Originally Published: Thursday, 13 July 2000 Author: Mark Stone
Published to: featured_articles/Featured Articles Page: 1/1 - [Printable]

Software Gates

Bruce Perens and Tim O'Reilly have engaged in an interesting and contentious debate on the notion of gated software communities. While both have thoughtful points of view, both ultimately fall prey to a common misconception about open source business models.

   Page 1 of 1  

Bruce Perens and Tim O'Reilly have engaged in an interesting and contentious debate on the notion of gated software communities. While both have thoughtful points of view, both ultimately fall prey to a common misconception about open source business models. The naive reader of these two essays could easily come to the conclusion that a commitment to open source must be a charitable act motivated by an ideological notion of what is politically and morally correct, and that this commitment inevitably conflicts with self-interested business practices. The truth is quite the opposite.

Bruce's sin is the lesser here, and at least consistent with his earlier views. So let me start there.

A gated software community is one in which the copyright holder of a particular piece of software and the licensed users of that software share source code, changes, updates, and can redistribute among themselves, but do not share or redistribute outside that group.

The problem, according to Bruce, is that this grants undue and unequal power to the copyright holder. To his way of thinking this is wrong, and not in the spirit of open source. I can't agree on either count. Frequently, though not always, the creator of a particular piece of software makes the largest contribution. There is certainly a presumption that that person has creative control over the direction of the project unless they demonstrate otherwise through lack of competence or lack of interest.

Furthermore, this asymmetry is consistent with open source licensing, and to some degree reflected in many open source licenses. The original Berkeley license, the artistic license, and the MPL all make some special provisions for originators over contributors. Certainly there is a long tradition in academic research of (a) making the results publicly available, but (b) retaining creative control over the direction of future research.

What open source does not countenance, as Bruce rightly points out, is a privileged class of users. Restrictions on making modifications to source code, and limiting the redistribution of those modifications, simply does not fit with the open source definition.

More importantly, though, gated communities simply are not in the spirit of open source. It's troubling to see Tim's lack of understanding of open source business models. Let's look at Tim's case in point: the CISpub system used by O'Reilly and some other publishers. Having worked for a couple of publishing houses, including O'Reilly, I'm familiar with this kind of software. CISpub certainly occupies a niche market. Its user base numbers in the dozens, not thousands, and that level of adoption represents success. It is written in Pick Basic, hardly a mainstream publishing language. The other software packages with which CISpub needs to interface are equally niche.

The conclusion that Tim wants to draw is that the CISpub community represents a perfect example of a community that would benefit from the gated software approach, but that would not benefit from a fully open source approach. Implied is that CISpub is typical of many niche software products, and hence there's a broader role for gated but not open source software communities. Also implied is that CISpub and others could be harmed by going fully open source.

Typically there are two instincts that incline people to resist open source; both come into play here. On the one hand is the fear that making source code freely available will spawn competitors, perhaps even, depending on the choice of license, proprietary competitors. On the other hand is the fear that making source code freely available will cause customers to defect and simply use the product for free. In a niche market, so the argument goes, neither of these risks can be offset by the marketing exposure gained from freely available software because a niche market is, by definition, just too small.

Let's consider the first point. How likely is it that another company would use CISpub's code to create a competing product? The universe of Pick Basic experts is quite small; the universe of Pick Basic experts with in-depth knowledge of publishing house IT needs is even smaller, and probably doesn't include anyone outside of CISpub and its customers. In the unlikely event that someone might think of starting a competing product, they'd have to not only figure out the technology but figure out how to make a competitive business on an open source pricing model. This really isn't going to happen. The "deterrent" value of open source business models has sufficed to limit competition even in fairly large markets (think Cygnus and C compilers).

What about the second point? Will customers flee in droves from a paid CISpub license if a free version is available? Hardly. What motivates most customers to pay for software is the need for accountability: customers want access to a technical team with expertise in the product in question, and want that team accountable for, and responsive to, problems that arise. Open source does nothing to detract from this business model, and instead just makes explicit that the real, long-term revenue model is based on service and support, and not on licensing. In fact a software product available as open source should be more attractive, not less: help is available not just from the vendor, but potentially from the community at large.

The truth is that with any large, complex piece of software, the business difference between proprietary and open source distribution is not great. Once code gets sufficiently complex, expertise in that code is difficult to come by, and resides in the oral tradition of people working on that code as much as in the code itself. Customers pay for access to that expertise, so even if you give them the code for free there's still something they need to pay for.

With open source business models there's a nice symmetry. The very features of a product like CISpub that seem to limit the advantages of open-sourcing it -- it's a niche product for a niche market -- also limit the risks of open-sourcing it. The corollary is that in large markets where open-source can be very risky -- compilers or operating systems, for example -- the gains can also be tremendous. No matter how you look at it, open source makes not just good development sense, but good business sense.

One other aspect of Tim's comments troubles me. By way of disclosure, let me say that as an employee of VA Linux Systems I have a vested interest in the success of VA projects like SourceForge. While I don't believe that colors what I'm about to say, readers may judge for themselves.

Gated software communities seem to have a lot in common with licensing schemes like Sun's Community Source License, an approach that Sun has repeatedly and unsuccessfully tried to wrap in the colors of open source. Sun, I note, is a customer of Collab.net, and Collab.net figures prominently in Tim's discussion of gated software. What isn't mentioned in Tim's discussion is that O'Reilly is a major investor in Collab.net. Could it be that the need to protect his investment has given Tim second thoughts about his commitment to open source?

I certainly hope not. Because taking that view simply buys into the false dilemma that with software one may do the enlightened thing, or the self-interested thing, but not both. In fact, open source has always represented, and still represents, the best path to enlightened self-interest in the software world.

Mark Stone is Media Publisher for VA Linux Systems, where he oversees web sites like Linux.com and Themes.org. He has been a writer and editor in Silicon Valley for many years, including two years with O'Reilly & Associates.

   Page 1 of 1