Seeking prior art where it most often is found in software

Open Invention Network plans to mine open source projects for patent busters

Patent ambushes are on the rise again, and cases such as Apple/Samsung shows that prior art really has to swing the decision–obviousness or novelty is not a strong enough defense. Obviousness and novelty are subjective decisions made by a patent examiner, judge, or jury.

In this context, a recent conversation I had with Keith Bergelt, Chief Executive Officer of the Open Invention Network takes on significance. OIN was formed many years ago to protect the vendors, developers, and users of Linux and related open source software against patent infringement. They do this the way companies prepare a defense: accumulating a portfolio of patents of their own.

According to Bergelt, OIN has spent millions of dollars to purchase patents that uniquely enable Linux and open source and have helped free software vendors and developers understand and prepare to defend against lawsuits. All OIN patents are available under a free license to those who agree to forbear suit on Linux grounds and to cross license their own patents that read on OIN’s Linux System Definition. OIN has nearly 500 licensees and is adding a new one every three days, as everyone from individual developers to large multinationals are coming to recognize its role and the value of an OIN license.

The immediate trigger for our call was an announcement by OIN that they are expanding their Linux System Definition to include key mobile Linux software packages such as Dalvik, which expands the scope of the cross licenses under the OIN license. In this way OIN is increasing the freedom of action under which a company can operate under Linux.

OIN’s expansion of its Linux System Definition affects not only Android, which seems to be in Apple’s sights, but any other mobile distribution based on Linux, such as MeeGo and Tizen. They have been interested in this area for some time, but realize that mobile is skyrocketing in importance.

Meanwhile, they are talking to their supporters about new ways of deep mining for prior art in source code. Patent examiners, as well as developers filing patents in good faith, look mostly at existing patents to find prior art. But in software, most innovation is not patented. It might not even appear in the hundreds of journals and conference proceedings that come out in the computer science field each year. It is abstraction that emerges from code, when analyzed.

A GitHub staffer told me it currently hosts approximately 25 TB of data and adds over 65 GB of new data per day. A lot of that stuff is probably hum-drum, but I bet a fraction of it contains techniques that someone else will try to gain a monopoly over someday through patents.

Naturally, inferring innovative processes from source code is a daunting exercise in machine learning. It’s probably harder than most natural language processing, which tries to infer limited meanings or relationships from words. But OIN feels we have to try. Otherwise more and more patents may impinge (which is different from infringe) on free software.

tags: , , , , , , ,

Get the O’Reilly Programming Newsletter

Weekly insight from industry insiders. Plus exclusive content and offers.

  • ComplyAnt

    This is where starting from a process model and linking it to the code can help. For example: Eclipse Use Case or BPMN diagrams. These can, and should, model WHAT something does at the high level and HOW it does it at the low level. It also helps re-use as you can search on the process or result (or the trigger) as well as helping users to understand how it might link to, or be used for, their business processes.