Prior art is defined as the body of prior knowledge relating to the claimed invention, including prior use, publications and patent disclosures [24]. During the patent life cycle (Section 3), prior art plays a role at different moments:
In the patent application it is usually indicated which previous patents are used or extended. As already discussed in Section 5.2, only the European Patent Office does not require to mention prior art that is known to the applicant.
We conjecture that in all the three cases mentioned above, the determination of prior art is identical, whether this is true prior art or posterior art as defined above.
A patent may describe a technique that computer scientists consider to be trivial. Nonetheless, it may turn out to be very hard to find prior art for it. Well-known techniques cannot be published in a scientific publication for the simple reason that they are already well-known and do not constitute a new research result. These well-known techniques may be used in the source code of many software systems, but this does not count as ``publication'' and cannot be used to illustrate prior art. At the same time, it may also be the case that they are not covered by any patent and someone can just file a patent application for this well-known technique.
Ullman [33] is among the most cited computer science researchers world-wide and he describes eloquently the difficulty to find prior art for a patent application about matrix triangularization that was later used in an unsuccessful attempt to bring suit to the large spreadsheet manufacturers.
In disciplines like chemistry and biology the patent literature forms the actual documentation of inventions. For software the unique situation exists that there is another powerful information source that plays no role in the patent process: the source code itself. This is a major handicap when searching for prior art. There is evidence that cross-citation between the patenting literature and the computer science literature is nearly absent [1]. Compared to software patents, business patents seem to contain relatively more references to the non-patent literature [2]. Nonetheless, the world of software and the world of patents seem mostly disjoint. From this follows that computer scientists are currently not well-aware of the patent literature.
We may conclude that it is urgent to find new ways to establish prior art. One way is the creation of public web sites that solicit and award proofs of prior art. It seems reasonable to include procedures in the patenting system where the public can submit prior art against patent applications.
Another way for establishing prior art is the patent system itself. Suppose the IsNot application is rejected. This fact can have a very positive impact: all the claims in the application are considered to be un-patentable and this blocks future patents on the issues stated in the rejected claims. In this way, a rejected patent application contributes to building up prior art. It is conceivable that major companies follow this strategy in order to prevent patent applications by competitors or to provide indemnity to clients against intellectual property claims.