Back in 1994 I was a member of the founding team of Open Market, an early eCommerce software company.
We started work in our basements in April 1994 to solve the challenges of online shopping, and on October 24, 1994 launched our product the “Open Marketplace”. That same day we filed a patent application, which eventually issued in 1998 as U.S. Patent 5,715,314, titled “Network Sales System”. The Wall Street Journal published two insightful articles about Open Market’s patents at the time: Open Market Receives Patents for Internet-Commerce Software and Patents Lift Open Market As Observers Guess Their Worth.
This posting isn’t about eCommerce particularly. Win Treese and I wrote a book on that subject, called Designing Systems for Internet Commerce. In two editions, it sold over 20,000 copies, was used for a number of college and business school courses, and was even translated into Chinese. This post is about the ‘314 patent, which is sometimes referred to as the “Shopping Cart Patent”.
Open Market grew rapidly, went public, made acquisitions, and outlived about 15 of its competitors, but was hit hard by the bursting of Internet Bubble 1.0 in 2000. The eCommerce software part of Open Market’s business, called “Transact”, and the related patents were acquired in 2003 by Soverain Software, which still develops and supports Transact and its customers today.
Soverain also filed a number of patent lawsuits, and settled or won all of them, up until this year. On September 4, the United States Court of Appeals for the Federal Circuit (CAFC), which is the appeals court responsible for patent cases, confirmed its January opinion that certain claims of the ‘314 patent (and some others) were invalid as obvious. Soverain has filed a petition for the case to be heard by the Supreme Court, and I hope it is accepted, because I think the CAFC got it really wrong.
Before you read any further, I want to mention that I have no ownership stake in any of Soverain’s patents – like all of Open Market’s employees, I assigned my rights to the company. I am, however, a paid consultant for Soverain, mostly to explain fine points of technology, and I have to admit to being a somewhat fascinated onlooker to the legal pageantry. I also have an inventor’s pride in my work, and an engineer’s natural alignment with facts, data, and truth.
In 2007, Soverain sued computer retailer Newegg for infringing various claims in its patent portfolio, including the so-called shopping cart claims in the ‘314. Soverain won the case in the trial court, but, crucially, the judge did not let Newegg’s assertion that the claims were obvious go to the jury because he thought the testimony and evidence on the issue was wholly inadequate. Newegg appealed the case, asking the Court of Appeals to send the case back for re-trial so that a jury could hear the evidence on obviousness.
A three-judge panel of the CAFC heard the case in September 2011, and then sat on it for over a year. In the meantime, a jury in a different trial decided that these patents were not obvious and the Patent Office re-examined the patents and also found that they were not obvious. In January 2013, the CAFC, surprisingly, went beyond Newegg’s request for a re-trial and instead decided on its own that the claims were invalid, saying they were obvious. Then in September, the same three-judge panel confirmed their January decision and the entire CAFC rejected Soverain’s request that the case be reheard by the full court.
To some, the claims of the ‘314 and Open Market’s other patents may seem obvious today, but the situation in 1994 was very new and different. An informed discussion about the prior art and how it might apply to these patents is proper, but the key is “informed.”
The way courts are supposed to work is that judges rule on the law and juries decide facts. Juries weigh the evidence and determine the credibility of witnesses.
According to the Supreme Court, obviousness is a matter of law “based on a set of underlying facts”. What bugs me is that the CAFC should not get to decide the underlying facts on its own, which is what it did in this case. The CAFC focused, more or less, on just two sentences in the trial record, in which one expert said “Compuserve already did this” and another expert said “no, they didn’t”. I shouldn’t have put quotes there, because I am paraphrasing, but you get the idea.
I could accept the CAFC’s decision if they understood the technical issues and still felt that the legal issues decided the case, but I can’t accept it when it is pretty clear that the judges misunderstood the technical issues and there was no opportunity to argue their misconceptions.
As just one example, claim 34 of the ‘314 patent requires that a message entering an item into a shopping cart contain a “product identifier” for the item. Newegg’s argument is that this is merely a truism. They say: the message has to identify the product or the system couldn’t work! This is disingenuous. And wrong. I will explain why.
