated against secrecy rent?

  35. Yes, we’re agreed on the benefit of closed source, it is a secrecy rent. However, secrecy is merely scarcity, ie property rights, for digital goods, which can be copied at minimal cost. In economics terms, digital content has the same fixed cost to produce but almost no marginal cost to distribute. However, you do have to recover that sometimes large fixed cost and the only way to do so is by enforcing some form of scarcity, which is likely to be some form of secrecy. This secrecy model has worked great for the proprietary software companies, who have spawned some of the richest people in the world. However, we’re now stuck paying a software tax to Gates and Ellison every year, because they maintain full proprietary control over their software. In contrast, open source has been a resounding failure monetarily, though it has spawned some interesting projects because of its open development process. The goal of my hybrid license is to combine the property rights of closed source with the open development process of open source. I believe that’s the ideal licensing situation, allowing us to fund an open development process with the money from closed source.

    As for the success of open source in producing software, I think the record speaks for itself. Almost nobody uses open source software compared to closed-source software, open source has been a resounding failure in how much software it has produced compared to closed source. If you think it’s just a matter of time, for users to get used to open source, keep dreaming: it won’t happen. I agree that you should be one of the more inclined of the open source crowd, that’s why I approached you. My goal is not to convince the open source zealots however, but to convince the pragmatists, who dominate the zealots of all camps in numbers, that this is the best way to license software. I had hoped that you were one of the pragmatists, but I’ve come to realize that you may have become as much of a zealot in your own way as Stallman.

  36. >As for the success of open source in producing software, I think the record speaks for itself. Almost nobody uses open source software compared to closed-source software, open source has been a resounding failure in how much software it has produced compared to closed source

    Nonsense. You can maintain this only if you begin with ignoring every market segment other than end-user desktops, then continue by ignoring Firefox which has pretty respectable share even there.

    You damage your case by beginning with the assertion that open source is Doomed Unless Something Is Done. This sounds like groundless panic-mongering to me, every bit as meretricious as Barack Obama insisting that the U.S. economy is doomed unless we pass his mile-high mountain of pork. It leaves me disinclined to trust your representations in other areas.

    There may be a rational case for the kind of license you advocate. But I don’t think you’ve made it yet. And if you can’t convince me, you might as well give up on convincing the open-source community. Try harder – which does not mean yelling “We’re doomed!” at higher volume. In fact, if you can find a way to discard that claim entirely I’d advise it.
    .

  37. While you may like to brand me as a mere panic-mongerer and summarily dismiss me as a result, I think I’ve made a more specific argument than that, which rests on two points. One is that open source has failed against closed source. If we multiply lines of code for each program by the number of minutes a user uses it and do a tally for open and closed source, open source would lose resoundingly: it’s tally would be a rounding error. The second is that hybrid-licensing will kill both pure open and closed source. So your entire premise is false, as my goal is not to save open source from Impending Doom but to replace it with something better. I’m sorry I haven’t convinced you yet, I’d like to try (again, I don’t care about the open source zealots who largely make up the community). But I don’t see how to when you summarily dismiss my conception with “something similar has been done.” If you can pose some specific questions or weaknesses, I’d be glad to discuss further. :)

  38. >If we multiply lines of code for each program by the number of minutes a user uses it and do a tally for open and closed source, open source would lose resoundingly: it’s tally would be a rounding error

    Again, this is complete nonsense sustainable only by ignoring everything but end-user desktops. The more insistently you repeat claims like this, the more tempting it is to dismiss you as a loon.

    >The second is that hybrid-licensing will kill both pure open and closed source.

    Given the number of developers that write open source for fun, nothing short of a civilization-wrecking catastrophe could kill it. But supposing you could – what makes you think you can appeal to the open-source community by promising to destroy a form of cooperation they value and enjoy? This seems remarkably tin-eared.

    >But I don’t see how to when you summarily dismiss my conception with “something similar has been done.”

    Go study the history of Ghostscript and Aladdin Software. When you can explain to me why your “hybrid license” is different enough from theirs that it can succeed where theirs failed and was eventually withdrawn, you’ll be ready to make a sounder argument.

