Skip to content

Pardo and Landau: LANPAR and the Spreadsheet Patent

Abstract

In the summer of 1969 two young Canadians, Rene Pardo and Remy Landau, wrote LANPAR, a budgeting language for time-sharing computers in which formulas could be typed in any order and the program worked out the order in which to calculate them. Bell Canada, AT&T and the American telephone companies used it for their budgets ten years before VisiCalc. Pardo and Landau applied for a US patent in August 1970. The Supreme Court’s first software-patent decision stopped it at the last step, and they got it back only by arguing their own appeal in 1982. A patent-licensing firm then used it to demand five percent of the sales of every major spreadsheet program. In 1995 Judge Sonia Sotomayor ruled it unenforceable: one of the “disinterested” experts whose affidavit had won over the examiner in 1972 had worked for the inventors that winter and been taught the program by Landau himself.

Two Canadians and a Budget Problem

Rene Pardo was born in Switzerland to parents living in Egypt and came to Canada at four. By his own account he began keypunching cards at the McGill University computer centre in the summer of 1964, spent the following summer at an IBM test centre, worked at the National Film Board from 1966, took a BSc at McGill in 1968 and then a master’s in education at Harvard. His résumé says that he and Landau organized an international exhibition of computer art at the Montreal Museum of Fine Arts in 1966 and built an educational computer game together at the Film Board and at Harvard.

Remy Landau had worked as a programmer at Bell Canada in the mid-1960s; a court later found that a UNIVAC Canada consultant had worked with him there “five or six years” before 1972. Bell Canada was also where the idea for LANPAR came from. Its plant budgeting forms had some 2,000 cells, and every change to a form had to go through programmers. Pardo recalled thinking that a manager at Bell Canada should not have to wait for a programmer to change a budgeting form: the manager should be able to type the formulas in whatever order they came to mind and let the computer sort out which to calculate first.

LANPAR

The two wrote the program in the summer of 1969. They called it LANPAR, for “LANguage for Programming Arrays at Random”, a name that also combines the first syllables of their surnames. It was written in FORTRAN and ran on a General Electric 425, one of GE’s 400-series time-sharing machines; the Lanpar company later licensed it on GE and Honeywell time-sharing services.

The problem it solved can be seen in the example the Patent Office’s Board of Appeals later used against it. A user supplies values for a, b, c and d, then writes A = X + Y, then X = a + b, then Y = c + d. A computer running the lines in order fails at the second one, because X and Y do not exist yet. LANPAR scanned the formulas, marked those whose inputs were all known as “defined”, executed them, and repeated the pass on the rest until every formula had been defined and executed. The user could write the model in any order, including formulas that referred forward to cells not yet computed. Pardo called this forward referencing; the later spreadsheet term is natural order recalculation. VisiCalc in 1979 still recalculated row by row or column by column, so a user whose sheet contained forward references had to recalculate until the numbers stopped changing.

LANPAR was sold to the plant budgeting divisions of Bell Canada, AT&T and the 18 operating telephone companies of the Bell System, and to General Motors in Warren, Michigan. In early 1972 Pardo and Landau adapted it for Ohio Bell, which used a Xerox Sigma 7 at Datalogics in Cleveland; a Datalogics programmer spent several days in March 1972 helping Landau move the program across. Their company, Lanpar Ltd., also sold a related program called OUTCOM.

The Application

Pardo and Landau filed their patent application, “A Process and Apparatus for Converting a Source Program into an Object Program”, on 12 August 1970. The claims described the compiler step: examine each formula, execute those that can be designated as defined, and repeat until all have been executed, “regardless of the sequence in which said formulas are presented”. The application contained a flow chart of the process but no program code.

In July 1971 the examiner rejected it for inadequate disclosure. Pardo swore an affidavit that the flow chart was enough for any competent programmer. The examiner’s final rejection of February 1972 answered that “an affidavit by applicant is self serving and therefore has very little probative value”. The inventors’ patent attorney, R. Gale Rhodes of Newark, New Jersey, then proposed affidavits from three outsiders of different skill levels, and asked Pardo and Landau to find them. They chose Robert Bullen, a Bell Canada supervisor of programmers; David Cikra, a programmer at Datalogics; and Peter Jones, whom they presented as a “computer scientist or compiler writer”. Bullen and Jones swore that a programmer could have written the program from the application alone. Cikra swore that he had actually done so. The affidavits went in on 29 June 1972, and on 28 July 1972 the examiner declared all claims allowable.

Benson

The formal notice of allowance never came. On 20 November 1972 the Supreme Court decided Gottschalk v. Benson, holding that a method for converting binary-coded decimal numbers into binary could not be patented because it was a mathematical algorithm. On 5 March 1973 the Patent Office withdrew its approval and rejected every claim as unpatentable subject matter. The Board of Appeals agreed, reading the inventors’ own use of the word “algorithm” in the specification as an admission. It also added a new rejection for obviousness, with the A = X + Y example and no prior art cited.

Pardo and Landau took the case to the Court of Customs and Patent Appeals without a lawyer. On 5 August 1982 the court reversed the board on both grounds. Judge Miller’s opinion noted that they had used the word “algorithm” more than two years before Benson gave it a legal meaning, and that it found no mathematical formula or calculation recited in their claims, only steps for controlling the order in which a computer executes a program. On obviousness he recorded the appellants’ point that computers had been in general use as sequential processors for 18 years before 1970, “and yet the PTO has been unable to cite even one reference against their claims.” The skill of a 1970 programmer, the court said, was not something it could take judicial notice of “no matter how simple a claimed invention may seem in hindsight.”

