How Many Patents Must Be Read to Clear All Patents Rights on Software?
There is a popular theory about the ills of the patent system that there is nothing wrong with software patents. According to this theory, all the problems are due either to bad actors called trolls or to poor quality patents. Then, as the logic goes, the patent system may be fixed by curtailing the activities of bad actors and by improving the quality of the patents.
Is that so?
Let's imagine all these reforms are a huge success. The patent office diligently makes thorough and accurate prior art search on every patent. No questionable or overly broad patent is ever issued. There are no abusive patent infringement lawsuits because no one ever sues unless he owns a valid patent and the defendant actually infringes. Unfortunately, this dream world is not yet ideal. There is one remaining problem. The cost of verifying that a piece of software doesn't infringe any patent is prohibitive. This alone ensures the patent system cannot function as intended.
People and operating companies who make, use, or sell goods and services can never be assured to own all rights to the goods and services they bring to the market because they cannot afford the costs of verifying that no patents are infringed. Those who make a business of suing patent infringers for profit have no problems finding targets. Operating companies can't help but infringe and be liable for it. The only way out is to stop writing software. This is like making house builders unable to pay the bill for checking they have all rights to the land before they build houses. They must either give up housebuilding or go ahead without clearing the rights hoping they are not sued. This nonsensical legal framework is harming innovation and the broader economy.
It is costly to verify that no patents are infringed because too many patents must be read to find out which ones are applicable. Let's assume a patent attorney works 2000 hours in a year.1 Let's assume this attorney needs 10 minutes on average to read a patent. If this attorney does nothing but read patents he can read 12,000 patents per year, or 1,000 per month. This is a significant speed limit because of the large quantity of applicable patents.
Here are some numbers given to us by Mark Lemley. Just reading the patents is several months if not years of work:2 (footnotes omitted)
A few examples: 3G wireless technology was subject to more than 7000 claimed "essential" patents as of 2004; the number is doubtless much higher now. WiFi is subject to hundreds and probably thousands of claimed essential patents. And the problem is even worse than these numbers suggest, since both 3G wireless technology and WiFi are not themselves products but merely components that must be integrated into a final product. Some industry experts have estimated that 250,000 patents go into a modern smartphone. Even nominally open-source technologies may turn out to be subject to hundreds or thousands of patents.
These numbers severely underestimate how many patents must be read because they count only patents that are applicable to the software. The costs are driven not only by the applicable patents but also by patents which don't apply but still need to be read to verify they are indeed inapplicable. There are presumably many more patents in the latter category than in the former.
It would be helpful if some competent scholar were to estimate how many patents must be read to clear all rights to some piece of software. The best I can produce is hypothetical scenarios. Let's consider a very small patent clearance project where a maximum of 6,000 patents are read. This inexpensive scenario requires six months of work and costs $91,000 assuming the patent attorney is salaried staff.3 A larger project where a maximum of 240,000 patents are read requires 20 attorneys working a full year and the costs rise to $3.64 million. I believe these two scenarios give us a range where the majority of patent clearance projects would probably fall. But a well-researched study is needed if we want more reliable numbers.4
This is just the cost of reading at a speed of 10 minutes per patent, without taking breaks, writing reports or doing anything else. In contrast, clearing all copyrights to software usually requires only a few hours or even minutes of work. In most circumstances it suffices to verify there is an adequate license for all code that isn't developed in-house. Copyright clearance is affordable while patent clearance is not.
This estimation is for a single clearance event. But patent clearance is not a one time event. Most companies develop several programs. Also, many programs are constantly modified and improved. Every modification may potentially infringe on some patents. Each new version of each program requires its patent clearance and these intolerably high costs accumulate.
There are nearly 15,000 software development companies with annual revenue of 500,000$ or less.5 For these companies five inexpensive patent clearance events would wipe out one year of revenue. One large project costs over seven years of revenue. These software developers are unable to own all patent rights to their own software.
Why do so many patents need to be read? Let's illustrate the point with an analogy. Imagine an hypothetical world where written English is patentable. Everyone writing text must verify he doesn't infringes on anyone's rights by searching some centralized database. In this hypothetical world there is no independent creation defense. Also, you don't need to copy verbatim to infringe. Different but equivalent text infringes. If you want to write some text you must verify there is no equivalent text in the database. This applies to single sentences or combination of sentences. This applies to every paragraph and combination of paragraphs. The same applies to sections, chapters and the work as a whole. How many searches are needed to clear any text of substantial size?
