FORTRANSIT -- Making FORTRAN a Winner

When I started the PRINT I project I had no thoughts of universal or standard programming languages. But that project was done in the same room with the original FORTRAN effort (not the Langdon Hotel of later stories, but a large room in an annex south of the main IBM buildingat 590 Madison). While not working on my own assignment I got the best outsider's view of that work that anyone ever had. And I could see that it was a big advance over what I was doing. It should have been, with about 10 times the effort.

Then those thoughts started to come. Because of my great familiarity with the IBM 650s (3 of them) at Lockheed Missiles and Space Division, I started to feel that it deserved a similar capability. Especially because there were so many of them out there, due in large part to IBM's generous discount to schools and universities.

Within IBM one got to know a great deal about what was happening in the outside world. One thing I knew was that another algebraic compiler [1] was being built under the direction of Prof. Alan Perlis, who had moved from Purdue to Carnegie Tech. It was called IT, and deserved fame more from its novel construction than from ease-of-use language.

I also knew that the language SOAP (Symbolic Optimal Assembly Language), written by Stan Poley of IBM, was the assembler of choice for almost every 650 user.

Everyone should read Don Knuth's homage to the 650, in that same issue of the Annals of the History of Computing as [5]. He describes his rapture upon first reading Stan Poley's code for SOAP, and the sheer beauty of the work. I was invited to Stan's for dinner one night. Outside on the penthouse-like terrace were two cushions his wife had embroidered. One said "SOAP", the other "650". May every great programmer's wife have the same appreciation and admiration for her husband's work.

I say manpower, although the name FORTRANSIT was created by Florence (Flo) Pessin, in a conversation I had with her. I had always liked puns in creating programming names (CODASYL was my play on the legal term "codicil", but COBOL was not). Flo liked anagrams and acrostics. We agreed that it meant either 1) FORTRAN-S(oap)-IT, or 2) FOR TRANSIT(ion), or 3) FORTRAN's IT (in the verb sense of FORTRANning the IT Compiler).

An overall description of the processor design was given in [2]:

"The FOR TRANSIT (sic) system consists of three major parts:

1. The translator, FOR TRANSIT, which accepts FORTRAN statements and produces corresponding IT statements.
2. The compiler, a modification of IT, which accepts IT statements and compiles 650 instructions in symbolic (SOAP II) language.
3. The assembler, a modified version of SOAP II [3], which produces an optimized machine language program from the symbolic instructions."

Flo argued, and I agreed, that the cascading method was inefficient. Flo went on to create, with others, a real FORTRAN for the various 650 manifestations. But one should never forget the dates that FORTRAN compilers were put in service:

    FORTRAN    (704)     1957 June
    FORTRANSIT (650)     1957 August (just 2 months later)
    FORTRAN    (705)     1958 September
    FORTRAN    (650)     1959 June

The key to getting FORTRANSIT out so fast was first the quality of the builders, and then the fact that SOAP and IT were already operational. It was very successful. It got much more usage than FORTRAN then, largely because the educational discount for 650s meant a larger market to cover. JAN Lee and Tony Pizzarello of Italy were among the many to apprise me of this.

Perhaps 10 times as many people were introduced to FORTRAN, via FORTRANSIT and the 650, as there were via the 704. That's why the title of this piece hints that FORTRANSIT clinched the preeminence of the FORTRAN language.

Certainly not the least accomplishment was proving that the same source language could be used, whether the hardware was decimal or binary in nature! It had never been done before.

Both FORTRAN and FORTRANSIT manuals were interesting from the viewpoint of authorship. I had always maintained within IBM that technical authorship or programs and manuals should be recognized, whereas IBM's policy had always been anonymity. For the PRINT I project for the 705, I got around it by naming a committee, and printing that membership in the manual (because I had used outside contract personnel from CUC). With this precedent, IBM folded. Both the FORTRAN and the FORTRANSIT manuals named names. John Backus was very pleased that this could be done.

In fact, on the 25th anniversary dinner that IBM held in 1982 June for the greater FORTRAN community (right on the dot even to the month) IBM gave us all boxed sets of the original programming manual and the educational manual. Completely printed anew to identical specs. You could say from this that IBM was very grateful to FORTRAN and derivatives.

How I assigned leadership in the project is vague. Dave Hemmes and Flo Pessin may have been joint leaders. If not, Dave was technically in charge. I was in favor of cooperative working rather than formal leadership (and still am).

Dave was ever a "character". While at Marquardt, I hired him from the RAND Corporation. I persuaded him to come to New York for the FORTRANSIT work. He deserves a large chapter of his own. For now I have many an anecdote in another vignette.

Another FORTRAN

FORTRANSIT served good purpose, but I wanted a real FORTRAN. The opportunity came with the GUIDE organization, the commercial counterpart to SHARE.

My records show "705 FORTRAN" for a trip on 1957 Sep 9-16. It was JFK-LAX-SFO-DCA-SFO-LAX-JFK (I forget what was going on in Los Angeles). San Francisco for a day of GUIDE, Washington Wednesday for an IBM seminar run by John McPherson, back to SFO for GUIDE. All on prop planes, with night sleep missed. The cooperative GUIDE FORTRAN project was quite interesting. We borrowed Jim Matheny from Shell, among others. It replaced my PRINT I product (of mid-1956) in 1958 September.

An Apology

I make it now to Flo (Florence H. Pessin). It has to do with the "Stack" or "Cellar" Principle, commonly attributed now to Fritz Bauer and Klaus Samelson, inasmuch as they authored jointly two papers on it [6,7].

My apology to Mrs. Pessin is that:

1. She invented the method independently for FORTRANSIT, sometime in the Spring of 1957, over two years before Bauer and Samelson published.

2. I wasn't checking closely enough to notice that she had made this great breakthrough for IBM.

3. I attended the IFIP Conference in Paris in 1959 June where they first presented it, and either did not hear it or took no note of it.

4. Most shamefully of all, I was the Editor of the Techniques Section of the Communications of the ACM, where their second paper on the method was published the following January, and made no reference to the Pessin work right under my nose!

REFERENCES

1. A.J.Perlis and J.W.Smith, "A Mathematical Language Compiler, in
   Automatic Coding, Franklin Institute Monograph No. 3, 87-102, 1957 Apr
   
2. B.C.Borden, "FORTRANSIT, a Universal Automatic Coding System for the IBM 650", Proc. Canadian Conf. for Computing and Data Processing, 349-359, 1958 Jun 9-10
   
3. SOAP II programmer's reference manual, Form 32-7646.
   
4. R.W.Bemer, "Nearly 650 memories of the 650",
   Annals of the History of Computing 8, No. 1, 66-69, 1986 Jan
   
5. D.A.Hemmes, "FORTRANSIT Recollections",
   Annals of the History of Computing 8, No. 1, 70-73, 1986 Jan
   
6. F.L.Bauer, K.Samelson, "The Cellar Principle for Formula Translation",
   Proc. Intl. Conf. on Information Processing, Paris, 1959
   
7. F.L.Bauer, K.Samelson, "Sequential Formula Translation", Commun. ACM 3,
   No. 1, 1960, 76-83.