Preview Mode Links will not work in preview mode

Aug 29, 2019

FORTRAN Welcome to the History of Computing Podcast, where we explore the history of information technology. Because by understanding the past, we’re better prepared for the innovations of the future! Todays episode is on one of the oldest of the programming languages, FORTRAN - which has influenced most modern languages. We’ll start this story with John Backus. This guy was smart. He went to med school and was diagnosed with a brain tumor. He didn’t like the plate that was left behind in his head. So he designed a new one. He then moved to New York and started to work on radios while attending Columbia for first a bachelor’s degree and then a master’s degree in math. That’s when he ended up arriving at IBM. He walked in one day definitely not wearing the standard IBM suit - and when he said he was a grad student in math they took him upstairs, played a little stump the chump, and hired him on the spot. He had not idea what a programmer was. By 1954 he was a trusted enough resource that he was allowed to start working on a new team, to define a language that could provide a better alternative to writing code in icky assembly language. This was meant to boost sales of the IBM 704 mainframe by making it easier to hire and train new software programmers. That language became FORTRAN, an acronym for Formula Translation. The team was comprised of 10 geniuses. Lois Haibt, probably one of the younger on the team said of this phase: "No one was worried about seeming stupid or possessive of his or her code. We were all just learning together." She built the arithmetic expression analyzer and helped with the first FORTRAN manual, which was released in 1956. Roy Nutt was also on that team. He wrote an assembler for the IBM 704 and was responsible for the format command which managed data as it came in and out of FORTRAN programs. He went on to be a co-founder of Computer Science Corporation, or CSC with Fletcher Jones in 1959, landing a huge contract with Honeywell. CSC grew quickly and went public in the 60s. They continued to prosper until 2017 when they merged with HP Enteprirse services, which had just merged with Silicon Graphics. Today they have a pending merger with Cray. David Sayre was also on that team. He discovered the Sayre crystallography equation, and molter moved on to pioneer electron beam lithography and push the envelope of X-ray microscopy. Harlan Herrick on the team invented the DO and GO TO commands and ran the first working FORTRAN program. Cuthbert Herd was recruited from the Atomic Energy Commission and invented the concept of a general purpose computer. Frances Allen was a math teacher that joined up with the group to help pay off college debts. She would go on to teach Fortran and in 1989 became the first female IBM Fellow Emeritus. Robert Nelson was a cryptographer who handled a lot of the technical typing and designing some of the more sophisticated sections of the compiler. Irving Ziller designed the methods for loops and arrays. Peter Sheridan, aside from having a fantastic mustache, invented much of the compiler code used for decades after. Sheldon Best optimized the use of index registers, along with Richard Goldberg. As Backus would note in his seminal paper, the History Of FORTRAN I, II, and III, the release of FORTRAN in 1957 changed the economics of programming. While still scientific in nature, the appearance of the first true high-level language using the first real compiler meant you didn’t write in machine or assembly, which was hard to teach, hard to program, and hard to debug. Instead, you’d write machine independent code that could perform complex mathematical expressions and once compiled it would run maybe 20% slower, but development was 5 times faster. IBM loved this because customers needed to buy faster computers. But customers had a limit for how much they could spend and the mainframes at the time had a limit for how much they could process. To quote Backus “To this day I believe that our emphasis on object program efficiency rather than on language design was basically correct.” Basically they spent more time making the compiler efficient than they spent developing the programming language itself. As with the Constitution of the United States, simplicity was key. Much of the programming language pieces were designed by Herrick, Ziller, and Backus. The first release of FORTRAN had 32 statements that did things that might sound similar today like PRINT, READ, FORMAT, CONTINUE, GO TO, ASSIGN and of course IF. This was before terminals and disk files so programs were punched into 80 column cards. The first 72 columns were converted into 12 36 bit words. 1-5 were labels for control statements like PRINT, FORMAT, ASSIGN or put a C in column 1 to comment out the code. Column 6 was boolean where a 1 told it a new statement was coming or a 0 continued the statement from the previous card. Columns 7 through 72 were the statement, which ignored whitespace, and the other columns were ignored. FORTRAN II came onto the scene very shortly thereafter in 1958 and the SUBROUTINE, FUNCTION, END, CALL, RETURN, and COMMON statements were added. COMMON was important because it gave us global variables. FORTRAN III came in 1958 as well but was only available for specific computers and never shipped. 