  39. Regarding closed source dominating open source, I counter that only a complete zealot would claim that open source has gained significant share in any broad market, consumer, enterprise, server, whatever. I don’t think cooperative coding will be destroyed, as I have claimed that hybrid licensing will still have a relatively open development process. Rather, my claim is that the pragmatists will jump to a development/business model that allows them to make money and still take part in an open development process. The resulting software will soon outclass both pure open and closed source and all but the open source zealots will move to an open development process that actually has a business model, hybrid licensing. So the goal isn’t to destroy open source cooperation, but to merge it with the property rights of closed source. Nothing would stop someone from solely focusing on the open parts of a mixed-source codebase and releasing all their source modifications openly, if they chose. My goal with hybrid licensing is to still have a bazaar but one with limited property rights, not a commons as the GPL enforces today, where you can show up to the bazaar the next day and someone else has taken your stall. ;)

    As for the Ghostscript situation, I was a little unclear on how that was a hybrid license at all, as it seems to be more viral than the GPL even, until I read this great clarifying interview with Deutsch. I wouldn’t call his AFPL a hybrid license, as it requires all source for a program to be openly released just like the GPL. However, by additionally requiring that all other software that is bundled with AFPL software be open-sourced (unlike the GPL), he did force the authors of proprietary Ghostscript frontends to license his code so I suppose it’s vaguely similar. I don’t know that he failed either, he does say in the linked interview, “At this point, I have made enough money from commercial Ghostscript licensing that I can retire.” That was in 1998. I suppose the AFPL could be considered a first step towards hybrid licensing, only the granularity is much coarser because the proprietary components are completely separate software that are only bundled together. My goal with my hybrid license is to have a software binary that is 50+% open source, with the remaining closed-source components of this same binary licensed from various contributors, so the granularity of this mixing is much finer. Unless you can provide a cogent argument for why the AFPL can be considered a failure, I actually consider the fact that Deutsch made a fair amount of money from Ghostscript, while funding the development of an open source product, an indication that this hybrid approach is a very promising line of thought. :)

  40. >Regarding closed source dominating open source, I counter that only a complete zealot would claim that open source has gained significant share in any broad market, consumer, enterprise, server, whatever.

    Your ignorance of the facts does nothing to make me into a zealot, however much you might wish it were otherwise. Now go look up Firefox’s market share, and learn about the trend over time. (There are many other good examples, this is just the one for which the research is easiest.)

    >Unless you can provide a cogent argument for why the AFPL can be considered a failure

    You’ve made a good start. Now continue your research and discover whether or not AFPL was ever used for any project besides Ghostscript, what eventually became of it, and why.

  41. Heh, yes, your evidence of Firefox’s insignificant 5-10% market share really proved me wrong. ;) I’m not really interested in researching the AFPL, which strikes me as merely the GPL taken one extra step in the hybrid direction, albeit by making it more viral. If you already know why the AFPL wasn’t more widely used, feel free to share but I suspect it’s only because the model where you write a open backend and then wait for somebody to write a proprietary frontend is not widely applicable, plus the groupthink in the open source community that leads them to cluster around the most viral license, the GPL. I have repeatedly given arguments why hybrid licensing is more evolutionarily fit, while you don’t really make an argument against the concept but keep repeating that the AFPL is some sort of social proof of the fact that hybrid licensing is doomed, even after I showed how successful the AFPL has been. Clearly, you cannot be convinced rationally, as you’ve gotten so attached to the conception of open source that it is beyond reasoning for you. That’s fine, I predict that you will soon be disappointed however as mixed-source will kill off both open and closed source. I don’t need to convince you of that, events will prove me right.

  42. >Heh, yes, your evidence of Firefox’s insignificant 5-10% market share really proved me wrong. ;)

    Guess again. Uncooked estimates are pushing 25% at this point, with share rising so fast that the Firefox dev team is publicly worried about becoming the next monopoly.

    But there is probably no point in arguing this with you, as you have revealed yourself to be a a crank. “mixed-source will kill off both open and closed source” and “events will prove me right”. Want to try for “Bwahaha! I’ll show them, I’ll show them all!” next?

    Best of luck with your license. Not that luck will help…

  43. Sigh, you really show how close-minded and silly you are with the silly insults, Eric. As for firefox, 25% market share is only in the firefox dev team’s dreams and I for one am a long-time firefox user that will be switching to google chrome as soon as it comes out for FreeBSD later this year.

  44. Although our discussion appears over, I just heard back from Mr. Deutsch with his emailed feedback, that I solicited, so I asked him if I can publish it here and he said it would be okay, with one obfuscation. I have made that change and I reprint most of his email verbatim below, followed by a few comments of my own:

    I retired from active involvement in computer science in 2003 and am not interested in an extended discussion of these issues, but I do have some comments about the AFPL experience. Please feel free to pass them along to Eric Raymond, and give him my greetings.

