I’ve been following a very interesting back and forth kicked off by legal scholar and author Tim Wu in The New Yorker on the merits of open vs. closed, and countered by author John Gruber. To catch you up…
Accuse me of overreading, but I propose a revision of the old adage: closed can beat open, but you have to be genius. Under normal conditions, in an unpredictable industry, and given regular levels of human error, open still beats closed. Stated a different way, a firm gets to be closed in exact proportion to its vision and design talent.
Allow me to start by putting forth an alternative rule of thumb for commercial success in any market: better and earlier tend to beat worse and later. That is to say, successful products and services tend to be those that are superior qualitatively and which hit the market sooner.
I liked Gruber’s response quite a bit, if only because my bias is strongly with Wu, and Gruber’s arguments were well formulated. But Wu responded today, and I think ultimately his response captures why the value of open systems can’t be underestimated:
The study of centralized and decentralized decision structures in an economic system is hardly my invention. It goes back to classic economic debates between Oskar R. Lange and Fredrick Hayek in the 1940s. Lange was an advocate of centralized planning, and argued that closed, state-run economies would be more efficient than open / decentralized market economies. Hayek, responding in 1945, argued that the advantage of an open system was largely informational. A theoretically perfect central planner would, Hayek conceded, outperform an open system, but in a reality of imperfect information, the open market system could usually be expected to perform better. There’s been much economic thought on the issue since that time, but I’ll skip it: the bottom line is simply that open and closed systems perform differently under different conditions and have differently strengths and weaknesses. I should add that this kind of analysis is relevant for any system and any product ecosystem, not just tech — it is really the study of institutional design.
(It’s worth noting that one of the points of Wu’s book is that often when closed systems do win, it’s at the expense of innovation.)
Zooming out like this is helpful to illustrate the relevant intellectual history; on the other hand, the case that open beats closed is strongest where the scope of he task is largest, because of the related uncertainty. Designing an economy is beyond the ability of any planner; designing a toothbrush is not.
In what follows, I’m referring to this particular formulation of “open” which Wu offers in his original piece:
First, “open” and “closed” can refer to how permissive a tech firm is, with respect to who can partner with or interconnect with its products to reach consumers. We say an operating system like Linux is “open” because anyone can design a device that runs Linux.
These questions get very complicated quickly, and “openness” not only becomes a matter of extent, but its merits will also differ layer to layer. How much control the maker of a piece of hardware exerts over what software can be run on it is separate from how much control an OS exerts over what applications can run on it is separate from how much control those applications have over what plugins can be added and which 3rd party apps can be integrated.
My intuition is that as the scope of the activity governed by the piece of the tech stack in question gets smaller, the case for (relatively more) closed technologies becomes stronger. In other words, the interoperabiltiy of an OS matters more than the interoperability of Snapchat. The complexity and uncertainty surrounding the latter should be less than in the case of the former, meaning the latter is more amenable to the closed planning model.
All of this is to say, I agree basically with Wu’s conclusion:
as Michael Arrington points out, you really can’t pretend to understand what has happened over the last twenty years without some understanding the relative advantages of open and closed systems (or if you prefer, decentralized and centralized decision hierarchies.)