1401 FORTRAN then came for the 1401 mainframe. The compiler ran from tape and kept the whole program in memory, allowing for faster runtime. FORTRAN IV came in the early 60s and brought us into the era of the System/360. Here, we got booleans, logical IF instead of that used in arithmetic, the LOGICAL data type, and then came one of the most important versions, FORTRAN 66 - which merged all those dialects from IV into not quite a new version. Here, ANSI, or the American National Standards Institute stepped in and started to standardize. We sill use DO for loops, and every language has its own end of file statement, commenting structures, and logical IFs. Once things get standardized, they move slower. Especially where compiler theory is concerned. Dialects had emerged but FORTRAN 66 stayed put for 11 years. In 1968, the authors of BASIC were already calling FORTRAN old fashioned. A new version was started in 66 but wasn’t completed until 1977 and formally approved in 1978. Here, we got END IF statements, the ever so important ELSE, with new types of I/O we also got OPEN and CLOSE, and persistent variable controls with SAVE. The Department of Defense also insisted on lexical comparison strings. And we actually removed things, which these days we call DEPRECATE. 77 also gave us new error handling methods, and programmatic ways to manage really big programs (because over the last 15 years some had grown pretty substantial in size). The next update took even longer. While FORTRAN 90 was released in 1991, we learned some FORTRAN 77 in classes at the University of Georgia. Fortran 90 changed the capitalization so you weren’t yelling at people and added recursion, pointers, developer-controlled data types, object code for parallelization, better argument passing, 31 character identifiers, CASE, WHERE, and SELeCT statements, operator overloading, inline commenting, modules, POINTERs (however Ken Thompson felt about those didn’t matter ‘cause he had long hair and a beard), dynamic memory allocation (malloc errors woohoo), END DO statements for loop terminations, and much more. They also deprecated arithmetic IF statements, PAUSE statements, branching END IF, the ASSIGN statement, statement functions, and a few others. Fortran 95 was a small revision, adding FORALL and ELEMENTAL procedures, as well as NULL pointers. But FORTRAN was not on the minds of many outside of the scientific communities. 1995 is an important year in computing. Mainframes hadn’t been a thing for awhile. The Mac languished in the clone era just as Windows 95 had brought Microsoft to a place of parity with the Mac OS. The web was just starting to pop. The browser wars between Netscape and Microsoft were starting to heat up. C++ turned 10 years old. We got Voice over IP, HTML 2.0, PHP, Perl 5, the ATX mother board, Windows NT, the Opera browser, the card format, CD readers that cost less than a grand, the Pentium Pro, Java, JavaScript, SSL, the breakup of AT&T, IBM’s DEEP BLUE, WebTV, Palm Pilot, CPAN,, the first Wiki, Cygwin, the Jazz drive, Firewire, Ruby, and NumPy kickstarted the modern machine learning era. Oh and Craigslist, Yahoo!, eBay, and Audible was also established that year but they weren’t owned by Amazon just yet. Even at IBM, they were buys buying Lotus and trying to figure out how they were going to beat Kasparov with Deep Blue. Hackers came out that year, and they were probably also trying to change their passwords from god. With all of this rapid innovation popping in a single year it’s no wonder there was a backlash as can be seen in The Net, with Sandra Bullock, also from 1995. And as though they needed even more of a kick that this mainframe stuff was donezo, Konrad Zuse passed away in 1995. I was still in IT at the university watching all of this. Sometimes I wonder if it’s good or bad that I wasn’t 2 or 3 years older… Point of all of this is that many didn’t notice when Fortran continued on becoming more of a niche language. At this point, programming wasn’t just for math. Fortran 2003 brought object oriented enhancements, polymorphism, and interoperability with C. Fortran 2008 came and then Fortran 2018. Yes, you can still find good jobs in Fortran. Or COBOL for that matter. Fortran leaves behind a legacy (and a lot of legacy code) that established many of the control statements and structures we use today. Much as Grace Hopper pioneered the idea of a compiler, FORTRAN really took that concept and put it to the masses, or at least the masses of programmers of the day. John Backus and that team of 10 programmers increased the productivity of people who wrote programs by 20 fold in just a few years. These types of productivity gains are rare. You have the assembly line, the gutenberg press, the cotton gin, the spinning Jenny, the watt steam engine, and really because of the derivative works that resulted from all that compiled code from all those mainframes and since, you can credit that young, diverse, and brilliant team at IBM for kickstarting the golden age of the mainframe. Imagine if you will, Backus walks into IBM and they said “sorry, we don’t have any headcount on our team.” You always make room for brilliant humans. Grace Hopper’s dream would have resulted in COBOL, but without the might of IBM behind it, we might still be writing apps in machine language. Backus didn’t fit in with the corporate culture at IBM. He rarely wore suits in an era where suit makers in Armonk were probably doing as well as senior management. They took a chance on a brilliant person. And they assembled a diverse team of brilliant people who weren’t territorial or possessive, a team who authentically just wanted to learn. And sometimes that kind of a team lucks up and change sthe world. Who do you want to take a chance on? Mull over that until the next episode. Thank you so very much for tuning into another episode of the History of Computing Podcast. We’re lucky to have you. Have a great day! The History of FORTRAN I, II, and III ::