    The AFPL was designed in the early 1990s to solve what I saw as the “free
    rider” problem. The GPL is based on a cultural regime in which those who
    use others’ contributions implicitly acknowledge a moral obligation to
    contribute to some extent themselves. (For a fascinating discussion of some related issues, see Lewis Hyde’s book “The Gift”, which I just discovered recently.) But the GPL includes no mechanism for enforcing or even encouraging this behavior. The AFPL tried to incorporate two mechanisms for doing this:

    1) It forbade distributing AFPL’ed software if any payment was involved, to
    prevent including it “free” with commercial products. (This was, in
    fact, what led to the creation of the AFPL in the first place: rampant
    inclusion of Ghostscript “free” with commercial fax software.)

    2) It required actual distribution of source code with binaries, since
    commercial distributors of GPL’ed software often obscure recipients’
    source code rights or make them difficult to exercise.

    While the AFPL had some success at curtailing free rides, it also prompted
    quite a bit of ill feeling in the Free Software community, and the “free
    rider” problem eventually dropped to insignificance as the commercial
    Ghostscript business* became better established. So sometime between 2000 and 2005 (I don’t know exactly when, since I was no longer involved), the new copyright holders switched back to the GPL, with no significant ill
    effects.

    –lpd
    —————————————————————-
    * Ghostscript — the same code — is also licensed commercially as an OEM
    product under a conventional commercial license. This is consonant with
    my philosophy that the recipient of code should be able to choose which
    set of rules they want to play by: open source, with no payment but an
    implicit obligation to contribute to the open source pool, or commercial,
    with the right to close the source but an obligation to pass along some of
    the payment they receive. This dual-license model has generated, by now, tens of millions of dollars in revenue that (among other things)
    supports a team of good engineers improving and extending the code,
    without impairing the rights of those who want to use the GPL release.

    Ajay’s comments – Mr. Deutsch’s statements seem to support my assertions that his license was very successful but was only retired because of free software groupthink. His company essentially moved from the AFPL hybrid model to the dual-licensing model employed by MySQL, TrollTech, and the few other successful L/GPL-licensed product software companies. However, this GPL dual-licensing model has problems, as it requires copyright assignments and the business model isn’t transferable: I can’t take MySQL code and dual-license it as I don’t own the copyright. My hybrid license aims to fix this problem, among others.

  45. >But I’m not sure I see any significant benefits of closed source. The only one worth even
    >bringing up would be the ability of developers to capture secrecy rent in order to fund
    >larger development projects than they could otherwise, and that’s only a benefit if it looks
    >like software would be underproduced without it. Sorry, but after the last decade that’s a
    >very difficult case to sustain.

    I might point at Apple OSX as the obvious example of a hybrid open/closed source that has not been replicated in open source. To get a reasonable (read any) return on investment on the attention to detail that is unsustainable in open source projects appears to require secrecy rent. Nitpicky stuff is rarely an itch that desires scratching. Neither is documentation, etc.

    There is no way that Apple could ever open source it’s UI, cocoa API, etc because that’s its primary competitive advantage over just running NetBSD with Gnome. Nor could OSX have every come to being based on a GPL license. The benefit to me for closed source is finally a desktop unix that doesn’t suck and (mostly) just works.

    Dual licensing using GPL is also a mechanism for secrecy rent. GPL effectively precludes everyone but you having a competitive advantage via a pro/enterprise version based on closed improvements or having commercial licensing fees. Compare commercial license revenue stream that MySQL has generated vs PostgreSQL.

    So arguably, dual-licensing under GPL is a hybrid open/closed license for the corporation that owns the core IP for that code. The number of projects that use GPL like this probably exceeds MPL projects but is harder to easily count as “secrecy rent” based (whether in whole or part). Likewise, you can’t ignore open source based ASPs like Google that mix open source with proprietary code in house as a variation on a hybrid environment that also depends on “secrecy rent”…although in Google’s case much of that secret sauce is their expertise in managing huge server farms vs their closed source repository. Still, Google hasn’t released their internal code base and derives some competitive advantage from it vs say Microsoft. It isn’t ALL scalability expertise.

    The problem with hybrid licenses is from a corporate standpoint, the weak copyleft buys you very little vs GPL or simply just chucking it out there as Apache (or any open source license with a CLA). It does buy you friction with the rest of the open source world unless you’re already operating in a MPL or CDDL ecosystem.

    As far as Firefox goes, while it does have significant share, without the Google revenue stream I sincerely doubt it would be much of a contender vs webkit based browsers or IE. It is relatively easy to be successful as an open source nonprofit when a company is pumping $50M or so into your coffers every year. Given webkit, I really doubt that the firefox devs have any need to worry about any imminent browser dominance.

