Sunday, October 30, 2005

Catching a tiger by the tail: Patents could destroy the software business

The software business is working itself into a situation where its economics are built on software patents. This will be catastrophic for investors unless such patents are constrained to be of high quality, limited scope and – especially – short duration. If this doesn't happen, licensing costs will dominate the business, and pure software companies will be at the mercy of patent trolls.

The drive to software patents is based on the argument that a knowledge-based industry needs a way to trade knowledge, and thus needs intellectual property. David Kaefer of Microsoft says it well: “Software is built on the shoulders of giants -- no one can build the whole thing. Patents are a property right that allows the innovation to be exchanged”. [1]

This logic is built on the fallacy that the commodities to be traded behave just like the physical goods we’re used to buying and selling. Our models of trade are based on exchanging goods and human services that are exhaustible and rivalrous, because they are essentially tangible. If you give me $20 for the service of cleaning your gutters, you no longer own the money; and I can only clean one client’s gutters at a time. Innovations are knowledge goods. Your use of my idea doesn’t limit another person’s ability to use it, and that idea is infinitely duplicatable. George Bernard Shaw summed it up well:

If you have an apple and I have an apple and we exchange these apples then you
and I will still each have one apple. But if you have an idea and I have an idea
and we exchange these ideas, then each of us will have two ideas.

One may object that the patent system has worked well for centuries; what’s suddenly the problem? Patents have always been on innovations, and thus we have a system that works for non-rivalrous intangibles. What’s changed is that software goods are not wrapped in matter, and it’s matter which is at the root of the metaphor of trade and exchange on which our economy is based. Software is pure ideas, and can be exchanged without being wrapped in stuff. Further, there are far fewer limits on the number of ideas that can be added together than on the number of components that can be added together to make a physical product [2]. I would predict that the number of patents per product increases as the knowledge content of the product increases; there are many more patents entailed in a $100 copy of Windows than in a $100 food processor.

Software is cumulative, as the Y2K experience has shown. Most software products beyond their first release are not only built on relatively ancient algorithms, but in fact are accreted the ancient code that implements those algorithms. Advancing the product is a additive process; little is taken away once it is written. Contrast that with physical products: a new and improved kind of soap can be patented, but one doesn’t have to license all the preceding patents in order to produce the new one. [3]

One of the assumptions implicit in the industry’s world view, and explicit in David Keafer’s contention, is that companies will cross-license with each other to obtain the knowledge inputs that they require without paying for it. This has clearly worked to date; in most industries the participants create license pools or build a web of cross-licenses that enable them all to operate. In the brave new world of knowledge goods, though, everybody is a software company. It will be impossible to draw boundaries around the industry players who’ll need to cross-license with each other, since everybody will be producing software innovations, though most will not sell software. Companies that focus on selling software will be worse off than anyone else – they will have to cross-license with everybody else.

Much has been written about “patent trolls”, those companies who live to invent, and who license their patents to productizers. There are fears that the trolls will extract fearsome tolls from those who want to build products. Their disproportionate power is a consequence of the disproportionate weight that has been given to the inventive step in innovation.

Many in the software industry argue that patents are necessary to guarantee investment in innovation. This only follows if patents incent the most critical step in the process. Let’s unpack the term “innovation”. It consists of the sequential steps of invention, synthesis, execution and distribution, that is, (1) coming up with an idea; (2) composing a number of ideas into a novel product concept; (3) building the product; and (4) putting the product successfully into the market. Patents reward Step 1; however, most of the value is generated further down the chain, in synthesis, execution and distribution. Some might describe patents as the keystone that hold together the arch of innovation. I think that exaggerates the importance of invention. Patents privilege a necessary but by no means sufficient step in the process of bringing new products to market.

If hundreds or thousands of ideas have to be licensed in order to ship a new software release, pure software companies will find that their input costs balloon, especially if the licensors are in industries unrelated or hostile to them, and if inventors are able to charge prices out of proportion to the importance of inventions in the innovation process.

Still: to the extent that patents can be used to anchor innovation transactions, they are useful. However, they are scaffolding rather than a keystone. They can be removed once the structure is complete. Hence, short patent terms are essential. The term of a patent should bear some relation to the rate of change in an industry; the software industry advances rapidly. Short terms are particularly important in software, where the development of a product is cumulative. By all means allow innovators to protect the new icing which makes their cake special; but allow the lower layers to pass into the public domain so that others can also innovate in making icing.


[1] The arms race, Economist, Survey of patents and technology, 22 Oct 2005

[2] There are surely limits to size of a given piece of software, as innumerable software projects who’ve missed their deadlines have demonstrated; Microsoft’s difficulties in shipping the Longhorn release of Windows is one in a long line of examples.

[3] For example, Lever 2000 bar soap is covered by patents 4,695,395 and 4,663,070.

No comments: