Joff Wild Responds to "End Software Patents"
Understandably, great exception was taken to the bombastic claim that the software industry spends $11.4 billion dollars on patent litigation. From this morning's post:
Clearly, what the [End Software Patents] authors have done is take 55 [the number of cases alleged to be filed each week], multiply it by 52 [weeks] and then multiply that figure (2860) to get the $11.4 billion. In other words, they have assumed that every single case filed in the US ends up going to a final decision at the first instance. However, the truth is that the vast majority of patent suits are settled well before they get to court and so the litigation costs incurred are significantly less. Any patent attorney that End Software Patents had cared to ask could have pointed out this simple truth. But maybe it is just a litte too inconvenient.
It gets worse than that, however. At the end of the third bullet point, the authors of the report state that the source for their statistics is a debate at the Brookings Institute held in 2005. One of the people taking part in the debate was the very same Dan Ravicher of the Public Patent Foundation who is credited with the stat about 55 software patent suits being filed each week. But, in fact, according to the transcript, he said no such thing. Instead, he said (at page 33): "How many patent lawsuits are filed every week? And Brian [Kahn, of the University of Michigan] already hinted at this. That's about 55 [inaudible]. Now, what percentage of those lawsuits are software-related? A small percentage, so they're not all software patents. But we still have an increase in patent litigation. And the cost of defending oneself from a patent lawsuit, as Brian indicated the statistics show, from a year or two ago, I think
that data came out, $2 to $4 million for an average size patent case."[W]hat we have here is spin, pure and simple. End Software Patents understandably wants to make a splash. It is the new kid on the block and feels it has to get its name out there so that it can start to have its voice heard. Unfortunately, however, it has chosen to do this by quoting figures that bear very little relationship to reality. Putting aside the various ethical issues in producing such dodgy data, what End Software Patents has done is wrong because these kinds of figures grab headlines and guide the debate. It is no coincidence that the organisation makes that $11.4 billion point one on its press release announcing the study's publication.
Read the post in its entirety here (link)
(Update)
. . . and read more on the fray in the comments below
Posts
Hi, Peter.
Here's what I posted on IAM. Since many people don't click through, I thought I'd post a comment here as well.
--------------
The main critique I've gotten for this number (from others) is that the statistics are out of date, and therefore too small.
First, this matter of using an average, $4 million. There are indeed many cases that are settled for on the order of tens of thousands, but there are others that drag on to cost on the order of a hundred million, all told. Your statement that many suits filed settle for cheap and the AIPLA's statement that the average cost of litigation over the entire range is $4 million are not inconsistent, especially since AIPLA's figures are (as far as I can tell) about lawsuits, not lawsuits drawn out to final decision.
But if you really don't like my using the AIPLA's number, Bessen and Meurer use a much more inclusive measure of the costs due to softpatent litigation, and find a "mean cost of $28.7 million", in 1992 dollars, equals $43 million in 2008 dollars. They are clear that they are looking at suits filed. The median is $2.9 million (1992 dollars), which verifies your point that most suits are relatively cheap, but a couple are blockbusters. [Bessen and Meurer: The Private Costs of Patent Litigation; download and critique at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=983736]
You're right: I was working off the slides and Dan's failure to veto my numbers, but the `small percentage' is just Dan being polite. In Bessen and Meurer's new book, they estimate that as of 2002, 25% of patent infringement suits are over software; all signs indicate that the current number is much higher.
So perhaps you'd like to use those numbers. Multiplying these numbers together (55 x 52 x .25 x 43m) gives us a value of $30.8 billion per year in costs due to softpatent litigation, probably more. Thank you for your corrections.
Finally, your seeming critique that multiplication is too simplistic touches on an interesting question about numbers in headlines. Let me assure you, being that my textbook on statistical computing is forthcoming from Princeton U Press, that I know another method or two for estimating the costs to patent litigation. But if I'd posted twenty pages of methods behind the headline figure instead of a single line of multiplication, none of us would be better off.
In an ideal world, every article would give a confidence distribution instead of a single number, to indicate the extent to which the authors know that the input data is inaccurately measured, that real-world definitions are never truly precise, and the fundamental fact that we make mistakes. In such a world, more sophisticated techniques could come up with much tighter confidence intervals, so I'd want to do better than the basic order-of-magnitude calculation here.
But in the world in which we live, the headline is a single sentence fragment in all caps. I'd love to have run a headline that more precisely describes the confidence interval: "Software patents cost businesses billions of dollars, probably around $10 bilion, but maybe more, maybe less, depending on your sources, definitions, and methods." Clearly, our publicist won't let it run like that. A smart reader reads it like that anyway.
By offering the derivation, sources, and discussion for your to poke at, we chose to invite critique such as this. You're right that there are other ways to run the data, and I did work off of Dan's slides instead of the transcript, but there's little benefit in calling a rough estimate "rubbish". Just call it the rough estimate that it is, just like every other headline in the newspaper today. You get bonus points, when calling somebody's numbers rubbish, by presenting your own. You're welcome to adopt the alternative numbers above which address both of your critiques, or you can present your own numbers. But whatever your definitions and methods, I'm confident that your numbers will fall in the same confidence interval as above: billions of dollars, probably around $10 billion, maybe more.
Again, thank you for checking up on the numbers.
Ben Klemens
Executive Director
End Software Patents
And in the same spirit, this is how I replied to Ben ...
Ben,
You are being disingenuous. Your own report did not link to slides, it linked to the transcript of the discussion. Are you really saying you did not read this? If so, what else haven't you read?
On your main point, even if 25% of all suits filed in the US related to software (and you have not shown they do, you have merely claimed it), that does not mean that they all go to court. If software cases are as likely to settle as other types of cases, only 14% of that 25% is likely to end up with a trial. In effect that means 75 to 100 each year - a far less compelling headline than 55 cases every day. Some of these will end up costing tens of millions of dollars, maybe even hundreds of millions, the vast majority will not.
Finally, the very nature of patents means that most people are not "smart readers" when they go through articles about them. Patents are technical, they are complicatd and they are difficult. People will not know what is right and what is wrong - they are very dependent on how the debate is reported. That's why figures and claims have to be accurate and not guessetimates based on prejudice.
I am not an advocate of software patents, I am a journalist who reports on IP. I have no responsibility to provide alternative figures to make my case, my only responsibility is to point out that your figures do not stand up to scrutiny and that, therefore, they are rubbish. That does not mean that software related litigation is not incredibly expensive. It just means that your claims, based on the data and references you have supplied, are demonstrably wrong. As we both agree that you have misquoted Dan Ravicher to get to the figure you did and that, even if you did not, you failed to take into account that most cases are settled and do not proceed to trial, it is surely incumbent on you to withdraw the claims and the press release. I hope that you do this.
Kind regards,
Joff
Joff Wild, IAM Magazine | 29 Feb 2008
OK, we've changed to use Bessen and Meurer's estimates. As per Joff's notes, we've taken care to use only numbers based on filings, not trials, throughout.
Bk
To me, the problem with the statistic, whatever the number, is that it implies that all the money spent on software patent litigation is "wasted", as though the enforcement of a property right is an illegitimate end (at least in the context of software).
And speaking of context, what are the numbers for monies spent on non-software patents (biotech, semiconductors)? How do they compare? Should these numbers be also treated as "wasted" assets that otherwise could be applied to innovative R&D?
Is the money spent on enforcing copyrights for SW (and criminal penalties) also "wasted" ????
How about money on anti-trust litigation for behemouths like M$ and Intel - is that "wasted" ???
To put it simply:
Ben Klemens and Dan Ravicher (and their yet to be unmasked corporate donors) produced some gross lies to fool the general public!
Liars, liars, shame on you !!!!!
It is a waste to spend money and time and other resources on the false existence of that which is based upon a lie.
The lie being that software is of patent-able qualities. Software is not of such qualities.
To use an analogy, the roman numeral system, in comparison to the hindu-arabic decimal system, is far to limiting to have been used to calculate the information we needed for us to have created much of what we have today, including computers. Yet it took 300 years for the more powerful and easier to use decimal system to overcome the wide use of the roman numeral system in mathematics and the elite accountants industry.
The key difference was the zero place holder. And of course it was easy to argue that only a fool would think nothing can have value.
Numbers and mathematics are abstractions and programming is also done using nothing but abstraction. Only here the abstractions used extend beyond math, though many will point out that it boils down to a mathematical algorithm.
What would happen should someone come along with the programming zero, that would make programming easy and powerful enough that most anyone can and would do it as they found need to. No different than using a calculator for any calculation one might need at the moment, or spreadsheet calculation, etc..
Non-novel would become more common, there would be a lot more "skilled in the art" or programming, etc.
But ultimately, software is not of patentable qualities as it is entirely made up of abstract ideas, using the natural law and right of human conscious ability to create not just abstraction but higher and higher levels of abstraction. All this has a physical phenomenon effect upon society (if in doubt, google the "trillion dollar bet" and read the transcript and realize the far reaching effect of this abstraction manipulation of value representation).
What is universally accepted as not being patentable:
three primary:
abstract ideas
Physical phenomenon
natural law
secondary:
mathematical algorithms
What happens when you try and contradict physics an nature? You have problems and even death!
It is of no surprise there are problems being caused with software patenting. It's predictable in a very calculating manner.
The mathematical zero, how simple it is, yet how empowering it is also to a small finite set of abstractions and their use.
So about the programmers "zero" well its not just for programmers, but rather an identification of a human quality in ability to create and use abstractions. google "Abstraction Physics".
Post a Comment