  46. >I can’t take MySQL code and dual-license it as I don’t own the copyright. My hybrid license aims to fix this problem, among others.

    That’s not a bug. That’s a feature. You don’t dual license something GPL because you wanted to allow some downstream company to beat you senseless in the marketplace with your own IP.

  47. Nigel, I agree with most of what you say, except for where you downplay the hybrid licenses. I think we’re agreed that the first tentative steps in the hybrid direction have been somewhat successful, whether Apple or Ghostscript or MySQL. However, there are problems with these tentative approaches that prevent them from really taking off. With Apple’s approach, since they’re taking BSD code, the problem is that there’s nothing stopping them from completely closing all the BSD source at any time and never releasing again. With the dual-licensing approach of Ghostscript or MySQL, only those companies can ever make money off commercial licensing and they’re unlikely to ever hand over that copyright to anyone else, as you note. However, I envision a model that’s better than either of these hybrid approaches, without the aforementioned dual-licensing bug.

    With my hybrid license, a codebase will likely have 50%+ open source code, most likely infrastructure code, with the remaining code consisting of closed source modules. The consumer knows that they will get the source for the entire binary within a couple years, so the codebase will never disappear into a black hole. The writers of closed-source modules know that they can profit off their work for a set period of time, but if they stand still and don’t keep adding to their code, someone will clone their module or grab the source in a couple years and pass them. The writers of the open source infrastructure code know that anybody who adds source to the codebase will be forced to open it in a couple years. Also, one could always fork the open infrastructure code, just like with open source, and then license the closed source modules for the new fork or recreate those modules themselves. This provides the optimum forking model: where you can fork a codebase AND sell it but are constrained by the fact that the programmers of the closed-source modules may not want to license their code to your fork, unlike with GPL code where there are essentially no constraints to forking and we have a hundred distros as a result. I believe that these varied benefits will give consumers and open and closed source programmers what they each want and they all can prosper symbiotically.

    For corporations, the benefit over existing open source licenses is that they will have closed source modules written independently by third parties, who will actually get paid by whoever is using them. The current alternatives are that the corporation pays for the work themselves as an open source patch to a GPL codebase, in which case the corporation pays all the cost and gains no competitive advantage if they’re shipping binaries with GPL source, that someone like Red Hat implements it for their consulting model, or they hope that some random person does it for free. For a BSD codebase, one can write closed-source modules but this increasingly turns the BSD codebase closed, as one sees in practice. My hybrid license combines the best features of both the GP and BSD licenses, the ability to add closed-source modules like the BSD license and the virality of the GPL, appropriately weakened to only take hold after a couple years. I believe that this will prove to be the best combination, allowing for the property rights of closed-source modules and the open development process and consumer guarantees of open source.

    I agree that hybrid approaches cause friction with the open source community, but this is limited to zealots like Stallman or Raymond, who take an ideologically pure but impractical approach, and their followers. This fringe can be safely ignored as pragmatists are the majority in every camp, including the much larger closed-source camp. We can go after the pragmatists of both camps with this hybrid approach and leave the zealots to their dogmas.

  48. I find that calling people zealots tends to increase friction…

    The hybrid model you propose will be liked by some but I think not enough to get critical mass. But what the heck, write it up and submit it to OSI. If you can refrain from calling everyone who disagrees with you a zealot you might get good critique back and it probably wouldn’t run afoul of anti-proliferation issues since I’m not aware of another time based open source license.

    I’m thinking the timeout scheme is too annoying unless my software repository automagically releases code after the timeout period. And it better be a large trusted entity that holds all code in escrow so even if the business disappears the code will always get opened at that later date. I might trust GoogleCode for this. Might. Any lesser site? Not too likely.

  49. I think it’s better to accurately label the irrational responses of zealots than to hope to curry favor with them by diplomatically ignoring their crazy critiques. I have painstakingly tried to convince esr that this is a valid line of thought and only labeled him a zealot when he showed no inclination to actually consider these issues. I’m not too worried about getting people to use my license over the long-term, as I believe it’s plainly better and will win on its merits. Over the short-term, convincing people will help get it started but the response so far has been lukewarm. That’s fine, I realize that the best way to bootstrap something like this is to build a showcase project that’s hybrid-licensed, demonstrating these benefits rather than just talking about them. Most people cannot reason whatsoever, so everything has to be demonstrated to them. I already have in mind what that showcase project will be and am content with talking up my license for now, until I have the resources to start that showcase project, so that others aren’t denied the possibility of creating a similar license in the meantime. As for submitting to OSI, I figure there’s no point as they’ve become as dogmatic in their own way as the FSF. Take for example this guy and the problems they’ve been giving him about accepting a similar time-limited license, where he simply took an OSI-approved license and added a single line about a time delay for source release. As for companies disappearing, people keep mentioning that but I don’t see it being a big deal. If we find companies disappearing to be a problem in practice, consumers can simply insist that the companies place the source code in escrow for later, which is what I told others with the same critique and as you yourself note.