Fortunately copyright law doesn't impose this burden on us. But patent law does impose such a burden on software authors. Source code is text. Lines of code may be combined to make more elaborated code, a bit like sentences can be combined to make larger units of text. Patents are applicable to simple operations that may be written in a handful of lines, like the
infamous XOR patent. Or they may be applicable to entire systems. Or they could be applied to anything in between. A developer must identify all portions of his software that could be interpreted as a patentable method and clear the corresponding rights. There is a large number of such portions in a piece of software. Hence, a large number of patents needs to be read.
Also, the burden spreads to the whole society because software development is done by a wide range of individuals and organizations.
Everyone interested in learning how to program can do it. Computers are inexpensive and quality software development tools may be downloaded at no costs from the Internet.
Free and open source software (FOSS) is often developed by individual volunteers.
The size of software companies range from a single individual working as a contractor to large multinational corporations.
Software development is not restricted to software companies because all employers of a sufficiently large magnitude have an internal IT department that develops software in-house.
Consulting firms or individual contractors may be hired to develop software.
Several pieces of software are customizable, including those for making web sites. It is possible to modify or extend their functions by writing code to that effect.
All these factors combine to make software development a very common activity in all areas of the economy. The costs to society of checking all patents rights for all software development is prohibitive.
If we assume all US corporations require a total of 300,000 patent clearance events per year and each event conforms to the above inexpensive scenario where 6,000 patents need to be read, 150,000 patent attorneys would be required and the total cost is $27.3 billion.6
Please remember, this is just the costs of reading patents to find out whether they are applicable.
For comparison there are roughly 41,000 patent attorneys and patent agents in the US7. More than 3.5 times this number would be needed just to read software patents during clearance projects. In 2011, the research and development spending in the United States for information and communications technology was $126.3 billion.8 Our estimated costs for reading patents is roughly 22% this amount. These numbers show the financial burden is intolerable.
As high as it is, this estimation is still unrealistically low for many reasons. There are roughly 100,000 employers in the US with over 100 employees.9 Most of these employers have an internal IT department that constantly writes new programs. The largest employers write dozens of programs every year. Some of these employers constantly write software for the purpose of selling it. These 100,000 employers on average surely need more than 3 patent clearance events per year because they surely write in-house an average of more than 3 pieces of software per year. This alone guarantees there should be more than 300,000 patents clearance events per year. Then there are the patent clearances needed by the 1,150,000 companies with 10 to 99 employees.10 Then there is the need to maintain the existing software. The code must regularly be modified to fix bugs, adapt it to new regulations or adapt it to changes to the business requirements. All the modifications require patent clearance to make sure they don't infringe. Finally, the estimation is based on an inexpensive scenario where only 6,000 patents are read per clearance event.
I computed a theoretical scenario based on unrealistically low assumptions. Of course it would be better if someone produced a well-researched study. I hope someone will eventually to do so.11
How do operating companies currently clear patent rights? According to Mark Lemley they don't.12
[B]oth researchers and companies in component industries simply ignore patents. Virtually everyone does it. They do it at all stages of endeavor. From the perspective of an outsider to the patent system, this is a remarkable fact. And yet it may be what prevents the patent system from crushing innovation in component industries like IT. Ignoring patents, then, may be a "workaround" that allows the innovation system to function in the face of overbroad patent protection.
The efforts of the President Obama to curtail overbroad functional claiming are welcome. The president tackles two real problems: a certain type of wrongly issued patents and the economic damage caused by patent trolls. But the problem of clearing the rights still remains because for purposes of calculating the costs it doesn't matter whether the patent is broad or narrow. Either way the patent must be read to make sure it is not infringed.
The source of the clearance problem may be tracked to this phenomenon mentioned in a report to the White House:13
[A] single piece of software or website might have several thousand "functions" that could be claimed in as many patents.
Thousands of patents must be read because there are thousands of patentable functions in a single piece of software. Even if the patents are narrowed down the specific implementations of the functions, all patents that may potentially be applicable to one of these functions must be read. It is not possible to read only the patents that are actually infringed because there is no way to tell which patents are infringed without first reading them.
