Searching Computer Implemented Inventions - The Truth is Out There

Members of the UK judiciary have recently expressed concern over the difficulty of prior art searching for computer program inventions.   Such sentiments were first espoused last year by Mr Peter Prescott QC, sitting as a Deputy Judge of the High Court in CFPH L.L.C.'s Application (see Footnote 1), and have since been echoed by observations from Robin Jacob LJ at a seminar arranged by the Society for Computers and Law (see Footnote 2).   At a time when there is so much controversy surrounding patents and software, such remarks provide ammunition for opponents of the patent system (see Footnote 3).   It is therefore important to investigate and understand the perceived difficulties in more detail.

Of course, computer programs per se are excluded from patentability by Article 52(2)c of the 1973 European Patent Convention (EPC), which is mirrored by Section 1(2) of the 1977 UK Patents Act.   According to the judgement in CFPH L.L.C.'s Application, one reason for this exclusion is that "it was felt that searching the prior art would be a big problem".   In support of this statement, the judgement cites Rule 39(1) of the 1970 Patent Convention Treaty (PCT), which excuses an International Search Authority from having to search an invention related to computer programs "to the extent that the International Searching Authority is not equipped to search prior art concerning such programs".

However, this interpretation must be open to question, since Rule 39(1) PCT is merely pragmatic in nature.   It recognises as a practical matter that some International Search Authorities might not have the resources or expertise to search inventions related to computer programs, but it also clearly contemplates that some International Search Authorities would have the resources and expertise to search such inventions.   For the founders of the EPC to have believed that the European Patent Office (EPO) would not be able to match or surpass the capabilities of other International Search Authorities seems uncharacteristically defeatist (see Footnote 4).

It is notable that in the recent judgement of Shopalotto (see Footnote 5), Pumfrey J is considerably more cautious than Mr Prescott QC as regards the origin of the computer program exclusion of Article 52(2) EPC.   Although Pumfrey J accepts that practical searching problems might have been one potential reason for the exclusion, he writes   that "it is difficult to discern any underlying policy".   It is also suggested in Shopalotto that the exclusion might have arisen instead from more theoretical considerations concerning the intrinsic nature of an invention - an idea which seems to be in better accordance with the historical record (see Footnote 6).

Irrespective of whether or not searching difficulties originally led to the computer program exclusion of Article 52(2) EPC, the perceived existence now of such difficulties is nevertheless significant.   This is especially true given that there are very many granted patents for inventions that are implemented by computer programs (rather than being computer programs per se) (see Footnote 7).   It is important for public confidence in the patent system that such patents are properly searched for potential prior art.  

In the CFPH judgement, Mr Prescott QC appears to feel quite strongly about this: "you cannot have a sensible patent system unless there exists a proper body of prior art that can be searched.   Not only are most computer programs supplied in binary form - unintelligible to humans - but most of the time it is actually illegal to convert them into human-readable form.   A patent system where it is illegal to search most of the prior art is something of an absurdity".   Furthermore, Jacob LJ is reported as having "considerable concern" regarding prior art searching in relation to computer program inventions.   Jacob LJ appears to view such searching as difficult in practice, with lapses in search quality leading to patents having been granted for "very ordinary things" (in the US at least).  

Given this concordance of judicial opinion, the neutral observer might well conclude that computer program inventions are indeed difficult to search.   However, there is no evidence from those at the coal-face of patent searching to support such a conclusion.   For example, the author has never encountered a PCT application where an International Searching Authority has actually chosen to avail itself of Rule 39(1) PCT in order to refuse to search a computer program invention (see Footnote 8).   Likewise for practitioners such as patent attorneys, searching computer program inventions is very much regarded as business as usual.

The discrepant viewpoints of the judiciary and practitioners may stem in part from some misapprehension on the bench about the nature of patents for computer-implemented inventions.   For example, Mr Prescott QC writes that "most computer programs [are] supplied in binary form", and draws from this the conclusion it would be illegal to convert (decompile) the programs into human readable form to search for prior art.   These comments are clearly based on the implicit assumption that the source code of a program is needed to determine whether or not it anticipates a patent.  