  50. Ah, I never bothered to join that mailing list when it split from license discuss and my mailbox has been happier for it.

    However the code escrow problem is significant and not hand wavable. Perhaps you and zooko can put together a GForge that has been modified for automatic 12 month release after check in. You may have to insist under the license terms that the code is escrowed there as part of the agreement.

    Extra bonus points if the source is encrypted until that 12 month expiry.

  51. Eh, that’s like step 25. First step is to get these ideas written up by lawyers who know what they’re doing and to release a showcase project under this new license. It’ll take awhile for me to get the resources to personally do those first two steps, so I’m just soliciting feedback and help for now. If nobody else runs with this in the meantime, I’m fine with just talking about it till I can bootstrap it myself.

  52. Ajay has the right idea. Work is work; work deserves to be paid for. At the end of the day you have to monetize something, and the most obvious source of income is the software product itself.

    Consider Mac OS X, the most successful desktop Unix ever. Its development is based on a hybrid model somewhat similar to Ajay’s: the kernel, BSD userland, and other bits are open source whereas the GUI and most of the “secret sauce” which makes an Apple product an Apple product (and therefore insanely great) are closed source. This has worked out spectacularly for Apple, end users, and developers, and today Mac OS X remains a superior platform for both open and closed source development.

  53. Thanks Jeff, I appreciate your support. Mac OS X has worked out great for Apple, not so great for the open source parts. People are understandably averse to contributing to an open source platform that Apple can just close up completely at any time. My goal with this hybrid license is to fashion a better compromise, where both the open and closed source contributions are respected. I think that it’s inevitable that this is the way most software will be licensed, but unfortunately you have to show this to most people, even the supposed intelligentsia, because almost nobody can reason from basic principles. In fact, if you look at what the intelligentsia at any time in history has contributed, they merely extrapolate from the past to the future: they have absolutely no ability to consider theoretical principles in figuring out when some new movement is likely to start or some old movement likely to end. This actual analysis is always left to lone voices who are left to shout at deaf ears, until what they said happens. It’s understandable why many good ideas are often rediscovered later as a result, because it’s the rare lone voice who perseveres through the stupidity surrounding them or has the drive and disparate skills to make it happen. Events will prove if I’m one of those voices, not just a crackpot like esr wants to paint me, but I’m confident that my theoretical principles and reasoning are sound and I’ve not heard any worthwhile contradictions from anybody yet (which doesn’t mean they don’t exist, merely that the people I’ve talked to so far haven’t been able to think of any).

  54. Whoops, posted that in the wrong thread, meant to post it here:

    http://esr.ibiblio.org/?p=690#comment-234671

    Regarding the debate about which business/licensing model will best serve the software market, my current idealistic thought experiment is that the Holy Grail would granulate the compensation of contribution.

    Also SOHO/consumer user’s prefer to use software for free, then only pay when they have serious economic task to accomplish with the software they’ve grown proficient in.

    I suppose there are plurality of business/licensing/marketing models, and that is good thing (a diverse free market).

    My thought experiment is about how to monetize more re-use and thus larger economies-of-scale.

    One thought is that software as a service (e.g. REST APIs), then you can charge for access, then only charge secrecy rent on the portions that you feel will maintain your competitive edge, and not inhibit the goal of more 3rd party integration and re-use:

    http://esr.ibiblio.org/?p=904#comment-234666

    One thing that might help is if projects get smaller in scope and more re-useable. Instead of donating my super-fast image scaling with bi-linear sampling code to Mozilla that used forward differences (I think did, I forgot), much better had I independently published a module that could be linked with Mozilla. Someone would develop fork tree, which could call up a version of modules as source snapshot. Some aggregators could compete on mix and match (there wouldn’t be just one Mozilla version). Does this increase complexity or reduce it? It reduces Entropy, thus there is more capacity for the system to do useful work (more permutations of order can be accomodated).

    As I said, this is thought experiment, not yet a coherent proposal.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>