US patent 4,398,249 was issued on 9 August 1983, thirteen years after filing. Pardo and Landau assigned it to Forward Reference Systems Ltd. of Downsview, Ontario, a company they controlled. It ran for seventeen years from issue, until August 2000. By 1983 VisiCalc, SuperCalc and Lotus 1-2-3 were all on the market.

Lanpar the Company

Lanpar Ltd., founded in Markham, Ontario in 1970, did not stay in software. Renamed Lanpar Technologies in 1976, it became by the early 1980s Canada’s largest independent distributor of computer terminals, selling, renting and servicing Digital Equipment Corporation terminals and printers. It distributed the Osborne 1 through a subsidiary, Osborne Canada, in 1982 and 1983, and in 1986 sold its own IBM XT-compatible, the LANPAR PC. Pardo, its chief executive, says he grew its revenue to $22 million between 1976 and 1982, followed by a $13 million share offering. The company turned to PC networking products in the late 1980s and was dissolved in January 2003.

Refac v. Lotus

Before any suit was filed, Forward Reference Systems made a deal with Refac International of New York, a company that licensed and litigated patents. Refac received a five-percent interest in the patent in return for promising to sue at least two infringers within a month, and the two firms would split the proceeds equally. On 26 July 1989 Refac sued six software publishers, among them Lotus, Microsoft and Ashton-Tate. The League for Programming Freedom used the case in its 1991 pamphlet against software patents: “Refac is demanding five percent of sales of all major spread-sheet programs.” The action was dismissed against every defendant except Lotus.

Lotus did not attack the invention. It argued that the patent had been obtained by deceiving the examiner in 1972, and in July 1993 the court held a bench trial on that question alone. Judge Sonia Sotomayor of the Southern District of New York decided it on 18 April 1995. The trial showed that none of the three “disinterested” affiants had been a stranger to LANPAR. Bullen had worked with Landau at Bell Canada and had been told the program was in use there. Cikra had helped Landau convert LANPAR for Ohio Bell less than three months before he swore that he had written a program from the application alone. Jones had worked at Lanpar for up to eight weeks in January and February 1972, had been taught the program’s internal logic by Landau, had read its source code and flow charts, and had drafted the introduction to the LANPAR Program Logic Manual. Landau conceded that after the two inventors, Jones “was the person who had the next best knowledge of [the P]rogram and how it worked.” Asked how he had prepared his affidavit, Jones said he had recognized the application’s flow chart “as being essentially the same one that Mr. Landau had shown me” in a training session.

Sotomayor found that Bullen’s and Cikra’s omissions were not proven to be intentional. Jones’s was. Rhodes testified that he had asked about the affiants’ backgrounds and would have dropped Jones had he known; Pardo and Landau said they had told him. “I do not credit this claim,” Sotomayor wrote, and concluded that the inventors had “purposely kept Jones’s employment history from Rhodes.” She held the patent unenforceable for inequitable conduct.

The Federal Circuit affirmed on 26 April 1996. Refac had argued that Jones’s affidavit could not have misled anyone, since it offered only opinion and so, under the examiners’ own manual, carried no weight. Judge Alan Lourie answered that arguing that “an affidavit submitted to persuade was defective as presenting only opinion, not fact” qualified “only for a chutzpah award”, not a reversal. He conceded that losing a patent over a gap in one witness’s employment history “might thus seem to be unduly severe”, but the trial court’s findings on intent were not clearly wrong.

Dead End

The patent failed as a way to be paid for the idea, for reasons that had nothing to do with whether the idea was new. It had to wait thirteen years to issue, in which time the spreadsheet had become a mass-market product that its inventors had no part in. The only enforcement attempt started in 1989, six years into the patent’s life, and it was lost on the conduct of the application rather than on its merits. The deciding facts dated from the spring of 1972, when the inventors needed three independent experts quickly and picked three people who already knew their program.

The 1982 decision had a longer life. In re Pardo applied the court’s two-step test for “mathematical algorithm” claims in the form the Supreme Court’s Diamond v. Diehr (1981) had just confirmed, and held that a method for controlling how a computer executes programs was not a mathematical algorithm at all. The software-patent disputes that followed are described in The Patent Wars. On the spreadsheet’s own history, LANPAR sits between Richard Mattessich’s 1961 budget matrices and the AutoPlan and AutoTab tools of 1968 on one side and VisiCalc on the other; see The Spreadsheet Revolution and Dan Bricklin and VisiCalc.

After Lanpar

Pardo kept founding companies: a digital image library on CD-ROM in 1991, a battery-capacitor venture called Mega-C Power in 2001, water purification, LED lighting, and from 2011 Brainwave Research Corporation. Mega-C ended in bankruptcy and before the Ontario Securities Commission. In a settlement approved on 17 August 2012, Pardo agreed that he had helped transfer Mega-C shares to more than 1,000 investors without registration or a prospectus, telling some of them that the shares would soon be listed on an exchange and that the company would buy them back on request; Mega-C had gone into Chapter 11 in 2004. He admitted breaching the Ontario Securities Act, was reprimanded, and was barred from trading securities and from serving as a director or officer of most companies for two years. He stated that he had relied on informal legal advice and had not been diligent enough in obtaining and reading it.

Landau left almost no public record after the patent case, in which he and Pardo were described as the Canadian citizens controlling Forward Reference Systems.

📚 Sources