However, it must be remembered that the primary purpose of a claim is to determine the scope of infringement.   It is highly undesirable to write a claim for a computer-implemented invention that requires source code analysis to determine whether or not a program infringes, since as Mr Prescott QC observes, such source code may be very difficult to come by.   Accordingly, there is already a strong incentive in the patent system for patents for computer program inventions to define the invention in functional terms, rather than in terms of particular source code structures.   In other words, a program would infringe such a patent if it performs the specified set of functions.   This generally allows infringement to be detected by use of the program, without recourse to source code analysis.   Many (probably most) patents for computer-implemented inventions are claimed in such functional terms.   Accordingly, the availability or otherwise of source code is simply not relevant for determining infringement or prior art issues for such patents (see Footnote 9).

Another potential area of misunderstanding concerns the nature of prior art.   For the overwhelming majority of patent applications, the Patent Office does not cite actual products as prior art, but rather relies upon published documents - patents, technical papers, web articles and so on (see Footnote 10).   Indeed, this type of prior art literature is generally much more readily available for software than for (say) computer hardware.   For example, a visit to a bookshop will quickly confirm that there are far more books available on software than on the internals of a computer display or printer.   Similarly, the number of academic journals focussed on computer software and applications greatly exceeds the number of journals that concentrate on computer hardware.  

In addition, even if a software product is to be invoked as prior art, old software may be much more accessible than old hardware.   For example, the number of 20 year old computers still in operation (or even still operable) is extremely small; most have long since been scrapped.   On the other hand, old code can be easily stored, and indeed, may frequently still be in use (as indicated by the concern over Y2K compatibility).  

The above considerations demonstrate that contrary to the suggestion in the CFPH decision, there is indeed a "proper body of prior art that can be searched" for computer-implemented inventions.   On the other hand, just because prior art is available, it does not necessarily follow that a Patent Office is good at searching it.   Consider for example the practice of the US Patent and Trademark Office (USPTO) in the 1980s and early 1990s.   At this stage, the main prior art resource for US Examiners comprised a library full of boxes, each box containing paper copies of existing US patents belonging to a particular technology classification.   When a US Examiner searched for prior art, he or she pulled the boxes of appropriate technology classifications, and skimmed through the paper copies of the patents for anything of potential relevance.

Unfortunately, this strategy suffered from several problems. Firstly, there was normally a visual element to the search, based on the drawing on the front cover of the patent, which did not work well for software inventions.   Furthermore, the search strategy relied upon the collection of earlier US patents providing a comprehensive representation of the prior art.   However, due to previous legal uncertainties regarding computer program inventions, and also the general nature of the software industry, much earlier programming technology was simply not described in patents.   As a result, the US patent literature only provided an incomplete record of the development of software.

One attraction for Examiners in searching US patents (rather than other documents) was the comprehensive and detailed technology classification - the US patents had been sorted into all the different boxes.   In theory therefore, an Examiner only had to look through certain specific classifications to find the most important prior art for a given technology.   Unfortunately however, this taxonomic approach was not always reliable for an immature and rapidly developing field such as software.   The patent classification system for software was still under development, and sometimes there simply might not be a good box to put a new patent into.   Alternatively, even if a good box was available, a patent might have been mis-classified into some other box through inexperience or ignorance.  

It seems likely that it was problems such as those described above that led to the granting of some dubious US patents, as per the concerns expressed by Jacob LJ.   However, these problems have now been largely overcome.   As software technology has matured, the classification scheme has become more stable and reliable.   In addition, the rising number of patents for computer-implemented inventions ensures that the set of patent prior art provides a much more comprehensive representation of the state of technology.    Perhaps most importantly, the fundamental method of searching has changed, and is now based on on-line keyword searching.   As a result, even if a patent has been mis-classified, it can still be found through the use of appropriate keywords.   Furthermore, since the keyword search strategy is no longer tied to the formalised technology classifications of patents, it is easy to extend the search to non-patent literature - in effect, just plug in additional databases (see Footnote 11).  

The searching of computer-implemented inventions is also still the subject of rapid and significant improvement.   Firstly, the last few years have seen a dramatic rise in Open Source software, including Linux, Apache and other projects.   Thus while the statement of Mr Prescott QC that "most computer programs [are] supplied in binary form" probably remains true, a very significant amount of programming is now also provided in source code format.   This significantly extends the body of available prior art, even in respect of those patent claims that do require source code analysis for a determination of infringement or validity.

