One would imagine that economists would have learned by now that they need to check their theoretical models against the real world. A recent article arguing that governments need to promote a “balance” between open source and proprietary software, rather than promoting open source, rests on no fewer than five incorrect assumptions, and is easily rebutted by what happens in the real world.
This article claims that economic models show that a market dominated by open source should produce less software than a market in which proprietary and OSS compete, with the optimum being less than 20% OSS. The problem lies in the assumptions they make to produce this model:
To see why, consider an all-OSS world in which each company offers consumers exactly the same shared code as every other company. By definition no company can then compete by writing more OSS code than its rivals.
How many incorrect assumptions are there?
- In an all OSS world each company offers consumers exactly the same shared code. Wrong. On real life multiple OSS products with entirely different code will compete with each other
- New code is immediately available in competitor’s products. Wrong. It takes time for competitors to rebuild and test. The original developer gets a first mover advantage.
- The only reason to write new code is to incorporate it into software to sell. Wrong. There are many other benefits to writing code: it helps sell services and hardware
- Only software vendors write code. Wrong. Users write and improve software for their own purposes and share it. There are a number of reasons why they cannot, or would not benefit from selling it.
- The only benefits to users are from the output of new software. There are also costs, and, especially, the hidden costs of the vendor-lock in of proprietary software. OSS with multiple vendors of the same code has option value from being able to switch vendors. All OSS has option value from lower migration costs and ability to fork.
Now lets look at the real world. Consider the web server market. According to the latest Netcraft survey not only is the market dominated by OSS, but by a single OSS product, Apache with a 66% share — a dominant position it has held for years. If one eliminates the Google Web Server (only used internally at Google) and separates the Unix and Windows web server markets, Apache’s dominance strengthens within the Unix market, and all its significant competitors are OSS.
This, if anywhere, is where you would expect to see the lack of progress the models predict. Of course, this is what you do not see. The last few release of Apache, Nginx or Lighttpd show continuing progress at a rapid rate for server software (where stability and security must always come first).
Mail servers and email delivery prove the same. One of the most influential essays written on open source, The Cathedral and the Bazaar, documents the delivery of the very widely used Fetchmail, and in doing so explains the reasons why people develop software to use rather than to sell. There goes another assumption.
Almost any kind of Linux desktop software also disproves the assumptions. Even if we exclude desktop environments and other critical software as influencing choice of operating system and therefore indirectly competing with their proprietary counterparts on Windows and MacOS, we are still left with lots of pure OSS dominated categories of software with variety and rapid, innovative development: text editors, bit torrent software, music players, etc. In fact, as far as torrent applications go, I believe that even Windows and MacOS usage is dominated by open source. Where is that stagnation?
Consider the benefits that companies get from writing code other than selling it. Red Hat makes sales because its contributions to Linux give it credibility as a company with the technical ability to fix client’s problems fast. IBM contributes partly to get Linux running well on its hardware: but most of those improvements benefit everyone.
Linux desktops would hardly evolve as fast if their evolution was driven by competing with Windows or MacOS. They are already well ahead of either in terms of features, but continue adding features to compete with each other. This applies to both desktop environments (KDE, Gnome, XFCE, Enlightenment. etc.) and to desktop distributions.
All this not only proves that OSS dominance mean progress not stagnation, but that the likely outcome of OSS dominance is multiple competing products. If a single product became dominant and did not progress it will be forked and the fork would progress: this is what happened to XFree86, which was the windowing system formerly universally used by LInux and similar operating systems.
In fact, it is proprietary software with its ability to lock-in users with high migrations costs that is likely to lead the market to stagnations. Consider what happened in the time when Internet Explorer dominated the market for web browsers — many years of stagnation. It is obvious that the same would not happen if Firefox or Chrome was dominant, as a new entrant could freely reverse engineer to ensure compatibility, or simply fork and compete.
The real lesson of the models is that a more realistic version of it would confirm the lesson we already know from real life: that dominance by a proprietary vendor leads to slow progress. This is a very strong argument for breaking up companies, like Microsoft, that dominate a software market.