The Compuserve environment, the supposed prior art, worked using dumb terminals connected to a time sharing system. A user would see a screen of items for sale, and could add the third item to the shopping order by typing “O 3”. Because there was a permanent connection between the terminal and the computer, the computer would remember what the third item on the screen was and add the right thing to the order. If, a few minutes later, the user was looking at a different screen of items and typed “O 3” again, a quite different item would be ordered, yet the messages to the computer – “O 3” – would be identical. If you look at the “O 3” messages alone, you would have no idea which product is specified. Since the same “O 3” message can order different things, it’s obvious the Compuserve system does not contain a product identifier.
The Open Marketplace system worked quite differently. It was intended to work on the then fairly new web, with stateless HTTP communications between client and server. A user looking at an electronic storefront web page would click on an item’s “add to cart” button, and the web browser would send an HTTP GET request with the attached URL to the specified server. The URL was quite a complex object. It specified the name of the server and the “add to cart” application, and it also carried a payload with a message authentication code signed by a key known only to the merchant and the payment computer. The Open Marketplace payload contained the identity of the merchant, the identifier of the particular product, the price of that product, and the period of validity of the price.
In the Open Marketplace web-based system, the add-to-cart message is self-contained, and if it were not, the system would not work at all. Unlike in the Compuserve system, the shopping cart computer of the ‘314 claims has no idea what is on the screen of each user! There is no way for “O 3” to mean anything, never mind to designate a specific product. The catalog page – which we now call a web page – might have come from the same server, but because HTTP is stateless, there was (in 1994) no way for the server to know that. The catalog page could have come from a different server (and in the Open Marketplace system it did) or it could have come from a web page on a CD ROM mailed to the consumer. For the ‘314’s system to work, the message had to contain a self-sufficient product identifier (which is just one element of claim 34).
At trial, Soverain’s expert clearly said the Compuserve order command did not contain a product identifier. And Compuserve’s own CTO said the same thing. The CAFC, however, said that “product identifier” was not given a special meaning in the specification and overlooked the fact that Newegg and Soverain had agreed to a meaning of “identifier”. So, the CAFC apparently accepted Newegg’s theory that since “O 3” was enough to order the third item, therefore it must be a product identifier. That is just nuts (and incorrect!).
The Supreme Court, I hope, will accept the petition to hear this case. (And if it does, I’ll be there, that would be something.) The Supreme Court doesn’t need even to discuss the technical merits here. All they have to do is remind the CAFC that judges must not pick and choose facts; because deciding facts is the job of the jury.
I wanted to mention one other thing – there’s a lot of furor about “junk patents” which generally cite little prior art and have a vague specification. The ‘314 patent isn’t like that. The section in the U.S. Constitution that allows for patents says they are intended to “promote the progress of science and useful arts”. The patent owner gets a temporary exclusive right and in exchange has to document the invention in sufficient detail so others can benefit.
That’s exactly what we did.
The specification of the ‘314 was written by an MIT faculty member for Open Market. The relevant source code of the Open Marketplace system as of October 1994 was included with the patent application for anyone to read – over 50 printed pages of code. In other words, Open Market showed that these inventions weren’t just a theory but an actual working system. Open Market submitted the source code to the Patent Office on microfiche since there was no way to submit machine readable appendices back in 1994. A scanned copy of the original Appendix G is available here if you are interested. Soverain has also now made a machine-readable version of the source code available here.
Also, after the initial ‘314 patent application filing and two subsequent rounds of patent office reexam, literally hundreds of pieces of prior and related art are now cited in the ‘314 patent. And for several years, the ‘314 patent was on the Patent Office’s list of the top ten patents most frequently cited by other patents. So the ‘314 is definitely not a junk patent.
Oh yes. Remember how Amazon received a patent for 1-click purchasing? We had that working in June 1994. Check the Open Marketplace source code.
About the Author
Dr. Lawrence C. Stewart is currently Chief Technology Officer of Serissa Research and a named inventor on 14 U.S. patents. He received a S.B. in Electrical Engineering from MIT in 1976, followed by M.S. and Ph.D. degrees from Stanford University in 1977 and 1981, respectively, both in Electrical Engineering. His Ph.D. thesis work was on data compression of speech waveforms using trellis coding. He began his career at the Xerox Palo Alto Research Center, and held technical positions at the Digital Systems Research Center in Palo Alto and Digital’s Cambridge Research Lab before joining Open Market in 1994 as its Chief Technology Officer.