Secondly, the WorldWideWeb provides an increasingly valuable resource for prior art relating to computer-implemented inventions.   More and more information sources are accessible via the Internet, such as archive libraries of papers, international standards, newsgroups and so on.   In addition, many documents that previously might never have become (readily) accessible to the public are now available via the Web (student project reports, etc).   For obvious reasons, the software community was one of the earliest and most prolific adopters of Web publication.

Thirdly, the rapid growth of the Web has driven dramatic and unprecedented improvements in fundamental searching technology.   Thus sites such as Google are still based around keyword searching, but perform a sophisticated analysis and ordering of the results to greatly improve usability (see Footnote 12).

The facilities are therefore now in place to allow better searches for software prior art than ever before.   Of course mean, this does not mean that every search performed by a Patent Office in the future for a computer program invention will be perfect.   In particular, the current processing backlogs for computer program inventions at the various Patent Offices put the search and examining staff under heavy workload pressure, and sometimes search quality may suffer as a result.   Furthermore, any long delay between filing and searching a patent application inevitably makes the search harder (irrespective of the field of technology) (see Footnote 13).  

Nevertheless, if the judiciary does perceive a problem in searching computer program inventions, it seems that this is either historical (and perhaps foreign) in origin, or else correctable by providing improved examining resources at the relevant Patent Offices.   Certainly, there is nothing intrinsic or fundamental that prevents a good search from being performed for a computer program invention.   Accordingly, from a policy perspective, the clear conclusion is that in terms of searching at least, the patent system can readily and properly accommodate computer program inventions in the same way as inventions from any other field of technology.

Footnotes:

1. [2005] EWHC 1589 (Pat); see in particular paragraph 35 of the decision.
2. As reported at: http://www.tmcnet.com/usubmit/2006/jan/1285737.htm. An audiocast of the presentation by Robin Jacob LJ is apparantly available via http://www.scl.org/default.asp
3. For example, the CFPH decision is extensively relied upon in: "Quietly Tying Down Gulliver - The Software Patent Fairy Tale" by Cristian Micelli (from the organisation Lawyers against Software Patents), January 2006, available from http://www.groklaw.net/article.php?story=2006115145429444
4. As an indication of the general sentiment that policy should drive practice rather than vice versa, a meeting organised by the UK Patent Office on the "Legal Protection for Software-Related Innovation", 19 October 1994, included a discussion on the potential problems of searching computer program inventions. According to the formal report of this meeting, "the view was expressed on behalf of the [UK] Patent Office that practical difficulties should not be allowed to stand in the way of reform if a policy decision were taken to extend patent protection in the software field" (see paragraph 4.19 of report).
5. [2005] EWHC 2416 (Pat); see in particular paragraph 8 of the decision.
6. See "computer Program Claims" by Simon Davies, [1998] EIPR 429-433
7. For brevity, such inventions will generally be referred to as "computer program inventions".
8. Consequently, any linkage such as suggested by Mr Prescott QC between the computer program exclusion of Article 52 (2)c EPC and Rule 39 (1) PCT would therefore be especially unfortunate, since it would imply that a significant aspect of substantive European patent law had been derived from an inconsequential and now apparently irrelevant procedural provision of the PCT.
9. In fact, claim 1 of CFPH L.L.C.'s Application itself was presented in just such functional terms, whereby access tot he source code of any potential prior art was irrelevant. (This is not explicitly recognised in the judgement, although the application was ultimately rejected as a business method, another excluded area under Article 52(2) EPC, not as a computer program).
10. The author has never encountered a situation where the Patent Office has cited anything other than a published document. As a former examiner, one would expect Mr Prescott QC to be aware of this practice. Of course, from a judicial perspective, the use of products per se as prior art is more common, especially in inter partes proceedings. However, litigation only occurs in respect of a tiny fraction of computer-related patents. Moreover, it clearly would be feasible for a court to compel access ot source code if so required (and section 45(1) of the 1988 Copyright, Designs and Patents Act precludes from copyright infringement anything done for the purpose of judicial proceedings).
11. The EP generally asopted keyword searching before the USPOTO, which may explain why it has developed a more favourable reputation for searching than the USPTO.
12. Google has applied for patents on some of these techniques.
13. Historically the EPO has had the worst backlogs for computer-related inventions, but while the situation at the EPO now appears to be slowly improving, pendency figures for the USPTO are deteriorating.

For further information please contact Simon Davies, D Young & Co Patent Partner, or your usual D Young & Co patent advisor:

View Contact Details

Return to full list of Publications