This hinders innovation because the innovative parts of a computer program are located in only a few of its functions. The bulk of the functions are old. An innovator may obtain some patents on the innovative parts of his software. This does not suffice. The patent rights to thousands of old functions must also be cleared, because otherwise the innovator doesn't have all rights to his own software.
Here is another point of comparison. Defendants and licensees paid patent assertion entities (i.e. patent trolls) $29 billion in 2011.14 Recall that my low-ball estimation of the costs of reading patents during clearance is $27.3 billion. These costs are comparable to the direct costs of patent trolling. The only reason the economy doesn't bear these costs is that, as Lemley notes, everyone ignores patents. This unhealthy solution makes every software author vulnerable to patent trolls. And it would do no good to force everyone to clear their rights because the economic impact is comparable to tolerating patent trolls.
How can we solve this problem? There is only one way. We must eliminate or at a minimum drastically reduce the number of software patents. Yet, this is exactly what Circuit Judge Moore's dissenting-in-part opinion in CLS Bank v. Alice Corporation, in which Chief Judge RADER and Circuit Judges LINN and O’MALLEY joined, is afraid of:
If all of the claims of these four patents are ineligible, so too are the 320,799 patents which were granted from 1998-2011 in the technology area "Electrical Computers, Digital Processing Systems, Information Security, Error/Fault Handling." See U.S. Patent & Trademark Office, Selected Technology Report, available at
ac/ido/oeip/taf/ec_dps_is_efh.htm. Every patent in this technology category covers inventions directed to computer software or to hardware that implements software. In 2011 alone, 42,235 patents were granted in this area. Id. This would render ineligible nearly 20% of all the patents that actually issued in 2011. If the reasoning of Judge Lourie's opinion were adopted, it would decimate the electronics and software industries. There are, of course, software, financial system, business method and telecom patents in other technology classes which would also be at risk. So this is quite frankly a low estimate. There has never been a case which could do more damage to the patent system than this one.
These Federal Circuit Judges clearly think that invalidating so many patents would be a bad thing, imagining negative consequences on innovation and the economy. This view is backwards. It is hard to see how any possible benefits of software patents could offset the harm of a system where operating companies are unable to clear patent rights to their own products.
That's looking at some theoretical numbers in a current scenario. Now think about the future numbers, as Carl Hewitt does in his talk, The Future of IP Software (here on YouTube), when computers will be able to write and file patent applications by the thousands on all known IP, and you can see that his prediction that without pruning back the system it will fall under its own weight is quite realizable, in fact inevitable.
1 This is 40 hours per week, 50 weeks per year.
2 See Lemley, Mark A., Software Patents and the Return of Functional Claiming (July 25, 2012). Stanford Public Law Working Paper No. 2117302. Available at SSRN: http://ssrn.com/abstract=2117302 or http://dx.doi.org/10.2139/ssrn.2117302 pages 24-25.
3 According to Patently'O the median salary of a patent attorney is roughly $130,000. Adding 40% for benefits and overhead gives $182,000 annually. A six-month project is half this amount.
4 To the best of my knowledge, the study that comes closest is Mulligan, Christina and Lee, Timothy B., Scaling the Patent System (March 6, 2012). NYU Annual Survey of American Law, Forthcoming. Available at SSRN: http://ssrn.com/abstract_id=2016968. This study supports the view that patent rights cannot be cleared because the number of patents that must be read is too high. But it doesn't produce the kind of numbers I contemplate in this article.
Great but Not Big.
6 This is the result of simple arithmetic. A single event takes six months and costs $91,000. Multiplying by 300,000 event in a single year gives the stated numbers.
7 Source: the USPTO.
11 To the best of my knowledge, the study that comes closest is Mulligan, Christina and Lee, Timothy B., supra.
12 See Lemley, Mark A., Ignoring Patents (July 3, 2007). Stanford Public Law Working Paper No. 999961; Michigan State Law Review, Vol. 2008, No. 19, 2008. Available at SSRN: http://ssrn.com/abstract=999961 or http://dx.doi.org/10.2139/ssrn.999961.
Patent Assertion and U.S. Innovation, (PDF),
Executive Office of the President, June 2013, page 8.
14 See Bessen, James E., and Michael J. Meurer. The Direct Costs from NPE Disputes. Boston University
School of Law, Law and Economics Research Paper No. 12-34, June 28, 2012.